Capacity and Coverage Calculation Model for the UMTS

© 2007 Journal of Computer Science Vol 1, No 1, June 2007, pp 05-10 Faculty of Computer Science ISSN 1994-6244

Capacity and Coverage Calculation Model for the UMTS M. Shamim Kaiser1, Md. Ezharul Islam2 and Md. Raihan Jamil1

ABSTRACT This paper deals with the coverage and capacity issues of upcoming 3G (Third Generation) or UMTS (Universal Mobile Communication Telecommunication System) and discuss the hurdles to implement the UMTS or 3G system in the place of existing 2G radio network. The radio interface of UMTS operating in the WCDMA (Wideband Code Division Multiple Access) FDD (Frequency Division Duplex) mode required the updating of 2G radio network. In this work an attempt has been made to find the total number of users that acquired the maximum adequate signal strength in a cell. Finally Optimization of coverage and capacity calculations for the UMTS has also been done. Keywords: UMTS, Network Planning, Capacity and Coverage, Propagation Model and WCDMA

I. INTRODUCTION 3G radio network, is also referred to as UMTS, is adopted to fulfill the user requirement for innovative services such as enhanced and multimedia messaging through high-speed data channels [1]. Wide-band CDMA (WCDMA), CDMA2000 and Time-Division Synchronous CDMA (TD-SCDMA) are the three standards accepted by ITU-T (International Telecommunication Union, Telecommunication Standard). Owing to the high costs and the scarcity of radio resources, an accurate and efficient mobile network planning procedure is required. The objective of network planning is to maximize the coverage, capacity and the quality of service (QoS) [2]. The planning of UMTS (or 3G) is very different from the 2G which adopts a time division multiple access (TDMA) technology. TDMA works by dividing a radio frequency into time slots and then allocating slots to multiple calls. In this way, a single frequency can support multiple, simultaneous data channels. 2G planning can be divided into a coverage planning phase and a frequency planning phase which are driven by coverage and capacity criteria respectively [3]. In the coverage planning phase, base stations (BSs) are placed so that the signal strength is high enough in the area to be served. In the frequency-planning phase, each BS is assigned a number of channels, and then network operator takes into account the traffic load, the level of services measures as the signal-to-interference ratio (SIR) as well. However, in the 3G planning, since all carriers in the network use the same frequency range, frequency planning is not required. Furthermore, coverage and capacity planning should be performed in tandem since capacity requirement and traffic distribution influence the coverage. UMTS or 3G mobile system has a significant impact not only on the RF (radio frequency) network, but also on the core network architecture. Care must be taken to allow current GSM (global system mobile) operators to protect their infrastructure investments when their networks are upgraded to support UMTS. This paper focuses on the differences between GSM radio system planning and UMTS radio system planning. UMTS uses WCDMA (wideband code division multiple access) as the radio transmission technology. It is claimed that TDMA (time division multiple access) RF planning is much more difficult than WCDMA-based systems. This is true, in part, because of the interference issues. However, UMTS serves users with various demands, and many aspects of planning are more closely interrelated to each other in UMTS planning than they are in GSM planning [1-5]. The major differences in the UMTS radio system planning process occur in coverage and capacity planning. In GSM, coverage is planned separately after the network is dimensioned (based on the market study), and capacity and frequency are planned in tandem. In UMTS, coverage and capacity are planned at the same time, because capacity requirements and traffic distribution influence coverage. Frequency and code can be planned separately. On the other hand, the wideband nature of WCDMA technology (5 MHz) compared with GSM (200 kHz) imposes new criteria in modeling the propagation environments. This paper outlines the challenges and solutions for planning and optimizing UMTS networks with respect to radio interface. WCDMA air interface specifications 1. 2.

Telecommunication Engineering. Training and Instrumentation Division, University Grants Commission of Bangladesh, Agargaon, Dhaka-1207, Bangladesh. e-mail: [email protected] Computer Science and Engineering. Training and Instrumentation Division, University Grants Commission of Bangladesh, Agargaon, Dhaka-1207, Bangladesh. e-mail: [email protected]

Md. Amiruzzaman et al.

and their implications on transmission channel behavior and modeling are described first. Next, solutions are provided for system design, including coverage, capacity, code, and frequency planning. The analysis captures both design processes and engineering calculations. The critical optimization and monitoring of WCDMA network performance are then discussed. II. UMTS ARCHITECTURE The UMTS network architecture is depicted in Figure 1. The core network handles call control and mobility management functionalities, while the UTRAN (UMTS terrestrial radio access network) manages the radio packet transmission and resource management. Packet routing and transfer within the core network are supported by definition of new logical network nodes called GGSN (gateway GPRS [general packet radio system] support node) and SGSN (serving GPRS support node). The GGSN is basically a packet router with additional mobility management features, and it connects with various network elements through standardized interfaces. The GGSN acts as a physical interface to the external packet data networks. The SGSN handles packet delivery to and from mobile terminals. Each SGSN is responsible for delivering packets to the terminals within its service area. GGSN and SGSN are capable of supporting terminal data rates up to 2 Mbps. A UTRAN consists of one or more RNSs (radio network subsystems), which in turn consist of base stations (Node Bs) and RNCs (radio network controllers). The RNS performs all of the radio resource and air interface management functionalities. The UMTS network architecture inherits most of its structure from the GSM model in the UTRAN.

Figure 1: UMTS Architecture [4-6] III. UMTS RADIO NETWORK PLANNING The UMTS radio network planning process can be divided into three parts; the initial phase (also called system dimension); detailed planning phase; and the optimization and monitoring phase. Each of the phases requires additional considerations such as propagation measurements, traffic demand measurements and so on. The process of the UMTS radio network planning is illustrated in Figure 2.

It is observed that the whole process needs to take into account coverage and capacity planning. In a cellular system where the entire air interface connections operate on the same carrier, the number of simultaneous users is directly influences the total noise of receivers. It is well known that the coverage of the cell has an inverse relationship with the user capacity of the same cell. An increase in the number of active users in the cell causes the total interference seen at the receiver to increase. This causes an increase in the power required to be received from each user. This is due to the fact that each user has to maintain a certain SIR at the receiver for satisfactory performance. For a given maximum allowable transmission power, an increase in the required power reception will result in a decrease in the maximum distance a mobile can be from the base station thereby reducing the coverage. The up-link and the ideal situation of the WCDMA technology are considered. That is, It is assumed that the coverage is limited by the maximum transmit power at the mobile and no blocking and outage take place in the cellular system since WCDMA technology can provide the enough codes for the new call to the cell in the ideal situation.

12

Journal of Computer Science, Volume1, Number 1, June 2007, ISSN : 1994‐6244

Embedded Wireless Network-Engine for VOD Sharing between Resource Constrained Devices

A. Initial Model for Capacity Calculation In order to calculate the maximum number of users in a particular cell the following assumptions are made: 1. No inter-cell and intra-cell interference is present with in the cell. 2. All signals arrive at the base station with equal power. 3. Un-limited number of spreading codes are available. The actual capacity of a WCDMA cell depends on many different factors, such as power control accuracy, interference power. If there are N users in a cell and the signal is denoted by S then the interference can be calculated as I = ( N − 1) S + η , where η is the thermal noise. Hence the SIR is given by

SIR =

S 1 (1) = ( N − 1 )S + η ( N − 1 ) + η / S

The value of the interference is based upon the three assumptions mentioned above and is not realistic. The inter-cell interference is attributed to mobile stations and base stations in one cell affecting the operations in another cell. Intra-cell interference is attributed to the effect mobile stations and has the transmitted signal of other mobile stations within the same cell. It is a simple step to change the interference value to consider the other two factors though in this paper it would only serve to complicate the formulae. Suppose the digital demodulator for each user can operate against the noise at energy per bit-to noise power density level is given by E b / I o , where Here E b = S / R and I o = I / W (2) where, W is the chip rate and R is rate of data communication. Hence using equations (1) and (2) gives

(N − 1) =

W/R η − Eb / I o S

(3)

In particular, if the user is not speaking during part of the conversation, the output of the coder is lowered to prevent the power from being transmitted unnecessarily. For a uniform population, this reduces the average signal power of all users and consequently the interference received by each user. The capacity is then increased proportional to this overall rate reduction. Speech statistics shows that a user in a conversation typically speaks close to 38% of the time using DTX (discontinuous transmission) and DRX (discontinuous reception) method controlled by transcoder (TC) as it has VAD (voice activity detection). The effects of voice activity and sectorization should be considered. This results in an increase in the Eb / I o by a voice activity gain factor, G v . Similarly, the sectionzation gain factor G s also increases the Eb / I o and the equation for the effective number of users, N ε , becomes

(N ε

⎛ W /R η⎞ − 1) = Gv G s ⎜⎜ − ⎟⎟ E / I ⎝ b o S⎠

(4)

Using speech statistics we can set G v =2.63 [1] and based upon the usual three sectors for a cell,

G s =2.63.

W/R is Eb / I o

typically a value between 20 to 100.. Finally, for a cellular system in which all users in all cells employ the common spectral allocation of W, we must evaluate the interference. Normally the total interference equal the amount of interference from other cells and interference from given cells. Mathematically,

Due to the interference, the actual numbers of user will decrease from ( N ε − 1 ) to ( N φ − 1 ) [7], the equation (4) will be modified into

⎛ W /R η ⎞ 1 − ⎟⎟ ( N φ − 1) = Gv G s ⎜⎜ ⎝ Eb / I o S ⎠ 1 + f

(5)

Normally, the total interference from users in all other cells equals approximately three-fifths of that caused by all users in the


13


given cell [9]. That is to say,

1 + f ≈ 1.6 . However we note that equation (5) gives a poor estimate to N φ due to the

many assumptions in the model. It does not reflect the relationship between the capacity and the coverage. B. Coverage versus Capacity The analysis in the above the capacity calculation can be isolated from coverage. However, in the 3G networks, it is important to derive the relationship between the coverage and capacity. The up-link and the ideal situation of the WCDMA technology are considered. That is to say, we will assume the coverage is limited by the maximum transmission power at the mobile and neither blocking nor outage takes place in the cellular system. Based on Equation (5), we can understand the performance of a WCDMA network by developing a simple expression for the ratio E b / I o as follows.

Eb S/R = I o ( N φ − 1 )S ( 1 + f )Gv G s + η W

(6)

Solving the equation (6) for S gives

S=

Eb Rη Io E W − ( b R( N φ − 1 )( 1 + f ) / G v G s Io

(7)

This equation can be used to demonstrate the dependency of S on the user data rates, R the voice and section gain G v , G s , respectively, and the total number of active users in a cell, N φ . We focus on the coverage by user 1 when the number of users in the cell is N φ . Let r be the distance of user 1 from the base station. The received power at the base station from mobile user l, S, is given by

S = S 1 − P( d ) − Z (8) where, S 1 The transmission power of the user, P( d ) is the propagation loss at distance d from the MS to BS, Z The shadow fading The propagation loss is usually calculated by using propagation models. The propagation model that is most commonly used is the Okumura-Hata. The model is developed by combining propagation theory and extensive measurement campaigns. The model considers several parameters such as effective antenna height, terrain type, terrain height, frequency, and so on. For details of the Okumura-Hata model the reader is referred to [2, 3]. The model predicts

(9) Where A, B are constants defined by the frequency and

f is the frequency. a( hms ) is a function that depends upon the height of the mobile station that is particular to the environment. C m is the area type correction factor. Equation (9) shows

the relationship between the propagation loss and the distance and can be expressed as

P( d ) = p 1 + p 2 log 10 d

(9)

Based on equations (7), (8), (9) the relationship among the received power, number of users, the coverage area, the gain of voice and gain of sectionzation is given by=

Eb Rη Io E W − ( b R( N φ − 1 )( 1 + f ) / G v G s Io

(10)

14



IV. UMTS (WCDMA) TRANSMITTER, RECEIVER, AND CHANNEL PARAMETERS Link budget is the main tool for planning WCDMA coverage. WCDMA link budget, takes into account the base station equipment and the base station antenna line configuration. A typical link budget for WCDMA is presented in Table 1. The link budget in Table 1 is divided into five parts. In general information, the frequency band, chip rate, and temperature constant are given. In service information, the bit rates and loads for uplink and downlink are defined. Receiving end and transmitting end define the radio links in the uplink and downlink directions, respectively. Finally, isotropic path loss is defined as the maximum expected path loss between the receiver and transmitter [6]. WCDMA Coverage and Capacity Planning Coverage and capacity planning in WCDMA are interrelated. In low traffic areas, WCDMA planning is quite similar to GSM planning, because the load does not have a great impact on coverage. Of course, many details differ between the systems, but the main principles can be applied to both. In high traffic areas, unlike for GSM, there is no clear split between coverage, interference, and capacity planning of WCDMA. A. Coverage Planning The propagation predictions for WCDMA require the same planning phases as in GSM. First, the base station configuration and the link budget have to be defined. Also, the coverage threshold has to be well defined to exceed the required quality criteria but avoid unnecessary additional investments for the radio network elements. Moreover, the capacity targets and forecasts have to be well known at this phase because they have a strong effect on the base station coverage area. When the base station antenna height, coverage threshold, and capacity requirements are defined and the base station configuration is clarified in the link budget calculations, the actual propagation predictions process can start. Propagation measurements can be performed to fine tune the propagation prediction model. When the prediction model is tuned, the final base station parameters can be used to make the propagation predictions. Optimized base station parameters can be evaluated when the planning criteria are defined. This planning threshold means that agreement must be reached on the reasonable QoS level required for the different geographical locations. The threshold also depends on whether the service has to be extended inside vehicles and buildings in different areas. The planning threshold is defined in GSM by starting from the mobile station sensitivity and by adding the required clutter planning margins to the sensitivity value in each particular planning terrain bin. B. Capacity Planning WCDMA capacity planning is directly related to the link budget and, thus, to the base station coverage area. In the link budget in Table 1, only one type of service (64/144 kbps data transmission) was introduced, and the base station coverage was fixed for this service. It is possible to have any type of service between the voice calls and 2 Mbps data traffic in the WCDMA base station. This means that the base station coverage area is different for different users. (Figure 3) General Value information Frequency 2100MHz Chip rate 3.84Mcpc Temperature 0 oC Service Urban information Uplink Downlink Load 30% 50% Bit rate 64Kbps 144Kbps Transmitting end Tx power/ connection Tx power Tx antenna gain Peak EIRP Isotropic path loss Receiving end Thermal noise density Rx noise figure Rx noise density

Urban Uplink Downlink 0.126W 1W 21dBm 30dBm 0 18dB 21dBm 44dBm 154.32dB 156.34dB Urban

Uplink

Downlink

173.93dBm/Hz

173.93dBm/Hz

3dB 170.93dBm/Hz

6dB 167.93dBm/Hz


15


Noise power Interference margin Receiving end Receiver interference power Processing gain Required Eb/No Receiver sensitivity Cable Loss Rx antenna gain Soft handoff diversity gain Required signal power Field strength

-105.09dBm

-102.09dBm

1.55dB

3.01dB Urban

Uplink

Downlink

-108.54dBm

-102.09dBm

17.78dB 5dB

14.26dB 4dB

-116.32dBm

-109.34 dBm

4dBm 18dB

0 0

3dB

3dB

-133.32dBm

-122.34dBm

10.32dB µ V/m

31.31dB µ V/m

Table1: Typical link budget for UMTS cell

Figure 3: Relative cell range and cell area versus user bit rate using WCDMA (cell ranges calculated by Okumura-Hata model at antenna height of 25m) Basically, the question is about the spreading power factor, SPF, which varies significantly when comparing the 12.2 kbps voice call (SPF = 25 dB) and 2 Mbps data transmission (PG = 2.8 dB) connections. In the uplink direction, the main objective in capacity planning is to limit interference from the other cells to an acceptable level. Network planning can increase the uplink load by reducing other cell interference. This can be achieved by obstacles to block the interfering cells. Also, downtilting is a very useful tool in limiting interference. In the downlink direction, two aspects should be considered: the interference from other cells and the power of the base station. The load equation for the downlink is similar to the equation for the uplink. However, in the downlink there is a new parameter called orthogonality and users are much more orthogonal compared with uplink. V. CONCLUSION AND FUTURE WORKS The efficient network planning is a vital aspect of UMTS networks. 2G and 3G networks are different from each other owing to the different levels of service offered. This article includes the optimization of capacity, coverage and quality of service. The equations derived in this work can be used in cellular system planning to set hard limits on the maximum number of users that can be admitted into the cell. Coverage and capacity in WCDMA are heavily coupled and can not be planned separately. The relationship given in (10) describing the relationship between coverage, capacity and data rates is very useful in planning the networks. However, there are shortcomings in the analysis and modeling provided herein. The modeling is largely based upon an ideal situation, wherein no call will be blocked. While WCDMA technology can provide enough codes

16



for the mobile terminals to assure the ideal situation these are not used in practice. In application the network operators take into account the cost of equipment and the network load. They use Erlang B to assure certain grade of service. To overcome the shortcomings of the modeling presented herein, in future work the outage probability and the loading control in the planning process are need. An analysis considering outage probability and cell loading will give a more accurate relationship between the coverage and capacity. VI. NOMENCLATURE AND ELABORATION MSC AuC HLR PLMN RAN RNS N

Eb / I o R Gv S

Mobile Switching Center Authentication Center Home Location Resister Public Land Mobile Network Radio Access Network Radio Network Subsystem Users in a cell Thermal Noise bit-to noise power density

EIR GGSN VLR RNC SGSN UTRAN S SIR

Data rate Voice gain

G

Received power from mobile user

S1

W

Gs

Equipment Identity Register Gateway GPRS Support Node Visitor Location Resister Radio Network Controller Service GPRS Support Node UMTS Terrestrial RAN Signal Signal-to-interference ratio chip rate Gain factor Section gain The transmission power of the user

VII. REFERENCES [01] Harri H., Antti T., 2005, "WCDMA for UMTS Radio Access for Third Generation Mobile Communication". Wiley, pp. 120-143. [02] Jaana L., Achim W., Tomas N., 2002, "Radio Network Planning and Optimization for UMTS", Wiley, pp. 38-54 [03] Wacker A., Laiho-Steffens J., 1999, "Static Simulator for Studying WCDMA Radio Network Planning Issues", Proc.1EEEVICSpwgCarf. [04] Hoppe R., Buddendick G., 2001, "Dynamic Simulator for Studying WCDMA for Radio Network Performance", Proc. IEEE International Symposium on Personal, Indoor and Mobile Radio Communications, pp: 32-37. [05] Jukka L., Matti M., 2001, "Radio Interference System Planning for GSM/GPRS/UMTS", Kluwer Academic Publisher, pp.42-48. [06] Venugopal V., 1999, "The Coverage-Capacity Tradeoff in Cellular WCDMA Systems", IEEE Transactions on Vehicular Technology. Vol.48, No.5. [07] Dinan E., December 2006, “UMTS Radio Interface System”, Bechtel Telecommunications Technical Journal, Vol.1(1), pp:10-14. [08] Gilhousen K. S., 1991, "On the Capacity of a Cellular WCDMA System", IEEE Transactions on Vehicular Technology. Vol.40, No.2. [09] Leung K. K., Mastey W. A., 1994," Traffic Modelling for Wireless Communication Networks" IEEE Journal on Selected Area in Communication. Vol. 12, No. 8. [10] Kyoung K, 2000, "Handbook of WCDMA System Design, Engineering, and Optimization" Prentice Hall.


17



Embedded Wireless Network-Engine for VOD Sharing between Resource Constrained Devices Md. Amiruzzaman1, Md. Mustafizur Rahman2 and Md. Mahbubul Alam Joarder3 ABSTRACT PDA, Smart Phone or other resource constraint devices has limited wireless communication capability and memory to serve the video on demand service. In this work, we propose a protocol to support and Video on Demand service on resource constrained devices and share among them. Designed protocol was successfully implemented and experimented on PDAs. Experimental results show that proposed protocol could successfully serve the video on demand using low rate wireless communication and share the video with similar devices efficiently. Keywords: Multimedia Stream, Embedded Network Engine, RTCP Multicast, Video-on-Demand

I.

INTRODUCTION

Resource constraint devices such as Personal Device Assistants (PDA) or Smart Phones became popular day by day and opened a new research area for the electronics and telecom industries. These devices are accepted by the people as they can access to the digital world from anywhere they can be carried easily1. Despite of their constraints in resources, such devices are getting more and more advanced, including the ability to act as hard drives on computers via USB cables, and even merging with Smart-Phones. Now-a-days they offer wireless networking, remote storage, data compression, transferring files including the multimedia. However, audio and video streaming services could not be implemented due resource limitation and limited network capability. As a result, the PDA or Smart-Phone users cannot access and share video on demand (VOD) and online radio services2. Modern PDAs and Smart Phones support both IEEE 802.11 (WLAN) and IEEE 802.16e (WiMax and WiBro) protocols with limited functionality. Our work mainly focuses on Video on Demand sharing over Wireless Broadband (WiBro) transmission between PDAs, and also covers the client-server communications between PDA and Server. Proposed network engine considers the memory capacity of PDA, transmission rate, bandwidth, connection procedure and user interaction. It is implemented on HP iPAQ PDAs and experimented with some real video clip.

Figure 1. Proposed Video Sharing between PDAs Advances in computing and communication in recent years have made video on demand sharing feasible between resourceful devices like computer systems or digital media devices. In VOD systems, movie database is stored in a central

server or in a group of interconnected servers. The video server is connected with high-capacity network interfaces through which local distribution centers (hubs) receive the stream data and broadcast to the households. Such systems can share video stream between them by high-speed networks and buffering in larger memories. However, similar VOD sharing between PDAs or Smart-Phone could not be offered due to PDA-to-PDA low rate wireless

18


Md. Saiful Islam et al.

transmission, insufficient memory for buffering, and absence of efficient communication tool. In the next section, we discuss some of the related works and in section III we present the proposed framework for VOD service in PDAs and sharing. Section IV describes the implementation issues of the VOD sharing followed by experimental results in Section V. Section VI concludes the paper.

II.

LITERATURE REVIEW

There is no remarkable work found on sharing video on demand for resource constraint device especially for PDA and Smart Phones. However, several works have been done on PC based video streaming engine and some of them can be considered in designing VOD service sharing in resource constraint devices. Jehan-Francois Paris et al. proposed a broadcasting technique for low bandwidth channels with multiple viewers3. Chiung-Shien et al. proposed a stream pumping engine for scalable Video-on Demand System and scalable video on demand server4-5. Lawrence A. Rowe et al. describes a strategic retreat to assess the current state of multimedia research and suggest directions for future research6. Some research works are also available on QoS issues7-8 for video on demand services in low bandwidth channels. III. PROPOSED MODEL In our proposed model, the clients register themselves for VOD access at the server using a client-server connection from the network engine. The video stream from the VOD server can also be shared with any of its neighboring PDA or the buddy list using a peer-to-peer connection. As the PDA to PDA wireless links is cannot support high data rate, and there is a very limited memory in both PDAs, we propose to establish a virtual connection between the PDAs During a video stream share, the source PDA transfers the same RTCP connection link to the destination PDA. Since the video servers delivers all packets as multicast packets and the wireless access point (WAP) broadcasts the packet within its transmission range, both sending and receiving PDAs receive the multicast packet and delivers to the application. Thus using this tricky methodology, we design the sharing protocol quite easy and there is no additional packet transmission in the wireless medium. Figure 1 below shows the video sharing technique between PDAs. The proposed model for VOD sharing comprises of two components: first, a client-server connection for user authentication and VOD database access from PDA to VOD server; and the second, is a peer-to-peer connection between two PDAs to share the Video stream with another PDA. IV. IMPLEMENTAION We implement the VOD network engine in HP iPAQ hw6945 Mobile Messenger with operating system Windows Mobile 5.0 Phone Edition. In contrast to the generic computing devices, programming tools in PDA environment offer limited functionalities. J2ME, embedded visual c++, embedded visual basic and embedded c (for Embedded Linux) are such programming compilers. Further, for rendering graphics OpenGL ES is used instead of OpenGL. In our work we have chosen the Embedded Visual c++ as our platform was Windows Mobile, for embedded Linux the Embedded C can be used.

Figure 2(a): Client-Server Connection Protocol

16

Journal of Computer Science, Volume1, Number 1, June 2007, ISSN: 1994‐6244

An Architecture of Active Learning SVMs with Relevance Feedback for Classifying E-mail

Figure 2(b): PDA-to-PDA Connection Protocol in case 1, when the source PDA is already receiving a video stream

Figure 2 (c) PDA-to-PDA Connection in case 2, when source PDA not receiving any video stream. PDA-VOD Server (Client-Server Connection): In implementing the PDA-VOD client server connection, we follow a three-phase connection: User Authentication, Video List transmission for selection and finally the video stream transmission. A Real-time Transport Control Protocol (RTCP) client socket is created up on request from the PDA application to initiate the User Authentication and billing at the VOD server. In the second phase, legitimate users request for the video list for selection and the server send the list to the PDA. Finally, when the PDA selects desired video program, the VOD server sends the multicast video frames one after one as stream. Figure 2(a) shows the PDA-Server connection functionality. We prefer to use multicast property of RTCP for a two-fold benefit: first to avoid the buffering at the PDA and second to avoid second transmission from the source PDA to the destination PDA during video sharing. PDA-to-PDA (Peer-to-Peer): We select a tricky method for the PDA-to-PDA connection for video sharing. As we propose to use the RTCP multicast protocol for video sharing, the repeated transmission from the source PDA to the destination PDA for each packet avoided, and thus, proposed method does not overwhelm the low rate wireless link. The PDA-to-PDA connection can be established in two ways for two cases. The first case is applicable when the source PDA has already been established a PDA-Server connection with the VOD server, i. e., when the source PDA is receiving a selected video stream from the server. The second type comes to the focus when the source PDA did not select any video. Two cases for the proposed video sharing protocol have been shown in figure 2(b) and 2(c). In case 1, when the source PDA already opened a client-server connection with the VOD server and it receives a sharing request from any of its buddies; the PDA can only share the on-going video transmission. The destination PDA (request PDA) opens a connection with the source PDA (sharing PDA) for video list. The source PDA then replies with single entry ongoing video information. If the requested PDA selects the entry, then the source PDA transfer the connection Journal of Computer Science, Volume1, Number 1, June 2007, ISSN: 1994‐6244

17

Md. Saiful Islam et al.

link to the destination PDA and then it can receive the multicast packet. This case is shown in Figure 2(a). In case 2, when the sharing/source PDA has not created any client server connection with the VOD server, it opens a client server connection immediately after receiving the sharing request from destination PDA and receives the available video list and send the list to the destination PDA for selection. The destination PDA selects the video and via the source PDA it initiates the video stream transfer. Source PDA transfers the link to the destination PDA. V.

RESULT ANALYSIS

The implemented protocol has been experimented with a number of test cases with variations in different transport layer packet sizes, and video compressions. Performance of the streaming service for both PDA-Server and PDA-TO-PDA connection has been observed. Table 1: PDA-Server Packet Transmission Delays Packet Delay Size No. of Loss Min. Max. Avg. (byte) Packets (%) (ms) (ms) (ms) 1024 50 0 7 209 66 2048 50 0 11 112 62 4096 50 0 19 117 69 8192 50 0 34 130 81 16384 50 0 59 157 107 32768 50 0 109 209 159 65535 50 1 212 329 264 Table 1 shows the observed per packet delay for video streaming in PDA-Server Connection and corresponding graph is shown in Figure 3. From the results, we find that the PDA-Server connection is capable of providing VOD service on PDAs. Only few packets are dropped with the maximum RTCP packet size (64KB). There is no packet loss below that packet size. Video streaming services generally use the 14KB, which is shown in shaded row in Table 1. Delay columns in Table 1 shows that 16KB packets can be delivered perfectly to cope up with the video information timing. In busy wireless network environment, larger packet sizes can experience larger delays that may degrade the video performance. Shaded region in Figure 3 shows the optimum tolerable video frame packet delay and we find that packet size of 14KB to 32KB the video frames appear in real-time.

Figure 3: Average RTCP packet delay In case of the PDA-TO-PDA connections for video sharing, 16 byte packets are suitable for data transmissions because no real-time multimedia is transferred using the connection. List of video and the connection link with the VOD server are transferred between them. Table 2 shows the observed delay summary for 16 bytes payload size in PDA-TO-PDA transfer. Table 2: PDA-to-PDA Packet Transmission Delays Delay Size No. of (byte) Packets Min. Max. Avg. (ms) (ms) (ms) 16

100

128

271

168

VI. CONCLUSION

18


An Architecture of Active Learning SVMs with Relevance Feedback for Classifying E-mail

We presented an embedded network-engine framework for Video on Demand service that is suitable for the PDA or other resource constraint devices with wireless communication environment. Proposed network-engine has been implemented on PDAs and the results we obtained from experiment justify its acceptance for resource constraint devices. The increasing popularity of VOD services to the users and the recent surge of PDAs and Smart Phones, we believe that such a system, which enables people to easily access video information anywhere and anytime, would be beneficial to many application areas such as education, entertainment, and business. Combining the developed network engine with real applications running in digital cameras and developing some data recovery techniques for large-scale VOD transmission are the prospective future work for this research. VI. REFERENCE [1]

Tarik Taleb, Nei Kato, Yoshiaki Nemoto. On-demand media streaming to hybrid wired/wireless networks over quasi-geostationary satellite systems. Computer Networks: The International Journal of Computer and Telecommunications Networking. Volume 47, Issue 2. pp. 287 – 306. 2005.

[2]

Video-on-demand service system and method of dynamic image distribution. European Patent EP1363456.http://www.freepatentsonline.com/EP1363456.html

[3]

Jehan-Francois Paris, Steven W. Carter, and, Darrell D.E. Long. A Low Bandwidth Broadcasting Protocol for Video on Demand. Seventh International Conference on Computer Communications and Networks (ICCCN '98) p. 690, 1998.

[4]

Chiung-Shien Wu Gin-Kou Ma and Bao-Shuh P. Lin “On the design of stream pumping engine for scalable Video-on Demand System.” IEEE serial number: 0-7803-3734-4/97, 1997, pp-72-73

[5]

Chiung-Shien Wu Gin-Kou Ma and Bao-Shuh P. Lin “A scalable architecture for Video-on-Demand Servers”, IEEE Transactions on Consumer Electronics, Vol. 42, No. 4, serial number: 0098-3063/97, Nov. 1996, pp1029-1036


An Architecture of Active Learning SVMs with Relevance Feedback for Classifying E-mail Md. Saiful Islam1 and Md. Iftekharul Amin2 ABSTRACT In this paper, we have proposed an architecture of active learning SVMs with relevance feedback (RF) for classifying e-mail. This architecture combines both active learning strategies where instead of using a randomly selected training set, the learner has access to a pool of unlabeled instances and can request the labels of some number of them and relevance feedback where if any mail misclassified then the next set of support vectors will be different from the present set otherwise the next set will not change. Our proposed architecture will ensure that a legitimate e-mail will not be dropped in the event of overflowing mailbox. The proposed architecture also exhibits dynamic updating characteristics making life as difficult for the spammer as possible. Keywords: Machine Learning, Artificial Intelligence, and Network Security, Support Vector Machines, Maximum Margin Hyperplane, Spam, Active Learning and Relevance Feedback.


19

I. INTRODUCTION Due to the rapid development of WWW and Internet, e-mail becomes a very efficient, convenient and shortcut way of communication. But at the same time despite e-mail’s numerous advantages, it has become the victim of abuse. The mass posting of unsolicited bulk commercial e-mail has become an increasingly large problem. This unsolicited and unwanted email is known as Spam in the literature. This increase in Spam is annoying. An interesting aspect of this rise in spam is that unique spam attacks are also on the rise. More about spam will be found in http://www.junke-mail.org, http://spam.abouse.net and http://cauce.org. There have been various learning machines to classify e-mail [1-3]. Solutions to the proliferation of spam includes either technical or regulatory [4]. Technical solutions include filtering based on sender address or header content. Microsoft junk and adult content e-mail filters work by looking for key words emphasizing on which words the filter will look for and where [5]. Although sender address based filtering may be useful but content based filtering may sometimes block valid messages [6-7]. Since missing legitimate e-mail is an order of magnitude worse than receiving spam, we have no intention to automatically reject e-mail that is classified as spam. Rather, we will allow user to mark (label) their e-mail as either spam or nonspam. After a finite number of examples are collected, the learning machine will trained and performance on new examples will be predicted. Our proposed mail server will deliver email to the user in decreasing order of probability that the e-mail is nonspam. It is then up to the user to either read the e-mail or trash the e-mail. He or she can also provide feedback to the mail server by indicating if an e-mail misclassified. This is what we want to define relevance feedback (RF). Then we will modify our classifier to take into account this. In the case of overflowing mailbox to provide rooms for less probable spam mail that has the high probability to be legitimate we will delete mails at the bottom of the list. This is one of the promising characteristics of our architecture. The paper is organized as follows: Section II. will describe recent trends and techniques used to classify e-mail using SVMs, Section III. will describe briefly about SVMs, Section IV. will describe active learning strategy, Section V. will describe several design choices and issues section VI. Will describe the proposed architecture and References are listed in Section VIII. II. RELATED WORKS The success of support vector machine (SVM) in solving real-life problems made it not only a tool for the theoretical analysis but also a tool for developing algorithms for real-world problems [1-2, 8-10]. Their efficiency in providing solutions to classification and function approximation problems is the ongoing research issue. Our intention is to classify e-mail using support vector machines. Since this is a new type of research in the area of machine learning, much of the works are still undone. Drucker et al. [2] have described SVMs for classifying e-mail and showed that SVMs perform best among different learning algorithms. They emphasized on rank ordering of the e-mail rather than rejecting it. But since their learner learns only once, it will not be able to cope with newer strategies adopted by spammer that can fake their classifier [7]. To take account into this problem our classifier will update itself upon receiving relevance feedback from users if any e-mail misclassified. Therefore it will be almost impossible for the spammer to misguide our classifier. The concept of relevance feedback we used here is as same as relevancy feedback (RF) in information retrieval (IR) of text documents [10]. The advantages of support vector machines with active learning strategy for spam classification has been described in [1]. But though their architecture exhibits active learning strategy, it is incomplete in the above sense that we claimed for [2]. In this paper we have described a complete architecture of active learning SVMs with relevance feedback for classifying e-mail that is free from of the above problems. It also presents other promising advantages that may prove it as a unique candidate. III. SUPPORT VECTOR MACHINES The key concepts of SVMs are the following: there are two classes, examples:

yi ∈{−1,1}

and there are N labeled training

( x1, y1), ( x 2, y 2),..., ( xN , yN ), x ∈ R d

where d is the dimensionality of the vector. If the two classes are linearly separable, then one can find an optimal weight vector w such that

w • xi − b ≥ 1 if yi = 1; w • xi − b ≤ −1 if yi = −1; or equivalently

1.

yi ( w • xi − b) ≤ 1.

Department of Mathematics, University of Dhaka, Dhaka-1000. Email: [email protected].

w

2

is minimum; and

Md. Showkat Ali.

Training examples that satisfy the equality are termed support vectors. The support vectors define two hyperplanes, one that goes through the support vectors of one class and one goes through the support vectors of the other class. The distance between the two hyperplanes defines a margin and this margin is maximized when the norm of the weight vector

w

is

minimum. Vapnik has shown we may perform this minimization by maximizing the following function with respect to the variables α j : N

N

N

i =1

i =1 j =1

W (α ) = ∑αi − ∑∑αiαj ( xi •xj ) yiyj subject to the constraint: 0 ≤ αj where it is assumed there are N training examples, xi is one of the training vectors, and • represents the dot product. If αj > 0 then xj is termed a support vector. For an unknown vector xj classification then corresponds to finding

F ( xj ) = sign{w • xj − b} where r

w = ∑αiyixi i =1

and the sum is over the r nonzero support vectors (whose α’s are nonzero). The advantage of the linear representation is that w can be calculated after training and classification amounts to computing the dot product of this optimum weight vector with the input vector. For the nonseparable case, training errors are allowed and we now must minimize N

w 2 + C ∑ ξi i =1

subject to the constraint

ξ

yi ( w • xi − b) ≤ 1 − ξ ξ ≥ 0

is a slack variable and allows training examples to exist in the region between the two hyperplanes that go through the

support points of the two classes. We can equivalently minimize

0 ≤ αj . Maximizing W (α )

W (α )

but the constraint is now

0 ≤ αi ≤ C

instead of

is quadratic in α and may be solved using quadratic programming techniques.

The advantage of linear SVM is that execution speed is very fast and there are no parameters to tune except the constant C. Drucker et. al. [2] has shown that the performance of the SVM is independent of the choice of C as long as C is large enough (over 50). Another advantage of SVM’s is that they are remarkably intolerant of the relative sizes of the number of training examples of the two classes. In most learning algorithms, if there are many more examples of one class than another, the algorithm will tend to correctly classify the class with the larger number of examples, thereby driving down the error rate. Since SVM’s are not directly trying to minimize the error rate, but trying to separate the patterns in high dimensional space, the result is that SVM’s are relatively insensitive to the relative numbers of each class. For instance, new examples that are far behind the hyperplanes do not change the support vectors. The possible disadvantages of SVM’s are that the training time can be very large if there are large numbers of training examples and execution can be slow for nonlinear SVM’s, neither of these cases being present here [2]. IV. ACTIVE LEARNING Pool based active learning involves selecting a training set of examples T from a pool of unlabeled examples U. The examples in T are labeled and used for learning. An iterative process involving active learning at each step is called the active learning cycle. This cycle proceeds as follows. At first, all of the examples in the pool are unlabeled and therefore members of U, the pool of unlabeled examples. An initial training set To must have at least one active and one inactive example in order for the maximum margin hyperplane to be found. Therefore we generate To by randomly selecting examples from U and labeling them until at least one example with each label is in To. We then remove the new examples from the unlabeled pool by making Uo = U − To. In turn, we can use

20


A Note On Symplectic And Contact Geometry

active learning to find a new training set T1. T1 contains all the examples from T0, using some selection criterion, we can choose a batch of N examples from Uo, label them and place them in set T1. And we get U1 = U − T1. Repeat this process we can find training set T2 and so on. The iteration of active learning process is: Step1. To find To by randomly selecting examples in U, until one of each label is found. Uo = U − To. Step 2. In iteration i, select a batch Bi of N examples from Ui-1 using some selection criterion. Step 3. Find the label of each example in Bi. Step 4. Ti = Ti-1 ∪ Bi, Ui = U − Ti. Step 5. Return to step 2 and repeat from step 2 to step 4. Here we are first given a set of both labeled and unlabeled data. The pool of training examples can be colleted from individual inboxes, thus it will almost impossible for spammer to write spam messages that can beat our e-mail filter [7]. V. DESIGN CHOICES 1. Feature Representation A feature is a word. In the development below, w refers to a word, x is a feature vector that is composed of the various words from a dictionary formed by analyzing the documents. We take a word as a feature only if it occurs in three or more documents. There is one feature vector per message. w refers to a weight vector usually obtained from some combination of the x’s. There are various alternatives and enhancements in constructing the vectors [2]. We will use the TF−IDF. TF-IDF uses the TF (term frequency) multiplied by the IDF (inverse document frequency). The document frequency (DF) is the number of times that word occurs in all the documents (excluding words that occur in less than three documents). The inverse document frequency (IDF) is defined as

⎛ D ⎞ ⎟ IDF ( wi ) = log⎜⎜ ⎟ DF ( w i) ⎝ ⎠ where

D is the number of documents. Typically, the feature vector that consists of the TF-IDF entries is normalized to unit

length [2]. This kind of representation scheme leads to a very high dimensional feature spaces containing 10000 dimensions or even more [2]. SVMs use overfitting protection mechanism, which does not depend on the number of features. They have potential to handle this kind of problems [1]. Use of stop list consisting of most frequent words like “of”, “and”, “the”, etc. and punctuation marks “.”, “,”, “;” etc. can be used to lessen the dimensionality [2]. But it is observed that there are more punctuation marks in the legitimate e-mail than in the spam [7]. This indicates that spammers are less interested in good punctuation than the other people. Looking at the most frequent words, it is not obvious which e-mail is spam and which is legitimate. Therefore we concentrated on using all the features rather than a subset. 2. Performance Criteria •Error Rate: Error rate is the typical performance measure for two-class classification schemes. However, two learning algorithms can have the same error rate, but the one which groups the errors near the decision border is the better one. • False Alarm and Miss Rate: We define the false alarm and miss rates as

The advantage of the false alarm and miss rates is that they are a good indicator of whether the errors are close to the decision border or not. Given two classifiers with the same error rate, the one with lower false alarm and miss rates is the better one. 3. Training Examples The pool of training examples consists of the individual inboxes. Initially there will be no classifier running. Users will receive e-mails whether it is spam or nonspam. But after a certain amount of time users will be requested to send both legitimate e-mail and spam they considered. This colleting process will be continued until pool is sufficient for the learner. When the learning procedure will complete from training examples, our classifier will be activated and users will receive e Journal of Computer Science, Volume1, Number 1, June 2007, ISSN : 1994‐6244

21

Md. Showkat Ali.

mails according to the probability of their e-mail of being spam. Selection Methods: Here we are given a set of both labeled and unlabeled data. The learning task is to assign labels to the unlabeled data as accurately as possible. To construct SVMs, we must construct an optimal classification hyperplane, i.e. the maximal margin hyperplane. Assume the training data: one subset I for which

( x1, y1), ( x 2, y 2),..., ( xN , yN ), x ∈ R d , yi ∈{−1,1} y =1, and another II subset for which y =-1can be separated by a hyperplane.

If the pattern datasets can be separated without error, the Euclidian distance between the support vectors and the hyperplane must be the largest, i.e. the margin. The maximal margin classifier for our task can be build based on the following theorem. Theorem: Given a linearly separable training examples

U = {( x1, y1),..., ( xN , yN )} the hyper plane that solves the problem min ( w • w) subject to yi (( w • xi ) − b) ≥ 1, i = 1,...N realizes the maximal margin hyperplane with geometric margin

γ= 1

w

2.

There are few criteria that can be used in each iteration of the active learning cycle as mentioned above. One method is to choose the closest examples to the maximum hyperplane. In step j we select the closest examples to the maximal margin hyperplane of T j . But it will be difficult to perform active learning if there are no unlabeled examples closer to the hyperplane than the support vectors. Another selection method is the furthest selection algorithm, which chooses the furthest examples from the hyperplane. This method is not theoretically motivated, and we expected to perform it poorly.

Fig 1. A maximum margin hyperplane with its support vectors VI. Overview of Architecture The block diagram of our proposed architecture is shown in figure-2. The architecture includes a user interface that is responsible for communicating with the user, a pool of training examples, active learning module, e-mail classifier and user mailbox. The total module can be in either Training mode (TM) or Active Mode (AM). In training mode the module will collect individual user e-mails that they considered as spam or legitimate. After collecting sufficient amount of both spam and legitimate e-mails the learning process will begin. Now the module will switch to active mode. In this mode the module classify user’s e-mails as spam or nonspam and place them in his or her mailbox according to their probability. When a user observes that an e-mail is misclassified, he or she will provide feedback with misclassified e-mail and its label he or she considered. After receiving feedback the module will switch to training mode and above described procedure will begin again.

22


A Note On Symplectic And Contact Geometry

Fig 2. Architecture of Active Learning SVMs with Relevance Feedback VII. SUMMARY AND FURTHER WORKS The paper described an architecture of active learning SVMs with relevance feedback for classifying e-mail. We have theoretically given reason why our provided architecture will be unbeaten from spammers. We are now bending ourselves in simulating our architecture and its practical implementation. Although a direct comparison has not been made here, the results of SVMs for spam categorization [2] made it promising. The paper actually motivated to find a framework for spam filter using SVMs. It is a try to explore dynamic updation in learning module that seems not to be written again due to change in spam. VIII. REFERENCES [01] L. Kunlun and H. Houkuan “An Architecture of Active Learning SVMs for Spam”, ICSP Proceedings, 1247-1250, 2002. [02] H. Drucker, D. Wu, V. N. Vapnik, “Support Vector Machines for Spam Categorization”, IEEE Transactions on Neural Networks, Vol. 10, No. 5, Sept. 1999. [03] W.W. Cohen, “Learning rules that classify e-mail”, In Proc. AAAI Spring Sump. Inform. Access, 1996. [04] L. Faith, Craner and B.H. LaMacchia, “Spam!”, Commun. ACM, Vol. 41, No. 8. pp 74-83, August 1998. [05] http://office.microsoft.com/en- us/assistance/HP010421791033.aspx [06] Stefanie Olsen, Staff writer, CNET News.com, “AT&T spam filter loses valid e-mail”, Published: January 24, 2003, 5:21PM PST. [07] Cormac O’Brien, “Compiling a Corpus of E-mail to Evaluate Spam Filters”, https://www.cs.tcd.ie/Cormac.OBrien/ecai2002.pdf [08] X. Wang and Y. Zhong, “Statistical Learning Theory and State of the Art in SVM”, Proceedings of the second IEEE international Conference on Cognitive Informatics (ICCI), 2003. [09] M.P.S. Brown, W.N. Grundy, D. Lin, N. Cristianini, C. W. Sugnet, T. S. Furey, M. Ares, Jr. and D. Haussler, “Knowledge-based analysis of microarray gene expression data by using support vector machines”, PNAS, Vol. 97, No. 1. 262-267, Jan. 4, 2000. [10] H. Drucker, B. Shahraray and D. C. Gibbon, “Relevance Feedback using Support Vector Machines”, In Proceedings of the 18th International Conference on Machine Learning, pages 122--129, 2001. [11]

[6]

Lawrence A. Rowe, Ramesh Jain. ACM SIGMM retreat report on future directions in multimedia research. ACM Transactions on Multimedia Computing, Communications, and Applications (TOMCCAP) archive, Volume 1 , Issue 1. pp. 3 – 13, 2005

[7]

Baschieri, F. Bellavista, P. Corradi, A. Mobile agents for QoS tailoring, control and adaptation over the Internet: the ubiQoS video on demand service. In Proceedings of the Symposium on Applications and the Internet, ISBN: 0-7695-1447-2, pp. 109-118. 2002.

[8]

Michael Ditze, Chris Loeser, Peter Altenbernd. Improving content replication and QoS in distributed peer-topeer VoD appliances. In Proceedings of the 24th International Conference on Distributed Computing Systems Workshops. pp.78- 84. 2004.


23

Md. Showkat Ali.


A Note on Symplectic and Contact Geometry Md. Showkat Ali1

ABSTRACT Contact geometry is the odd-dimensional twin of simplistic geometry. The connection between them is similar to the relation of projective and affine geometry. In this paper some new approach to simplistic and contact geometry are studied. Finally a main theorem in complex contact manifold in contact geometry is established. Keywords: Symplectic Manifolds, Contact Manifolds.

I. INTRODUCTION Contact geometry was born more than two centuries ago in the work of Huygens, Hamiltona and Jacobi as a geometric language for optics. It was soon realized that it has applications in many other areas, including non-holonomic mechanics and thermodynamics. One encounters contact geometry in everyday life when parking a car, skating, using a refrigerator or watching the beautiful play of light in a glass of water. Sophess Lie, Elie Cartan, Darboux [1] and many other great mathematicians devoted a lot of their work to this subject. However, till very recently most of the results were of a local nature. With the birth of symplectic and contact topology in the eighties, the subject was reborn, and the last decade witnessed a number of break through discoveries. There they found new important interactions with Hamiltonian mechanics, simplistic and sub Riemannian geometry, foliation theory, complex geometry and analysis, topological hydrodynamics, 3dimensional topology, and knot theory. Simplistic geometry is the mathematical apparatus of such areas of physics [2] as classical mechanics, geometrical optics and thermodynamics. Simplistic geometry is the geometry of a closed non-degenerate 2-form [3] on an even-dimensional manifold [4]. Contact geometry is the geometry of a maximally non-degenerate field of tangent hyper planes on an odd-dimensional manifold. The symplectic structure is fundamental for Hamiltonian dynamics and in this sense symplectic geometry is as old as classical mechanics. Symplectic manifold are an intermediate case between && hler) manifolds. Symplectic manifolds still play an important role in recent topics in physics, such as real and complex (K a string theory. The contact manifolds are the odd-dimensional analogues of symplectic manifolds. In this paper a theorem on complex contact manifolds in contact geometry is focused that generalizes on symplectic manifolds in symplectic geometry.

II. PRELIMINARIES In this part, we will give some definitions in vector spaces and prove some Propositions. 1. A symplectic vector space is a pair (V , ω ) consisting of a finite-dimensional real vector space V and a non-degenerate skew-symmetric bilinear form (i)

ω : V ×V → ∇. This means that the following conditions are satisfied:

Skew-symmetry: for all

u , v ∈ V , ω (u , v ) = −ω (v, u ).

(ii) Non-degeneracy: for every u, v ∈ V , ω (u , v) = 0, ∀ v ∈ V ⇒ u = 0. The vector space V is necessarily of even dimension since a real skew-symmetric matrix of odd-dimension must have a kernel. A linear symplectomorphism of the symplectic vector space (V , ω ) is a vector space isomorphism φ : V → V which preserves the symplectic structure in the sense that

φ *ω = ω

24


Measurement of Specific Electrical Resistance of Nine Pure Liquid Metals at High Temperature ( 15000C) using Rotating Magnet-field Apparatus.

where

φ *ω (u , v) = ω (φ u , φ v)

for u , v ∈ V . The linear symplectomorphisms of (V , ω )

form a group [5] which we denote Sp (V , ω ). In the case of the standard symplectic structure on Euclidians space we use the notation Sp ( 2n) = Sp ( R2n,

ω0 )

ω0 =

where

n

∑

dx ∧ dy . j j

j =1 2. Let (V , ω ) be a symplectic vector space and F ⊂ V be a subspace. The symplectic complement of F is the subspace F ⊥ = { X ∈ V | ω ( X , Y ) = 0, ∀ Y ∈ F } . The properties of the symplectic complement are summarized in the following: If F and G be the subspaces of a symplectic vector space (V , ω ) , then

⊂ F ⊥ whenever F ⊂ G , ⊥ ⊥ (ii) ( F ) = F , (i) G

⊥

( F + G) ⊥ = F ⊥ ∩ G ⊥ , ⊥ ⊥ ⊥ (iv) ( F ∩ G ) = F + G , ⊥ (v) dim F = dim V − dim F .

(iii)

3. A subspace F of a symplectic vector space V is said to be

F ⊂ F⊥, ⊥ (ii) coisotropic whenever F ⊂ F , ⊥ (iii) symplectic whenever F ∩ F = {0} , ⊥ (iv) Lagrangian whenever F = F . (i)

isotropic whenever

F ⊥ is coistropic and conversely. It follows from the above properties that a Lagrangian 1 subspace of V is the same as an isotropic subspace of dimension dim V . The categories are not exhaustive, a general 2 Note that if F is isotropic then

subspace is neither isotropic, nor coisotropic, nor symplectic. Proposition 1

2

1.

Let

(V , ω )

be

a

2n-dimensional

symplectic

vector

space.

Then

V

has

a

basis

n

{ X , X , L , X , Y1 , Y2 , L , Yn } such that ω ( X a , X b ) = 0, 2ω ( X a , Yb ) = δ ba , ω (Ya , Yb ) = 0

where a, b = 1,2, L , n and such a basic { X , Yb } is called symplectic frame. a

To proof this Proposition, let Q be a Lagarngian subspace of V and let W be some other n-dimensional subspace such that

V = W ⊕ Q . Then W identifies W with Q * by mapping Z ∈ W to the linear form 2ω ( Z ,⋅) ∈ Q * . 1 2 n * Let {Y1 , Y2 , L , Yn } be a basis for Q and let {Z , Z , L , Z } be the dual basis for W = Q . Then ω (Ya , Yb ) = 0 and

2ω ( Z a , Yb ) = δ a . Put λab = ω ( Z a , Z b ) and X a = Z a + λab Yb (with the summation convention). b

Since

λab = −λba , we have


25

Matiur Rahaman Mian

ω ( X a , X b ) = ω (Z

a

+ λ ac Y c , Z

b

+ λ bd Y d )

= ω ( Z a , Z b ) + ω ( Z a , λ bd Y d ) + ω ( λ ac Y c , Z b ) + ω ( λ ac Y c , λ bd Y d ) = λ ab + λ bd ω ( Z a , Y d ) −

λ ac ω ( Z b , Y c ) + 0 1 ba 1 λ − λ ab 2 2 1 ab 1 ab − λ − λ 2 2

= λ ab + = λ ab = 0

and

2ω ( X a , Yb ) = ω ( Z a + λac Yc , Yb ) = ω ( Z a , Yb ) + ω (λac Yc , Yb ) = δ ba + λac ω (Yc , Yb ) = δ ba + 0 =0 a

Therefore, { X , Yb } is a symplectic frame. Hence completes the proof. Proposition 2. Let F be a subspace of a symplectic vector space (V , ω ) and let onto a bilinear form

ω′

on

V ′ = V /( F ∩ F ⊥ ). Then ω projects

V ′ and (V ′, ω ′) is a symplectic vector space.

For the proof of this Proposition, let

π : F →V′

denote the projection along

F ∩ F ⊥ . Define ω ′ by

ω ′( X ′, Y ′) = ω ( X , Y ) where X , Y ∈ F and X ′ = π ( X ) and Y ′ = π (Y ) . This is clearly well-defined, skew-symmetric and non-degenerate. Hence (V ′, ω ′) is a symplectic vector space. Hence completes the proof. III. SYMPLECTIC AND CONTACT MANIFOLDS 4. A symplectic manifold is a pair ( M , ω ) consisting of a manifold M of dimension 2n and a differential 2-form ω such that (i) ω is closed, i.e., d ω = 0, (ii) ω is non-degenerate, i.e., ∀ x ∈ M and any v x ∈ Tx M , one has ω (v x , wx ) = 0, ∀ wx ∈ Tx M if and only if

v x = 0.

If we consider the space M = R2n with coordinates form ω =

n

∑ dx i =1

i

x1 , x 2 , ... , x n , y1 , y 2 , ... , y n equipped with a differential

∧ dy i . Then the pair ( M , ω ) is a symplectic manifold.

Note that the above manifold M is called complex symplectic if M is a complex manifold and

ω

is a holomorphic 2-form.

1. Darboux Theorem 1. (For symplectic Manifold) Let ( M , ω ) be a 2n-dimensional symplectic manifold and p be any point in M. Then there is a coordinate chart

(U , x1 , x 2 , ... , x n , y 1 , y 2 , ... , y n ) centered at p such that on U, n

ω = ∑ dx i ∧ dy i

.

i =1

26



The chart

(U , x1 , x 2 , ... , x n , y 1 , y 2 , ... , y n ) is called a Darboux chart.

5. A contact manifold is a pair (Y , D ) consisting of (2n + 1)-dimensional manifold Y and a rank 2n-subbundle D ⊂ TY of the tangent bundle to Y which is maximally non-integrable in the sense that the Frobenius form [6]

φ : D×D →TY /D (u , v ) a [u , v] mod D

is everywhere non-degenerate. The condition of non-integrability has a very useful reformulation, which we shall now describe. Given a co-dimension 1 subbundle D ⊂ TY , let L := TY / D denote the quotient line bundle on Y, we then have an exact sequence θ 0 → D 2 n → TY 2 n +1 → L → 0 , is the tautological projection and D = ker θ . But we may also think of θ as a line bundle-valued 1-form ∈ H 0 (Y , Ω1Y ⊗ L) , and so attempt to form its exterior derivative dθ . Unfortunately, this ostensibly depends on a

where

θ

θ

choice of local trivialization; for if

α

is any 1-form, d ( fα ) = f dα + df ∧ α . However, it is now clear that

well-defined as a section of L ⊗ Λ D , and an elementary computation shows that 2

*

the Frobenius form mentioned above. Now if the skew form 2n, so that Y must have odd dimension

dθ D is

dθ D thought of in this way, is exactly

dθ D is to be non-degenerate, D must have positive even rank

2n + 1 ≥ 3 . Moreover, the non-degeneracy condition exactly requires that [6]

θ ∧ ( dθ ) n ≠ 0 . Proposition 3. The map φ : D × D → TY / D is a OY linear (or a tensor). For the proof of this Proposition, by the definition of Frobenius form

φ ( u , v ) = [ u , v ] mod D ∈ L.

Now consider for all local function f ∈ OY, we have

φ ( fu , v) = [ fu , v] mod D = =

f [u , v] mod D − v( f )u mod D (Since u ∈ D and v ( f ) is a function, therefore, v ( f )u ∈ D so that f [u , v] mod D − 0

=

f φ (u , v).

v ( f )u mod D vanish).

φ ( u , v ) = −φ ( v ,u ) . So we can conclude that φ ( u , fv ) = f φ ( u , v ). Therefore the map φ is a OY linear or a tensor. Hence completes the proof. Note that

M n be a manifold and T * M its cotangent bundle. Then the projectivization of T * M , P (T * M ) is a contact manifold whose dimension is 2 dim M − 1 i.e., 2n-1.

If we consider

2. Derboux Theorem 2. (For contact Manifold) [7] Every differential 1-form defining a non-degenerate field of hyperplanes on a manifold of a dimension (2n + 1) (contact manifold) can be written in some local coordinate system in the “normal form”

ω = xdy + dz

where

x = ( x1 , x 2 , ... , x n ), y = ( y1 , y 2 , ... , y n ) and z- the local coordinates. IV. THEOREM ON COMPLEX CONTACT MANIFOLDS


27

Matiur Rahaman Mian

* 6. Let M be a complex manifold of complex dimension 2n + 1. Let

{U i } be an open covering of M. We call M a complex

contact manifold if the following conditions are satisfied: (i)

On each

U i there exists a holomorphic 1-form ω i such that ω i ∧ ( dω i ) n is different from zero at every point of

Ui . (ii)

U i ∩ U j is nonempty, then there exists a nonvanishing holomorphic function f ij on U i ∩ U j such that ω i = f ij ω j on U i ∩ U j .

If

1. Theorem 3. (Main theorem) If M is a complex contact manifold of complex dimension 2n + 1, then (i) The structure group of the tangent bundle of M can be reduced to

U (1) × ( Sp ( n) ⊗ U (1)).

(ii) Let ci (M ) be the i-th Chern class of (the tangent bundle of) M and characteristic class of the line bundle over M defined by

α

the

{ f ij } . Then

1 + c1 ( M ) + c 2 ( M ) + L L =

(1 + α )(1 + nα + L L)

In particular, c1 ( M ) is divisible by n + 1 ; c1 ( M )

= (n + 1)α .

(iii) There exists a principal fiber bundle P over M with structure group U (1) such

that P is a real contact

manifold. Moreover, both P and a real contact form on P can be constructed in a natural way from M and

ωi .

It has been shown in [2] that the structure group of the tangent bundle of an orientable real contact manifold of real dimension 2n+1 can be reduced to SO (1) × U ( n)( = SO (1) × ( SU ( n) ⊗ U (1)).

Tx (M ) be the complex tangent space to M at a point x. Assume x to be in U i . As ωi ≠ 0 at x, ωi = 0 defines a 2n – dimensional complex vector subspace Fx of Tx (M ) . Let F be the vector bundle over M with fibers Fx . Let E be the line bundle T ( M ) / F . Then, T ( M ) ≅ F ⊕ E . From the definition of contact form, it follows that dω i , on Fx ,

To proof of (i), let

is of maximal rank and is defined up to a factor, i.e.,

dω j Fx = f ji dω i Fx . Now (i) follows immediately from the definition of Sp ( n) . For proof of (ii), let us consider

ω i = f ij ω j , it follows easily that

ωi ∧ (dωi ) n = ( f ij ) n +1 ω j ∧ (dω j ) n . Since

ω i ∧ (dω i ) n is a holomorphic form of degree 2n + 1, { f ij −( n +1) } defines the canonical bundle K of M. The

characteristic class of K is − c1 ( M ) . The line bundle E = T ( M ) / F is defined by obtain

{ f ij } . From K = E −( n +1) , we

c1 ( M ) = (n + 1)α .

Let ci (F ) be the i-th Chern class of the vector bundle F. By the Whitney Duality Theorem,

1 + c1 ( M ) + c 2 ( M ) + L =

(1 + α )(1 + c1 ( F ) + c 2 ( F ) + L)

From c1 ( M )

28

= (n + 1)α , it follows that c1 ( F ) = nα . Hence completes the proof. Journal of Computer Science, Volume1, Number 1, June 2007, ISSN : 1994‐6244


Finally for the proof of (iii), let L be the line bundle over M defined by

{ f ij −1} and p the projection of L onto M. Let

hi : p −1 (U i ) → U i × C be the coordinate map. If v ∈ L x and x ∈ U i ∩ U j , then hi (v) = ( x, z i ),

h j (v) = ( x, z j ),

z i = f ji z j .

z i . p * (ωi ) = z j . p * (ω j ),

Hence

showing that { z i . p (ω i )} defines a holomorphic 1-form *

(ω + ω ) ∧ (dω + dω ) 2 n +1 = A( z i z i ) n ( z i dz i *

n

n

− z i dz i ) ∧ p (ω i ∧ (dω i ) ∧ ω i ∧ (dω i ) )

where

ω on L. A simple calculation shows that

A = (2n + 1)! /(n!) 2 .

B x be a positive definite Hermitian form on Lx differentiable with respect to x. Let Px = {v ∈ L x ; B x (v, v) = 1}, i.e., Px is a circle in L x . Then P = U x∈M Px is a principal fibre bundle over M with group U (1). We shall show that the restriction of ω + ω to P defines a real contact structure on P. If Px is given by

We define a bundle P as follows. Let

z i z i = b( x) 2 ; (b( x) > 0). As

ω i ∧ (dω i ) n ∧ ω i ∧ (dω i ) n ∧ db is a form of degree 4n + 3 (> real dimension of M) on M, it vanishes identically.

Hence, the restriction of

(ω + ω ) ∧ (dω + dω ) 2 n +1 to P is

2 Ab 2n z i dz i ∧ p * (ω i ∧ (dω i ) n ∧ ω i ∧ (dω i ) n ), which is clearly different from zero at every point P.

V. DISCUSSIONS In this paper we will provide two examples of proof of (ii) and (iii) for the Main Theorem. * 1. Complex projective spaces of odd dimension: Let complex vector space C

2n + 2

z 1 , z 2 , ... , z 2 n +1 , z 2 n + 2 be a coordinate in the (2n + 2)- dimensional

and let P 2 n +1 (C ) be the (2n + 1)- dimensional complex projective space. Then, C

2n + 2

-{0}

is the principal fibre bundle associated with a line bundle L over P 2 n +1 (C ) . Set

ω = z1dz 2 − z 2 dz1 + L + z 2n+1dz 2n+ 2 − z 2n+ 2 dz 2n+1 Let

{U i } be an open covering of P 2 n +1 (C ) and si a holomorphic cross section of the principal bundle C 2n + 2 -{0} over

U i. Set ωi = si (ω ). Then {ω i } defines a complex contact structure on P 2 n +1 (C ) . Considering C 2n + 2 as the real (4n *

4n+4

, we obtain the (4n + 3)-dimensional real projective space P 4 n +3 (P), P 4 n +3 (R) is a principal fiber bundle over P 2 n +1 (C ) with a structure group U (1). Every odd dimensional real projective space is a real contact manifold. The standard real contact form on P 4 n +3 (R) is the one derived from the contact form of P 2 n +1 (C ) in the

+ 4)-dimensional vector space R

manner described in the proof of (iii). * 2. Complex projective co-tangent bundles: Let V be a complex manifold of dimension n +1 and

ω

a holomorphic 1-form

~ on the dual complex tangent bundle T (V ) (= the space of complex co-tangent vectors) defined by ~ u ∈ Tv (T (V )) ω (u ) = v (δ π (u )), ~ ~ where π is the projection of T (V ) onto V and δ π is the differential of π ; δ π : T (T (V )) → T (V ) . In terms of local ~ 0 1 n coordinate z , z , ... , z of V and the induced coordinate z 0 , z1, ... , z n , ζ 0 , ζ 1, ... , ζ n of T (V ) , Journal of Computer Science, Volume1, Number 1, June 2007, ISSN : 1994‐6244

29

Matiur Rahaman Mian

ω = ζ 0 dz 0 + ζ 1 dz 1 + ζ 2 dz 2 + L + ζ n dz n . Fibre is the n-dimensional complex projective space (complex projective co-tangent bundle). This bundle of complex dimension 2n + 1 is our M. Considering the fibre of T (V ) as the (2n + 2)-dimensional real vector space, we take as P the co-

~

~

tangent sphere bundle over V (i.e., the fiber of P is a sphere in the fibre of T (V )). T (V ) − V is the principle fibre bundle associated with a line bundle L over M. The definition of the complex contact structure on M is similar to the one in the first example. The classical real contact structure on the co-tangent sphere bundle P is the one derived from the complex contact structure on the complex projective co-tangent bundle M as described in the proof of (ii). VI. REFERENCES [01] [02] [03] [04] [05] [06] [07]

Arnol’d, V. I. and Novikov, S. P., 1990, Dynamical Systems IV, Springer-Verlag, Berlin Heidelberg. Franel, T., 1997, The Geometry of Physics, Cambridge University Press. Darling, R. W. R., 1994, Differential Forms and Connections, Cambridge University Press. McDuff, D. and Salamon, D. A., 1998, Introduction to Symplectic Topology, Oxford University Press, New York. Libermann, P. and Marle, C-M., 1987, Symplectic Geometry and Analytical Mechanics, D. Reidel Publishing Company. LeBrun, C., 139(1), 1991, 1-43, Thickenings and conformal gravity, Comm. Math. Phys. Arnol'd, V. I., 1978, Mathematical Methods of Classical Mechanics, Springer Verlag.


Measurement of Specific Electrical Resistance of Nine Pure Liquid Metals at High Temperature (~15000C) using Rotating Magnet-field Apparatus. Matiur Rahaman Mian1,PhD

ABSTRACT Specific electrical resistance of nine pure molten metals, mercury, tin, bismuth, cadmium, lead, zinc, antimony, aluminum and silver were measured at different high temperatures using rotating magnet-field apparatus. The results for pure metals obtained were much the same with those of previous investigations.

Keywords : electrical resistance, high temperature, rotating magnet-field and apparatus. I. INTRODUCTION Many investigations have been already made of the electrical resistance of molten metals, especially in connection with the discontinuous change of electrical resistance during the molten of the metals. L. de la Rive [01] first studied the electrical resistance of six metals, namely tin, bismuth, cadmium, lead, zinc and antimony at high temperatures and found that in case of tin, zinc, cadmium and lead, the value of the electrical resistance increases abruptly during fusion, but in the case of antimony and bismuth, on the contrary, decreases abruptly during melting. G Vincentini and D Omodei [02] measures the electrical resistance of five metals namely tin, bismuth, thallium, cadmium and lead in the molten state and found that the specific resistance of molten metals at the melting point is proportional to the atomic weight. E F Northrup [03] studied the resistance of twelve pure metals, namely sodium, aluminum, potassium, copper, zinc, cadmium, tin, antimony, gold, quicksilver, lead and bismuth, at high temperature, and found that the electrical resistance of molten metals increases linearly with the rise of temperature in a wide range. H. Tsutsumi et al [04], Vassura et

30


A Comparative Study of Cell Based and Tree Based Call Admission Control for Mobile Cellular Network

al [05], Müller et al [06], Hackspill et al [07], Bridgman et al [08] and Bernini et al [09] also studied the resistance of molten metals and their results are much the same. II. EXPERIMENTAL TECHNIQUES The electrical resistively of the liquid alloys has been measured by several experimental arrangements which are in principle are based on the following two different methods [10]: 01. Direct method 02. Indirect (Electroden loss) Method: (a) Inductive Coupling Method (b) Rotating Magnet field Method The direct method is similar to the current-potential methods used with solid specimen. This method involves the introduction of thin capillaries [11], in which the potential drop along a capillary filled with the liquid is measured and converted into electrical resistivity by means of an accurate knowledge of the dimensions of the capillary. This method is in principle, capable of measuring very high precision, but in practice the method is suitable for above 2000C temperature. This is due to the experimental technique and thermoelectric e.m.f. Also in this method, most metals used to form connections to the melt because of their resistance to attack by liquid metals from thin rectifying oxide films, even at low temperatures and oxygen pressure that are impossible to reduce in hydrogen atmosphere. The presence of the films make precise measurement of the potential drop along the capillary very difficult and sometimes cause large discrepancies between values found with the different magnitudes or directions of the capillary current A second method involves an “H” cell with two large vessels connected by a thin capillary. The current and potential leads are contained in reservoirs. Matuyama [12] used a single capillary of length 300 mm and diameter 5.5 mm tube containing iron lid inserted in the surrounding large vessel of liquid. The absolute measurement of resistance in this method is not very accurate. The inductive coupling method developed by Nyberg and Burges [13] and Tomlinson and Licter [14] was the first scientist to use that method for the determination of electrical resistivity of liquid metals. The technique involves placing the sample in a solenoid producing an alternating magnet field, which induces eddy current in the sample. The observable effects are the change of impedance of the solenoid, more specifically, an increase in resistance and decrease in inductance of the solenoid, which is related to the sample conductivity (resistivity). In the four-prove method, the voltage drop along a well-defined specimen is measured using d.c current and the resistivity is calculated by Ohm’s law [15]. For higher temperatures, the rotating magnet-field method is more reliable. Several variants of this electrodeless method have been developed which are not always generally applicable. In the high-frequency method, the sample is dropped through a field of 10 MHz, the damping of the field being a measure of the electrical resistivity [16]. The rotating magnet field method was developed by Braunbeck, et al [17] and later modified by [18-23]. In this method, the specimen, in the form of a short cylinder, is suspended with its axis vertical in a magnet field rotating about the axis of the specimen. The eddy currents induced in the specimen interact with the applied field to produce a rotation of the specimen about the vertical axis. This rotation is measured and converted into electrical resistivity. This method is suitable for both solid and liquid specimens III. APPARATUS FOR MEASUREMENT OF RESISTIVITY The electrical resistivity was measured by the electroden loss Rotating Magnet-field Techniques developed by [17] used in this investigation. In the Soviet Union, the method was independently developed by Reget and described in details by Glazev et al [24]. The general principles of this techniques is further modified by Grube and Spiedel [18], Roll et al [19], Regel [20], Loffe and Regel [21], Samarin [22], Takeuchi and Endo [23].


31

Md. Shahid Hossain et al.

IV. THE APPARATUS The apparatus is illustrated in figure-1 The main furnace is a vacuum tight aluminum tube of length 333 mm, internal diameter 20 mm and outer diameter 27 mm is directly heated by a bifilar – wound with Mo-resistance wire ( d = 5 mm) and opens at both ends. The central 250 mm is surrounded by an austenite (25-20 Cr-Ni) stainless steel tube, which carries the furnace winding, insulated from the tube by a layer of mica. The two tubes are separated by a gap of 1.06 mm and are held concentric by a pyrophyllite plug at each end of the outer tube. The furnace winding of 0.0359 inch in diameter tray in three non-inductive separate sections, each having a resistance of about 10Ω and each externally shunted by a variable resistance. The stainless-steel tube is surrounded by alumina brick and powder insulation contained inside 220 mm diameter asbestos of outer tube. The heating of the magnetic coil is prevented by radiation shields, by a vacuum of 10-2 Pa, and by water cooled outer shell. The rotating field is provided by a stator coil from a three phase stabilized alternating current regulator. The current intensities of the three phases can easily measured by a digital multi-meter. The field in the centre is provided approximately by H = 10i, where i is the coil current (max.30A) and H is field strength in gauss. The stator coil surrounds the asbestos outer tube of the furnace, which is approximately 225 mm in diameter. The 18 mm gap between stator and tube contains copper cooling coils which serve to maintain the stator at constant working temperature. The very weight of the stator is supported on three Duralium legs which are bolted to the brass base-plate. The base-plate is rigidly tied to the top-plate by the three tension brass rods. When assembled within its Dexion frame, the apparatus may be leveled by the three screws. The stator current is controlled by a three-phase voltage stabilizer and a variac auto-transformer. The currents are measured to ± 0.1% accuracy. The currents in all the three phases are adjusted to exactly equal. As the temperature coefficient of resistivity of liquid metals is usually about 5 x 10-8Ωdeg C-1. The control of furnace temperature is very important. A standard saturable-reactor type controller was used with a platinum resistance thermometer. The resistance thermometer is placed against the central furnace winding and can control furnace temperatures up to 15000C. The temperatures are measured with a calibrated Pt-Pt/13%Rh thermocouple contained in an alumina sheath whose tip lies about 2 mm below the suspended crucible. Thus a zone of 20 mm length with a temperature variation of ± 2oK could be obtained. The dimensions of the specimen are measured over the complete range of temperatures to be used for the resistivity measurements. Two probes, one of which can be moved in a vertical position, are placed in the liquid metal contained in a crucible in an inert atmosphere. They are connected electrically to a battery so that a current flows between them when both are immersed in the liquid. The liquid surface may then be located by withdrawing the sharp-pointed movable probe until the current ceases to flow. The probe is then again lowered until contact is made, and the true position of the surface recorded as the mean of the two positions of the probe, measured by the cathetometer. The bottom of the crucible may then be located exactly as the lowest point attainable by the probe. The difference between the two readings, corrected for the expansion of the probe, is then the correct length of the specimen

32



The specimen was placed in a pure Boron Nitride (BN) crucible. The boron nitride is inert, has a low electrical conductivity, low thermal expansion coefficient ( α lin = 0.9 x 10-6 ) and can be easily machined. The BN crucible consists of internal diameter 14.3 mm and height 20.7 mm is shown in figure-2. The crucible is screwed and fitted with an additional BN suspended rod reaching into the cooler part of the furnace which carried at the upper end of the hexagonal concave mirror and the aluminum vanes. The mirror reflects a He-Ne Laser light beam through the optical window on to a linear scale of rigidly fixed distances of one meter. The caps also contain a small hole to allow the escape of gas during evacuation. The suspension Phosphor-bronze wire of length 200 mm and outer diameter 0.15 mm is attached at its upper end to a brass rod which enters the chamber through an O-ring seal, and which can rotate to zero the image on the scale or moved vertically to adjust the height of the crucible. The tungsten wire is clamped at either end by means of two chucks. The rotational movements of the suspension are damped by two arms which dip into a bath of silicone oil. Thus the angular movements of 0.01 radian can easily detect. The entire torsion system (Crucible, BN-rod, Hexagonal mirror, Al-vanes) is suspended by the wire. The torsion angle α can be determined with a He - Ne Laser light beam by measuring the distance on two calibrated scales normal to the incident beam and a well defined distance d = 129 mm.

b j −a j b o −a o − arctan ] ………(1) d d Where a o and b o are the readings on the scales a and b respectively before the measurements and a j and b j the 1 2

α = [arctan

readings on the scales at the measurement. The use of He -Ne Laser resulted in a sharp image on the light beam simultaneously on both the scales. To ensure good reproducibility and high accuracy, the BN crucible has to be completely filled with the melt at each measurement. The excess of the melt could rise in the capillary in the crucible lid and had thus no influence on the results. At the bottom of the crucible is a Pt-Pt/13%Rh thermocouple that measured the temperature. V. WORKING PRINCIPLES OF THE APPARATUS When a metallic specimen is placed in the centre of a rotating magnet field, an eddy current is induced in the specimen. By the mutual interaction between the external magnetic field and an additional magnetic field induced by the eddy current, a mechanical moment, inversely proportional to the electrical resistivity of the specimen. In case of liquid specimen, the rotational moment of the liquid in its container causes rotation of the container through the viscosity of the liquids and the rotating moment M is given by

M =

π

4

.σ .ω.L.R 4 .H 2 −

π

1 . .σ 2 .ω.L.R 6 .H 4 ...(2)

192 η

Where ω = the angular velocity of the rotating magnetic field, H = the external rotating magnetic field strength, η = the viscosity of the specimen, σ = electrical conductivity of the specimen, L = length of the specimen and R = radius of the specimen respectively. By proper control of the dimensions of the specimen and the field strength H, the second term of the equation (2) becomes negligibly small. In case of metals of unknownη , the rotating moment induced in the specimen can be balanced by the torsion of the strip, the lower end of which is attached to the container containing the specimen and the resistivity can be obtained by measuring the torsion angle as

M =

σ =C

α H

2

π

4

σωLR 4 H 2 = Kα …………………...(3)

=C

α i2

and

as α ≈ H ……………..( 4 )

4K = Apparatus constant, depends on the specimen dimensions, frequency of the rotating magnetic field πωLR 4 and magnetic field strength, α = torsion angle and i = current intensity.

Where C =

The exact derivation of the equations can be found elsewhere [17, 25-26] . VI. SAMPLE PREPARATION Samples were prepared using high purity metals from the American Smelting and Refinery Company, USA and from Ventron Company, FRG. To remove the oxide layer (if any), the metals were further purified by melting the metals under vacuum and filtering the metal through quartz wool. The metals were weighed on an analytical balance and placed in a graphite crucible, which had exactly the same inner diameter as the BN- crucible used for the measurements. The graphite Journal of Computer Science, Volume1, Number 1, June 2007, ISSN: 1994‐6244

33


crucible was sealed under a vacuum of 10-2 Pa in quartz; the metals were melted in a oven. The melted metals were kept for a week at 1273oK. After quenching in normal water, the specimens were removed from the graphite crucibles and grounded to the desire length of 20.5 mm. This procedure had to be followed since some of the metals were very brittle. VII. EXPERIMENTAL PROCEDURE The samples were put into the BN crucible which were connected to the torsion system and the whole arrangement was placed into the apparatus with the crucible at the centre of the alumina tube. The system was evacuated up to 10-2 Pa under argon atmosphere. The samples were then heated slowly under slow flow of argon gas. At different temperature 5-7 measurement were taken in the solid state. The sample was slowly heated (50oC/hour) up to about 20oK above the liquidus temperature and remained about 2 hours at that temperature. After a constant temperature was attained, the zero readings ( a o , bo ) of the torsion angles were taken. By changing the coil currents from 0.5A to 5.5A, up to ten measurements of the torsion angles ( a j , b j ) and of the corresponding current readings were taken. The arithmetic mean of the current readings of the three phases were calculated from

i1 + i2 + i3 ………………………(5 ) 3 2 Thus for each temperature 10 values of α / i were calculated. Since these values showed slide drift, the arithmetic mean of i=

these data point was used to calculate the value of the conductivity of the given temperature. The temperature was then again raised to 20oK and waited till the temperature was again constant, the next measurements were taken. Since it was not possible to complete a series of measurements for a sample in one day, the samples were kept overnight in the molten state in the apparatus. The results of the measurements on the consecutive day agreed well with one another. In this way about 10-12 measurements from liquidus temperature up to 200oK above the liquidus temperature was carried out. Each time the samples were heated to the desired temperature and after a constant temperature was attained, the readings were taken. To cheque the reproducibility of the results, experiments were also carried out when the samples were cooled. The results on both the heating and cooling were in very good agreement. These samples were used to utilize to measure the value of the electric conductivity of the metals. VIII. EXPERIMENTAL RESULTS & DISCUSSION 1. Calibration of the apparatus Constant C : For the evaluation of the results, the apparatus has been calibrated using sample of pure magnesium, gallium and lithium. The results are tabulated in table-1.

α /i2 ρ

Metals used

0

(973 K

Gallium (Ga)

1.14 A

Mean

C

C 3.54 x10 A Ω 4

2

24.75 x1 cm-1 cm

Magnesium (Mg)

1.04 A

3.47 x10 4 A 2 Ω 3.496 x10 4 27.72 x10 cm-1 cm-1

cm Lithium (Li)

0.218 131.90 x1 3.48 x10 4 A 2 Ω cm

cm-1

This value of the apparatus constant C was calculated using equation (4). The results of these values are well fit with the results of Busch et al [27], Cusack et al [28] and Roll et al [29] respectively. 2. Experimental Date : The experimental electrical resistivity measurements were carried out on nine pure metals over the temperature range from liquidus temperature up to 1000K. The data obtained from heating and cooling were in good agreement. The experimental points about 12-17 data for each metals were taken. The results for few temperatures are listed in Tables 2 ~ 10.

Table 2: Mercury(Hg) of M.P.-38.9 oC*

34



Measurement Temperature ( 0C) -19 -11 0 13 44 100 145 200 240 275 320 340 362 400

Measurement Resistivity σ x 106 94.8 95.2 96.4 97.5 100 105.7 110.7 120.3 128.8 135.2 146.7 152.2 155.7 157.3

M.P.=Melting Point, * = Data from [12 ] Table 3: Tin (Sn) of M.P. 231.9 oC Measurement Temperature ( 0C) 265 295 325 348 379 440 485 563 600 668 718 783 847

Measurement Resistivity σ x 106 48.6 49.5 50.1 50.9 51.5 52.2 52.9 53.8 54.6 55.6 56.2 57 57.6

Table 4: Bismuth (Bi) of M.P. 231.9 oC Measurement Measurement Temperature ( 0C) Resistivityσ x 106 279 128 325 129.7 375 133 400 134.8 440 136.8 471 138.1 526 140.9 550 142.8 590 144.9 639 146.8 680 148.9 Journal of Computer Science, Volume1, Number 1, June 2007, ISSN: 1994‐6244

35


709 752 795

150.9 152.1 155.8

Table 5: Cadmium(Cd) of M.P 320.9 oC Measurement Measurement Resistivityσ x 106 Temperature ( 0C) 351 32.01 392 32.18 419 32.23 457 32.32 494 32.48 528 32.56 550 32.67 574 32.72 633 32.92 650 33.98 700 33.18 740 33.26 762 33.29 Table 6: Lead (Pb) of M.P. 327 oC Measurement Measurement Resistivity σ x 106 Temperature ( 0C) 348 93.6 375 94.6 420 96.7 472 99.6 517 102.5 550 103.5 578 104.6 602 106.8 641 109.1 682 110.6 731 112.9 776 114.8 836 119.6 900 121.8

Table 7: Zinc (Zn) of M.P. 419.39 oC Measurement Temperature ( 0C) 445 484 512 539 570 627 669 695 719 737 749

36

Measurement Resistivity σ x 106 37.2 37 36.8 36.6 36.3 35.6 35.1 34.8 34.6 34.4 34.2



763 801 852

33.9 33.6 33

Table 8: Antimony(Sb) of M.P 630 oC Measurement Resistivityσ x 106 Measurement Temperature ( 0C) 658 115.5 708 116.1 746 116.6 808 118.3 861 119.2 913 120 945 120.7 970 121.2 990 121.9 1019 122.6 1049 123.3 1091 124.2 1118 124.5 1152 125.3 Table 9: Aluminium (Al) of M.P.658.5 oC Measurement Measurement Resistivityσ x 106 Temperature ( 0C) 686 26 715 26.4 745 26.8 774 27.4 806 27.9 831 28.1 853 28.6 871 28.9 900 29.4 930 29.8 980 30.4 1023 30.9 1060 31.5 1100 31.9 Table 10: Silverc (Ag) of M.P. 960 oC Measurement Temperature ( 0C) 980 1001 1025 1047 1070 1096 1121 1143 1161 1173 1188 1199

Measurement Resistivity σ x 106 17.5 17.9 18.3 18.8 19.1 19.6 20 20.4 20.7 20.9 21.2 21.5


37


1206 1210

21.7 21.9

The calculated values of the resistivity varied for Aluminum from 26.0 x 106 – 31.9.x 106 between temperature range 6861100 oC; Mercury from 94.8 x 106 - 157.3 x 106 between temperature -19 – 400 oC; Tin from 48.6 x 106 - 57.6 x 106 between temperature 265 -847 oC; Bismuth from 128 x 106 – 155.8 x 106 between temperature 279-795 oC; Cadmium 32.01 x 106 – 33.29 x 106 between temperature351-762 oC; Lead from 93.6 x 106 – 121.8 x 106 between temperature348-900 oC; Zinc from 37.2 x 106 - 33.00 x 106 between temperature 445-852 oC; Antimony from 115.5 x 106 - 125.3 x 106 between temperature 658-1152 oC; and Silver from 17.5 x 106 - 21.9 x 106 between temperature980 - 1210 oC. The experimental data were compared with the data from the literatures. It was found 98% of the data were well fit with the literature. In some metal a slide variation of 2-3% were observed, this may due to impurities in the metal. IX. REFERENCES [01] [02] [03] [04] [05] [06] [07] [08] [09] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29]

38

de la Rive L, Arch. des. Sci. Phys. (Geneva) 17, 18, 362, 1863 Vincentini G and Omodei D, Atti. Acc. Soc. Torino, 20, 1884 Northrup E F, Trans. Amer. Elektrochem. Soc. 25, 373, 1914 Tsutsumi H. et al, Sci. Rep. 7, 93, 1918 Vassura et al, Neuvo. Cimento, 1890 Müller et al , E. T. Z, 13, 1892 Hackspill et al, Compt. rend. 151, 305, 1910 Bridgman et al, Proc. Amer. Acad. 56, 61, 1921 Bernini et al, Phys. Zeits. 5, 406, 1904 & 6, 74, 1905 Matiur R Mian et al, Bangladesh J.Sci Ind. Res., XXXI, 1, 1996 Scala E and Robertson W D , J Metals, N Y, 5, 1141-47, 1953 Matuyama Y, Science Report, Tohuku University, 16, 447, 1927 Nyberg D W and Burges R E, Can J Phys, 40, 1174, 1962 Tomlinson J L and Lichter B D, Trans.Met.Soc, AMIE, 245, 2261, 1969 Miller E, Paces J and Komarek K L, Trans. Met. Soc. AIME, 230, 1557, 1964 Schurnberger W, Thurn H and Krebs H, Ber. Bunsen Ges., 80, 730, 1976. Brawenbeck W, Z Phys, 73, 312 – 334, 1932 Grube G. and Spiedel P Z, Elektrochem., 46, 233 – 242, 1940 Roll A, Felger H. and Motz H , Z. Metallkunde, 47, 707 – 713, 1956. Regel, Ph.D Dissertation, University of Leningrad, U.S.S.R., 1957 Loffe A F and Regel A R , Progr. Semiconductor, 4, 237 – 291, 1960 Samarin A M , J. Iron St. Inst., 200, 95 – 101, 1962 Takeuchi S. and Endo H., Trans. Japan Inst. Metals, 3, 35 – 41, 1962. Liquid Metals, Chemistry and Physics by Sylvan, Z Beer, Maercel Dekker, Inc, New York, 1972 Eder O J, Kunsch B, Lacom W and Neumann W, Ost. Studienges f. Atomenergie, SGAE Ber No. 2194, Seibersdorf, Austria, 1973 Eder O J, Kunsch B, Lacom W and Neumann W, Ost. Studienges f. Atomenergie, SGAE Ber No. 2379, Seibersdorf, Austria, 1974 Busch G and Tieche Y, Phys Kondens. Mat., 1, 402, 1967 Cusack N E and Kendall P, Proc. Phys. Soc., 75, 309-311, 1960 Roll A and Motz H, Z. Metall, 48, 272-280, 1957.


I. Introduction Mobile cellular network is growing rapidly and simultaneously the user of the cellular network is growing. As a result the number of new calls and handover calls in limited cells are increasing in a tremendous rate. Handling handover calls and new calls became a big challenge for cellular network. There are different methods for call admission for new calls and handover calls given in [1-2]. Two types of call admission, cell based call admission and tree based call admission are compared in this paper. Figure 1 shows a structure of a mobile cellular network. The fixed network is attached with two different switches. Each of the switches is covering 7 cells. Tree base call admission is shown under switch A and cell based call admission is shown under switch B. A mobile phone subscriber (also called mobile host) use one of this cell, when he/she receives a new call or while he/she travels among the cells. In this paper we will try to compare whether a cell based call admission or tree based call admission is more efficient for a MH (mobile host). Here λ is considered as new call arrival rate/cell, h is considered as handover rate/cell, µt is considered as call termination rate for new call and µh is termination rate for handover calls. Here number of user is unlimited. Details of call admissions are explained in section II. Network traffics are analyzed using state transition diagram in section III. In section IV results of the analysis are discussed. Finally section V gives concluding remarks.

Call Admission Control TREE BASED CALL ADMISSION CONTROL In case of tree based call admission control a new call or a handover call is admitted to a group of 7 cells which is shown in Figure 1. So, new call arrival rate for the group of 7 cells is 7λ. Handover rate for the tree is 3h as there are 18 outside edge for the tree and handover rate to an edge of the cell is h/6 (18 X h/6 = 3h) which is also mentioned in [1]. Call termination rate is µt for new call arrival and µh for handover calls. Here 7C is total number of channels for the tree where N number of channels used for both new calls and handover calls. 7C-N channels are reserved for handling only handover calls.

Figure 1. Mobile Cellular Network

7C handover

3h N

new + handover

7λ 0 Figure 2. Blocked diagram for tree based call admission

CELL BASED CALL ADMISSION CONTROL As the new call in admitted cell wise the new call arrival rate for the cell based technique is λ and call termination rate for new call is µt., handover rate is h and termination for handover is µh. Here C is total number of channels in a cell where m channels are both for new and handover calls and C-m is reserved for only handover calls.

1. 2.

Department of Computer Science and Enginering, IBAIS Universit. e-mail: [email protected]. Warid Telecom, Gulshan, Dhaka. e-mail: [email protected]

Modeling a Medical Diagnosis System using Bayesian Belief Network

C handover

h m

new + handover

λ 0 Figure 3. Blocked diagram for cell based call admission 3h

7λ+3h 7λ+3h 7λ+3h

7λ+3h 0

2

1 µt

3h

N

2µt

3µt

Nµt

3h N+2

N+1 Nµt+µh

3h 7C

Nµt+2µh

Nµt+(7C-N)µh

Figure 4. State transition diagram for tree based call admission control λ+h λ+h

λ+h

λ+h 0

2

1 µt

h

h

2µt

3µt

mµt

h

m+1

m mµt+µh

h

m+2

C

mµt+2µh

mµt+(C-m)µh

Figure 5. State transition diagram for cell based call admission control

III. Traffic Analysis State transition diagram [3] for tree based call admission is shown in figure 4. And for cell based call admission control shown in figure 5. By analyzing the state transition diagram in figure 4 and using cut equation we find following derived equations: Blocking probability for new call arrival in a tree based technique is 7. λ Pbn( λ )

3. h

N

3. h

N

µt N

7. λ

i= 0

i

7. λ

3. h . 1 i! µt

µt

.1 N! .1 . N!

7. C N

j ( 3. h )

j

j= 1

( i. µh ) i= 1

3.h

7. λ

N

µt

Probability of handover failure in a tree based technique is

7 .C N

. . 1 . ( 3 h) N ! 7.C N

( i. µh ) Phf( λ )

i= 1 N i= 0

7. λ

i 3. h . 1 i! µt

7. λ

3. h µt

N

.1 . N!

7. C N

( 3. h )

j

j

j= 1

( i. µh )

i= 1 By analyzing the state transition diagram in figure 5 which is also used in [4-10] and using cut equation following equations are derived Blocking probability for new call arrival in a cell based technique is

λ Pbn( λ )

h

m

µt m i= 0

λ

i

h .1 i! µt

λ

h µt

m

. 1 m!

. 1 . m!

C m j= 1

h

j

j ( i. µh ) i= 1

Probability of handover failure in a cell based technique is


31

Md. Mizanur Rahman et al.

λ

h

m

µt

C

h . 1 . m! C m

m

( i. µh ) Phf( λ )

i= 1 m

λ

i= 0

i

h .1 i! µt

λ

h µt

m

. 1 . m!

C m

h

j

j

j= 1

( i. µh ) i= 1

IV. RESULT AND DISCUSSION

Blocking probability of new arrival and handover failure of both cell and tree based admission is plotted against new call arrival rate taking number of allocated channel of new arrival call and guard channel of handover call as parameters shown in figure 6-9. Here handover arrival rate h=0.2 calls/min, handover termination rate µh =1.2 calls/min, total termination rate µt = 2.6 calls/min and total number of channels C=8 taken for cell based call admission technique. For tree based admission, number of channels for new call N=14 and average number of channels per cell C=3. Both blocking probability of new arrival call and handover failure increase with increment of call arrival rate at the same time both probability decreases with increment of allocated channel of respective traffic. When large number of users shares a common string of channels usually show better performance than the situation when several group of channels are allocated for individual group. The situation is also visualized in two call admission schemes i.e. tree based admission shows better performance with less number of channels/cell. Blocking Pr. of new call

Tree Based Call Admission 0.1 0.01 3

1 .10 4 1 .10 1 .10

5 6

1 .10 7 1 .10 1 .10

8

1 .10

9

1

2 3 New call arrival rate (calls/min)

4

N=14, C=3 N=16, C=3 N=18, C=3

Pr. of Handover filure

Figure 6. Blocking probability against new call arrival rate for tree based call admission 1 .10 1 .10 1 .10

Tree Based Call Admission

6

7

8

9

1 .10 10 1 .10 11 1 .10 12 1 .10 13 1 .10 14 1 .10 15 1 .10 16 . 1 10 17 1 .10

1

1.5

2 2.5 3 New call arrival rate (calls/min)

3.5

4

N=14, C=3 N=16, C=3 N=18, C=3

Figure 7. Probability of handover failure against new call arrival rate for tree based call admission

32


Modeling a Medical Diagnosis System using Bayesian Belief Network

Blocking Pr. of new call

Cell Based Call Admission 0.1 0.01 1 .10

3

1 .10

4

1 .10

5

1 .10

6

1 .10

7

1


4

m=5, C=8 m=6, C=8 m=7, C=8

Pr. of Handover filure

Figure 8. Blocking probability against new call arrival rate for cell based call admission 1 .10

3

1 .10

4

1 .10

5

1 .10

6

1 .10

7

1 .10

8

Cell Based Call Admission

1


4

m=5, C=8 m=6, C=8 m=7, C=8

Figure 9. Probability of handover failure against new call arrival rate for cell based call admission

V. Conclusion All the graphs of previous section yield logical results. Here traffic is modeled using one dimensional Markovian chain but result would be more accurate if two dimensional chain (where any probability sate (i,j) ; i=number of new call and j=number of handover call) were used. Inclusion of queue with guard channel could further enhance carried traffic of handover part.

VI. References [01] [02] [03] [04] [05] [06] [07] [08] [09]

K. Wang and L. Lee, ‘Design and Analysis of QoS Support Frequent Handover Schemes in Microcellular ATM Networks’, IEEE Transaction on Vehicular Technology, Vol. 50, No. 4, pp942-953 July 2001 Y. Fang and Y. Zhang, ‘Call Admission Control Scheme and Performance Analysis in Wireless Mobile Network’ , IEEE Transaction on Vehicular Technology, Vol.51, No.2, pp. 371-382, March’2002 J.F.C. Kingman, ‘Markov population processes.’ J. Appl. Prob., Vol. 6 (1969), 1-18, 1969 G.V Tsoulos, ‘Expermental and theoretical capacity analysis of space-division multiple access (SDMA) with adaptive antennas’ IEEE proccommunications, vol.146, No.5, pp. 307-311, Oct’1999 V. B. Iverson, ‘Teletraffic Engineering Handbook’, Technical University of Denmark, November 2001 M. I. Islam, M. Q. Maula, and L. J. Rozario, ‘Analysis of Two Dimensional Limited Source and Mixed Traffic Model for a BTS of a Small Mobile Cellular Network’ Proceedings of ICCIT’ 2001, pp190-196 M. S. Hossain and M. I. Islam ‘A Proposed 2-D Queueing Model of PCT-I Traffic’ Proceedings of ICCIT’ 2003, vol. I, pp114-118. M. S. Hossain and M. I. Islam, ‘A Theoretical Analysis of SDMA Traffic of Limited User Network for Duplicate at Last Case’, Proceedings of ICCIT’ 2004, pp 728-732. M. S. Hossain and M. I. Islam, ‘Analysis of SDMA PCT-II Trafficfor Duplicate First Case -A Theoretical Approach’, Proceedings of 2005 International Conference on Wireless Communications Networking and Mobile Computing,Vo. I, pp 423-426, Wuhan University, Wuhan , China


33

Md. Mizanur Rahman et al.

[10] [11]

Kwan L. Yeung and Sanjib Nanda, ‘Channel Management in Micro/Macro Cellular Radio Systems’ IEEE Transaction on Vehicular Technology,Vo.45, No.4, pp.14-23, Nov’1996 S. A. El-Dolil, W.C. Wong, and R. Steele. Teletraffic performance of highway microcells with overlay macrocell. IEEE Journal on Selected Areas in Communications, 7(1):71,78, January 1989.


Modeling a Medical Diagnosis System using Bayesian Belief Network Md. Mizanur Rahman1 and Md. Sharrukh Zaman2

ABSTRACT Bayesian belief networks are useful tools for coherent representation of uncertain knowledge. They are widely used in medical diagnostic reasoning. There are many expert systems to help physicians diagnosing different classes of diseases. In this paper we propose a system using belief networks to do the primary diagnosis of common diseases. Here a detailed study of physicians’ working procedure has been performed and it is modeled in a belief network. So far as we know, this is the first of its kind. This will be a very much helpful tool for primary diagnosis.

Keywords: Bayes’ Rule, Belief Network, Medical Diagnosis. I. BACKGROUND Many AI techniques have been used in medical diagnostic reasoning. Among them Bayesian belief network remains in a higher position because it allows human to apply expertise and prior knowledge directly into the computations.

Ex-

= P (Ex-|X, Ex+) P (X|Ex+)/P (Ex-|Ex+), Keeping as fixed background evidence.

U1

Bayes’ rule of computing conditional probability is the foundation of belief networks. A belief network is a directed acyclic graph whose nodes correspond to random variables. Links denote direct dependencies and the conditional probabilities for directly dependent variables are part of the network. A generic singly connected network is shown in the figure1. Here Ex+ is the casual support and Ex- is the evidential support for X, which combines to form E. P(X|E) is calculated using backward-chaining algorithm derived by Russell and Norvig [3]. For convenience, derivation of the algorithm is given below. Our aim is to compute P(X|E) given that X is not in E. To mention again, E is divided into two parts - one is Ex+, the casual support for X, i.e. the evidence variables that are its parents or connected via the parents of X. The other part is Ex- - the evidential support for X, i.e. the evidence variables that are children of X or connected via children of X. Hence we have, P (X|E) = P (X| Ex-, Ex+) 34

EX+

Um

X

Znj

Znj

-

EX

Ym

Ym

Figure 1: Generic singly connected network.


Structural Analysis of Unicode Based Keyboard: Bengali Perspective From An Operational Standpoint

Using conditional independence and normalizing constant, P (X|E) = α P (Ex-|X) P (X|Ex+). To calculate P (X|Ex+), let U be the vector of U1,…, Um, and u be an assignment of values to them. Then we have, P (X|Ex+) = Σu P (X|u, Ex+) P (u|Ex+) = Σu P (X|u) Πi P (ui|Ex+), using d-separation and conjunction of independent variables. Therefore, P(X|Ex+) = Σu P(X|u)Πi P(ui|EUi\X), partitioning Ex+ into EU1\X,…, EUm\X and using the fact that EUi\X d-separates Ui from all other evidence in Ex+. Now to calculate P (Ex-|X), we find that the evidence in each Yi box is conditionally independent of the others given X. Therefore, P (Ex-|X) = Πi P (EYi\X|X) = Πi Σyi Σzi P (EYi\X|X, yi , zi) P (yi, zi|X), averaging over Yi and zi, where Zi is the parents of Yi other than X and zi is - an assignment of values to the parents. = ΠiΣyiΣziP(EYi-|X, yi , zi)P(EYi+\X|X)P(yi, zi|X), dividing EYi\X into EYi- and EYi+\X. = Πi ΣyiP(EYi-|yi)ΣziP(EYi+\X|zi)P(yi, zi|X), applying conditional independence. =ΠiΣyiP(EYi|yi)Σzi(P(zi|EYi+\X)P(EYi+\X)/P(zi)) P(yi,zi|X), applying Bayes’ rule. = Πi ΣyiP(EYi-|yi)Σzi(P(zi| EYi+\X)P(EYi+\X)/P(zi)) P(yi|X, zi) P(zi|X), rewriting the conjunction Yi, zi. = Πi ΣyiP(EYi-|yi)Σzi βi P(zi| EYi+\X) P(yi|X, zi), applying conditional independence and replacing P(EYi\+X) by normalizing constant βi Finally, multiplying the Zij’s as they are independent of each other and combining the βi’s into one big normalizing constant β, we have P(Ex-|X ) = β Πi Σyi P(EYi-|yi)ΣziP(yi|X, zi)Πj P(zij| EZij\Yi). Now we are ready to calculate P(X|E) = α P(Ex-|X) P(X|Ex+). II. STUDY AND FINDINGS After getting the patient’s identity, physicians note the age, sex, religion, locality, occupation, and economic condition of a patient. These parameters are essential because probability of many diseases depends on them. Then details about the complaints of the patient are asked. For example when a patient complaints of fever or sweating etc. Answers to these queries help the physician to differentiate the complaints into specific group of diseases. We have selected diseases with a common symptom, fever, for primary study. Most common of them are non-specific viral, malaria, visceral leishmaniasis, enteric fevers, rheumatic fever, tuberculosis, pneumonia, urinary tract infection, meningitis, dengue and tonsillitis. Then fifteen characteristics to distinguish these diseases were identified. They are – percentage among fever diseases, region, age, sex, and economic condition

dependency, magnitude, onset, duration and time dependency of temperature, presence of body pain, shivering or sweating, association of burning sensation while passing urine, soar throat or cough. To determine the network topology, we grouped the diseases characteristics into two- influence and symptom. First five of the above lists fall in the influence category and the rest fall into symptom category. Then the network for a particular disease can be expressed as in figure-2.

I1

…

I2

Im

D S1

S2

…

Sn

Figure 2: Network for a particular disease. Here D stands for the disease node, I1, I2,…, Im for influence nodes and S1, S2,…, Sn for symptom nodes. The assignment of probabilities P (D|Ii) and P (Sj|D) can be done using the experience of practicing physicians and statistical data provided by some hospitals. Medial textbooks also helped in this matter. We applied Bayes’ rule to calculate some probabilities not directly found from collected data. In this rule, we devised a simulator using some test data to determine the performance of our model. It came out that four diseases are successfully diagnosed more than 90% of the time. Success rate for two other diseases are around 70%. Less frequent three diseases are correctly diagnosed around 50% of the time. As our model used probability theory, correctness of diagnosis of two-rate disease was significant in percentage. We show the summary of 100,000 simulation runs for two diseases below. The summary shows the number of times each position (by probability value) obtained by the simulated disease in the diagnosis phase. Also average position and standard deviation are calculated shown in table-1.

Summary of 100,000 simulation runs for the disease influenza Number of times each position obtained Position 1: 96911 Position 2: 2533


37

Mohammad Shafkat Amin et al.

of Internal Medicine Vol. 108; No. 1, pp. 80-87. January 1988.

Position 3: 412 Position 4: 115 Position 5: 22 Position 6: 4 Position 7: 3 Percentage of Success: 96.911 Average position: 1.03828 Standard deviation: 0.23726 Summary of 100,000 simulation runs for the disease malaria Number of times each position obtained Position 1: 91624 Position 2: 5206 Position 3: 1839 Position 4: 590 Position 5: 331 Position 6: 243 Position 7: 102 Position 8: 58 Position 9: 7 Percentage of Success: 91.624 Average position: 1.14267 Standard deviation: 0.58454 Table 1: Summary of simulation runs for influenza and malaria. III. CONCLUSION There is a high possibility of this field in our country and other poor countries like ours, since physicians are inadequate in number and not available everywhere. If we can provide software in at least every Thana health complex in our country, it can bring a revolutionary change in the service of health sector. Besides, such software can be very much helpful for novice doctors by giving them a guideline about identifying the disease of their patients from primary complaints. It can also be used as a time saving tool for experienced and specialist physicians.

IV. REFERENCE [01] Long, W. Medical Diagnosis Using a Probabilistic Casual Network. Applied Artificial Intelligence, 3: 367-368, 1989. [02] [2] Macleod, J., Editor. Davidson’s Principles and Practice of Medicine. English Language Book Society / Churchill Livingstone. [03] [3] Russell, S.J. and Norvig, P., 1995. Artificial Intelligence: A modern Approach. Prentice Hall International, Inc. [04] [4] Szolovits, P., Patil, R.S. and Schwartz, W.B. Artificial Intelligence in Medical Diagnosis. Annals 38


Structural Analysis of Unicode Based Keyboard: Bengali Perspective From An Operational Standpoint


Structural Analysis of Unicode Based Keyboard: Bengali Perspective from an Operational Standpoint Mohammad Shafkat Amin1, Fahima Amin Bhuiyan2 and Mohammad Shyiq Amin3

ABSTRACT This study investigates the Unicode specifications and the possibilities for their implementation in Bengali Script Language. Unicode provides a unique number for every character, irrespective of the platform, program and the language. These characters cover principal written languages and symbol system of the world. The study explores an efficient way of incorporating Unicode based Bengali keyboard in contemporary operating systems. This paper systematically investigates an in depth analysis of Unicode based Bengali keyboards and its implementation aspects unfolding the details of supports provided by the operating system as well as explaining the necessary technical prospects of open type fonts. Keywords: Unicode, Glyph, lparam, Codepage, OTF.

I. INTRODUCTION Unicode provides a unique number for every character irrespective of the platform, program and the language. Fundamentally, computers just deal with numbers. They store letters and other characters by assigning a number for each one. Before Unicode was invented, there were hundreds of different encoding systems for assigning these numbers. No single encoding system could contain enough characters: for example the European Union alone requires several different encoding to cover all its languages. Even for a single language like English no single encoding was adequate for all the letters, punctuations, and technical systems in common use [1]. These encoding systems also conflict with one another, that is, two encodings can use the same number for two different characters, or use different number for the same character. Any given computer (specially servers) needs to support many different encodings, yet whenever data is passed between different encodings or platforms, the data always runs the risk of corruption [2]. This paper proposes an efficient way of implementing Unicode based Bengali keyboard in modern operating systems. This approach facilitates the linguistic computing in the local domain and enables the support of multi-lingual computing as well. II. THE UNICODE STANDARD The Unicode worldwide character standard is a character coding system designed to support the interchange, processing and display of the written texts of the diverse languages of the modern world. In addition, it supports classical and historical texts of many written languages [1, 2]. In its first 256 characters, it includes the repertoire of characters from numerous other corporate, national and international standards as well. The Principles of Unicode standards A team of computer professionals’ linguists and scholars created the Unicode to become a worldwide standard, one easily used for text encoding everywhere. To the end, it follows a set of fundamental principles. The Unicode standard is simple and consistent. It uses fixed width 16 bit character codes, and does not depend on states or modes for encoding special characters. The Unicode standard incorporates the character sets of many existing standards, for example, it includes Latin-I, character set as its first 256 characters [2]. It includes repertoire of characters from numerous other corporate, national and international standards as well. The Unicode standard uses Han unification to consolidate Chinese, Korean and Japanese ideographs. The Unicode standard allows for character composition in creating marked characters. It encodes each character and diacritic or vowel mark separately, and allows the characters to be combined to create a marked character. It provides single codes for marked characters when necessary to comply with preexisting character standard. III. UNICODE CAPABILITIES 1. Role of Unicode specification in Internationalization Journal of Computer Science, Volume1, Number 1, June 2007, ISSN : 1994‐6244

39

Upgrading Trust Factor Evaluation in AODV Protocol for MANET

Unicode is the new foundation for this process of internationalization. Older code pages were difficult to use and had inconsistent definitions for characters. Internationalizing different codes while using the same code base is complex, since different character sets with different architectures – for different markets, was needed to be supported. But modern business requirements are ever stronger; programs have to handle characters from a wide variety of languages at the same time. The EU alone requires several different older character sets to cover all its languages. Mixing older character sets together is a nightmare, since all data has to be tagged and mixing data from different sources is nearly impossible to do reliably [3]. With Unicode, a single internationalization process can produce code that handles the requirements of all the world markets at the same time. Since Unicode has a single definition for each character, data corruption problems that plague mixed code set programs, are easily avoided. Since it handles characters for all the world markets in a uniform way, it avoids the complexities of different character code architectures. All of the modern operating systems, from PCs to mainframes, support Unicode now, or are actively developing support for it. 2. Unicode specification Vs using classical character sets for application programs Different character sets have very different architectures. In many, even simply detecting which bytes from a character is a complex, contextually dependent process. That means either having ultiple versions of the program code for different markets, or making the program code is much, much more complicated. Both of these choices involve development, testing, maintenance, and support problems. These make the non-US versions of programs more expansive and delay their introduction, causing significant loss of revenue. 3. Using classical character sets for databases Classical character sets only handle a few languages at a time. Mixing languages was very difficult or impossible. In today’s markets, mixing data from many sources all around the world, that strategy for products fails badly. The code for a simple letter like “A” will vary wildly between different sets, making searching, sorting and other operations very difficult. There is also the problem of tagging every piece of textual data with a character set, and corruption problems when mixing text from different character sets [4]. 4. Bengali in Unicode The problem for Indic languages is more complex compared to European languages. Unlike Latin-based scripts, here there is a very basic distinction between a sequence of characters and the rendering of these characters in a visually acceptable form. It should be noted that Unicode says little about how to actually display a sequence of abstract characters. Unicode attempts to list all alphabetic characters (as opposed to glyphs [5, 6], the visual form representing one or more characters) used in all the major languages of the world. Different groups of languages use different sets of characters, many characters are common to more than one language (e.g., Hindi and Sanskrit use the same character set; the French character set is a superset of the English one). Unicode tries to list all these characters. Bengali characters are numbered from 0980 to 09FF (Devanagari from 0900 to 097F). The advantage of Unicode is that Unicode Bengali characters are uniquely and unquestionably Bengali. Even though the question of how they should be displayed is not clear, the intended content of the file is uniquely defined as shown in figure 1.

Figure 1: Bengali Codepage in Unicode IV. BENGALI KEYBOARD IN WINDOWS All windows in a Windows program are created based on a window class, which is a template that defines the attributes of a window. Multiple windows can be created from a single window class. The window class must be registered with Windows using the “RegisterClassEx()” Win32 API function. After a window class has been registered, it can be used to create as many windows wanted. Window classes are represented by a data structure in the Win32 API called WNDCLASSEX, which defines the attributes of a window [7]. 41 Journal of Computer Science, Volume 1, Number 1, June 2007, ISSN:1994‐6244

Muhammad Ruhul Amin Khandaker et al.

Windows communicates with application programs by sending it messages. A message has three pieces of information associated with it: • A window • A message identifier • Message parameters The window associated with a message is the window to which the message is being sent [7]. The DispatchMessage(&msg); } This code essentially processes all messages at the application level and routes them to the appropriate window procedures. Code in each different window procedure is then responsible for taking action based on the messages they receive. Windows message processing system is depicted in figure 2. In order to programmatically incorporate a Unicode based keyboard in windows, a system wide message hook has to be established. A hook is a mechanism by which a function can intercept events (messages, mouse actions, keystrokes) before they reach an application. The function can act on events and, in some cases, modify or discard them. Functions that receive events are called filter functions and are classified according to the type of event they intercept. For example, a filter function might want to receive all keyboard or mouse events.

Figure 2: Windows Message Processing System. For message identifier is a number that specifies the message being sent. The Win32 API defines numeric constants that represent each message. The message parameters consist of two pieces of information that are entirely specific to the message being sent. These 32-bit values are called wParam and lParam, and their meaning is completely determined by the message being handled [7]. When a message is delivered to a program by Windows, it is processed in the WndProc() function. Although WndProc() is responsible for handling messages for a given window class, programs must still take on the task of routing messages to the appropriate window procedures. This is taken care of by a message loop, which must be placed in the heart of the WinMain() function: while (GetMessage(&msg, NULL, 0, 0)) { TranslateMessage(&msg); DispatchMessage(&msg); } Windows to call a filter function, the filter function must be installed— that is, attached—to a Windows hook. Attaching one or more filter functions to a hook is known as setting a hook. If a hook has more than one filter function attached, Windows maintains a chain of filter functions. The most recently installed function is at the beginning of the chain, and the least recently installed function is at the end. This is depicted in figure 3. When a hook has one or more filter functions attached and an event occurs that triggers the hook, Windows calls the first filter function in the filter function chain. This action is known as calling the hook. To maintain and access filter functions, applications use the “SetWindowsHookEx” and the “UnhookWindowsHookEx” functions [7]. The following code snippet when executed establishes a windows hook.

Figure 3: The filter function chain in Windows [7]. HOOK=SetWindowsHookEx(WH_GETMESSAGE,(HOOKPROC)Key,hDLL,0); Here the WH_GETMESSAGE parameter installs a hook procedure that monitors messages posted to a message queue and the parameter Key is a pointer to the hook procedure. The zero in the last parameter specifies that, the Key parameter must point to a hook procedure in a dynamic-link library (DLL). Otherwise, this parameter can point to a hook procedure in the code associated with the current process. Journal of Computer Science, Volume1, Number 1, June 2007, ISSN : 1994‐6244 42


The WM_CHAR message is posted to the window with the keyboard focus when a WM_KEYDOWN message is translated by the “TranslateMessage” function. The WM_CHAR message contains the character code of the key that was pressed. Here the Callback function Key actually intercepts the Windows messages and checks to see whether it is a WM_CHAR message and if so it performs the actual keyboard translation and sends back a Unicode character, which can be any of the valid Unicode character code i.e. 0x995. V. OPEN TYPE FONT SPECIFICATION The OpenType font format is an extension of the TrueType font format, adding support for PostScript font data. The OpenType font format addresses the following goals: • broader multi-platform support • better support for international character sets • better protection for font data • smaller file sizes to make font distribution more efficient • broader support for advanced typographic control [8]. PostScript data included in OpenType fonts may be directly rasterized or converted to the TrueType outline format for rendering, depending on which rasterizers have been installed in the host operating system. OpenType fonts can include the OpenType Layout tables, which allow font creators to design better international and high-end typographic fonts. The OpenType Layout tables contain information on glyph substitution, glyph positioning, justification, and baseline positioning, enabling text-processing applications to improve text layout. As with TrueType fonts, OpenType fonts allow the handling of large glyph sets using Unicode encoding. Such encoding allows broad international support, as well as support for typographic glyph variants. Figure 4 represents the Glyph positioning in Bengali script development.

Figure 4: Glyph position in Bengali Alphabet [9] Whether TrueType or PostScript outlines are used in an OpenType font, the following tables are required for the font to function correctly. Tables 1,2,3,4 and 5 describe the names of these tables. Table 1: Required Table for OTF [9] Name Character to glyph mapping Font header Horizontal header Horizontal metrics Maximum profile Naming table OS/2 and Windows specific metrics post PostScript information Tag cmap head hhea hmtx maxp name OS/2

For OpenType fonts based on TrueType outlines, the following tables are used: Table 2: Tables Related to TrueType Outlines [9] Tag Name cvt Control Value Table fpgm Font program glyf Glyph data loca Index to location prep CVT Program 43 Journal of Computer Science, Volume 1, Number 1, June 2007, ISSN:1994‐6244

Muhammad Ruhul Amin Khandaker et al.

The PostScript font extensions define a new set of tables containing data specific to PostScript fonts that are used instead of the tables listed above. Table 3: Tables Related to PostScript Outlines [9] Tag Name CFF PostScript font program (compact font format) VORG Vertical Origin Table 4: Tables Related to Bitmap Glyphs [9] Tag EBDT EBLC EBSC

Name Embedded bitmap data Embedded bitmap location data Embedded bitmap scaling data

OpenType fonts may also contain bitmaps of glyphs, in addition to outlines. Hand-tuned bitmaps are especially useful in OpenType fonts for representing complex glyphs at very small sizes. If a bitmap for a particular size is provided in a font, it will be used by the system instead of the outline when rendering the glyph. There are also several optional tables that support vertical layout as well as other advanced typographic functions: Table 5: Advanced Typographic Tables [9] Tag Name BASE Baseline data GDEF Glyph definition data GPOS Glyph positioning data GSUB Glyph substitution data JSTF Justification data

Thus all the information controlling the substitution and relative positioning of glyphs during glyph processing is contained within the font itself. This information is defined in OpenType Layout (OTL) features that are, in turn, associated with specific scripts and language systems. VI. CONCLUSION This paper investigates the Unicode specification and its possibilities in terms of Bengali script language and multilingual processing. This research paper explores the internal specification of Bengali Unicode, OTF and its support in modern operating systems for Bengali script processing. Specifically the support for Bengali script processing has been investigated in terms of Windows operating system and the windows message passing architecture has been explained and its implementation aspects have also been discovered. Nowadays, as the need for multi-lingual support in operating systems have become eminent, the importance of Unicode and its implementation aspects have gained greater attention. This paper simplifies the architecture of Unicode based Bengali language processing. VII. REFERENCES [01] [02] [03] [04] [05] [06] [07]

The Unicode Consortium. The Unicode Standard, Version 4.0. Reading, MA, Addison-Wesley, 2003. 0-321-18578-1.z Memon, A.P.” Study of Unicode specifications and their implementation in Arabic script languages by designing a multilingual Unicode editor”.Multi Topic Conference, 2001. IEEE INMIC 2001. Technology for the 21st Century. Proceedings. IEEE International 28-30 Dec. 2001 Page(s):229 - 233 Bishop,Avery,F.,Brown,David,C., Meltzer, David, M. “Supporting Multilanguage Text Layout and Complex scripts with Windows 2000.” Microsoft System Journal. Becker, J.D. “Multilingual word processing.” Scientific American. 251, 1 (July 1984), 96-107. Becker, J.D. “Arabic word processing.” Communications of the ACM 30, 7 (July 1987), 600-610. Davis, M.; Collins, L. “Unicode”, IEEE International Conference on Systems, Man and Cybernetics, 1990. Conference Proceedings. 4-7 Nov. 1990 Page(s):499 - 504 Charles Petzold, “Programming Windows, The Definitive guide to the Win32 API”, 5th Edition, Microsoft Press, 2001.

Journal of Computer Science, Volume1, Number 1, June 2007, ISSN : 1994‐6244 44


[08] [09]

Mudawwar, M.F. “Multicode: a truly multilingual approach to text encoding”, Computer IEEE Journal, Volume 30, Issue 4, April 1997 Page(s):37 – 43. Microsoft, Typography, http://www.microsoft.com/typography”.


Upgrading Trust Factor Evaluation in AODV Protocol for MANET Muhammad Ruhul Amin Khandaker1, Md. Javed Kaiser2, Jugal Krisna Das3 Md. Shahid Hossain1 and Khandaker Razia Sultana3

ABSTRACT

In an ad hoc network, users communicate with each other from a temporary network, without any form of centralized administration. Each node participating in the network acts both as host and a router and must therefore be willing to forward packets for other nodes. The existing wireless routing protocols do not accommodate enough security for MANET and are highly vulnerable to attacks as malicious nodes may enter and leave the immediate radio transmission range at random intervals. To reduce this vulnerability and ensure security some researchers introduced trust levels of each node to select the neighbors. In this paper we introduce a technique for upgrading the trust levels of each node. Key words: MANET, AODV, routing, secure neighbor, malicious node, trust factor/ level.

I. INTRODUCTION Ad hoc network does not rely on any stationary infrastructure. The concept behind these infrastructures less networks is the collaboration between its participating members, i.e., instead of making data transit through a fixed base station, nodes consequentially forward data packets from one to another until a destination node is finally reached. Typically, a packet may travel through a number of network points before arriving at its destination. The operation of the system depends on distributed cooperation among all nodes in the network and fairly needs to trust the intermediate nodes between sender and receiver and the neighbors. Because of the improvised nature of ad-hoc networks, routes are built dynamically as and when nodes are regrouping (Discovery). Hence, ad hoc networks are more responsive to topology changing than any wired networks. Consequently, routing protocols for ad-hoc networks should be able to cope with link breakages and make sure that the network won’t collapse as nodes are moving or shutting down. This paper describes secure neighbor detection for mobile ad hoc networks. In particular, we employ a trust-level based technique to find the nodes which might be neighbors for the ad hoc on-demand distance vector (AODV) routing protocol, a widely adopted ad hoc routing protocol. AODV is a reactive and stateless routing protocol that establishes routes only as desired by the source node. AODV is vulnerable to various kinds of attacks [15]. This paper analyzes some of the vulnerabilities, specifically discussing attacks against AODV that related to neighbors. We propose a solution based on the trust level to upgrade trust level as a malicious node may enter in network any time and ensue it cannot be chosen as the neighbor. The remainder of this paper is organized as follows: Section –describes the secure neighbor in AODV. Section describes relative work on AODV security and intrusion detection on ad hoc networks. Section describes the upgrading technique of trust level in AODV. And finally in section we summarize the work II. SECURE NEIGHBOR DETECTION IN AODV Differentiating between a node and malicious in ad hoc networking environment is a challenging task. Malicious nodes may behave maliciously only erratically, further complicating their detection. A node that sends out false routing information could be the one that has been compromised, or merely one that has a temporarily stale routing table due to volatile physical conditions. Dynamic topologies make it difficult to obtain a global view of the network and any approximation can become quickly outdated. Therefore, secured neighbor detection is the first task for secured routing in AODV. In a network a node initiates path discovery process while it needs to communicate with another node if it does not have sufficient information. This process is accomplished by broadcasting a route request (RREQ) packet to its neighbor. After receiving a RREQ packet 45 Journal of Computer Science, Volume 1, Number 1, June 2007, ISSN:1994‐6244

A Novel Segmented Display for Arabic Numerals

a node may reply back by forwarding RREP packet or rebroadcast RREQ to its neighbors. Before rebroadcast it will increase the hop field and remaining same the destination sequence number. The previous one is done if the receiving node is destination and the next one if it is an intermediate node. For a secured neighbor detection we propose to add several security modules with the existing AODV. In AODV protocol, a secure node wishing to communicate with a destination node, first broadcasts a RREQ packet to its neighbors. Upon receipt, the destination node reply RREP packet to the source. Each node maintains only the hop information to reach to destination. The route selection criteria of AODV based upon hop count and destination sequence number. Hop count determines how short the route is, and the sequence number of the destination speaks about the freshness of the route information. Therefore, the route selection metric is clearly independent of the security level of the application and trust factor of the participating nodes. Each node maintains a local database of its neighbors, where trust factor is dynamically updated 3]. Trust factor of each node calculated from the trust level is stored in the local database. Trust level of any node is defined from an integer value. Trust value is calculated from the activities of a node when routing occurs. Every node dynamically upgrades its trust level upon observing its neighbor. This protocol allows both the initiator and target to verify that both are within their maximum transmission range. It is a simple three-round mutual authentication protocol. In 1st round, initiator sends a neighbor solicitation packet by. The target sends the reply by a solicitation packet after receiving the packet. In last round, the initiator sends neighbor verification, including broadcast authentication of a timestamp and the link from the source to destination. Now the total delay between these two subsequent messages is found by δ = t1 – t0. The distance between them (with respect to initiator) is bounded by: d ifc, Driver=pegasus, 12M Port 4: Dev 8, If 0, Class=hub, Driver=hub/3p, 12M Port 1: Dev 9, If 0, Class=HID, Driver=hid, 12M And in PROC file system: /proc/bus/usb/devices T: Bus=01 Lev=01 Prnt=01 Port=01 Cnt=01 Dev#= 2 Spd=480 MxCh= 0 D: Ver= 2.00 Cls=00(>ifc ) Sub=00 Prot=00 MxPS=64 #Cfgs= 1 P: Vendor=05ab ProdID=0060 Rev= 2.10 S: Manufacturer=In-System Design S: Product=USB Storage Adapter S: SerialNumber=2201010A0247C8AB C:* #Ifs= 1 Cfg#= 2 Atr=c0 MxPwr= 98mA I: If#= 0 Alt= 0 #EPs= 3 Cls=08(stor.) Sub=06 Prot=50 Driver=usb-storage E: Ad=01(O) Atr=02(Bulk) MxPS= 512 Ivl=125us

80


Minkowski Space Time In Theory of Relativity

E: Ad=82(I) Atr=02(Bulk) MxPS= 512 Ivl=0ms E: Ad=83(I) Atr=03(Int.) MxPS= 2 Ivl=32ms The serial number mentioned here is changed dynamically with the change of USB devices. So according to no. 4 digital identification technique we can use it for uniquely identifying a user. Software running in a daemon can trace it and make a decision to access permission or do some predefined work associated with the user. We have implemented this principal in our project.

Figure 2: USB Token plugs into the USB drive of a computer

III USB Token

USB Token offers a secure and unique serial number. The serial number is stored in the token and it cannot be downloaded on to another device. The hardware is also tamper proof. Smart keys are technologically similar to USB token but interface with computers using a USB port. USB Token deploys two factor authentication schemes in which the user has to have something (USB Token) in possession and know something (password). The token is with the client at all times and therefore stealing it is not easy. The token is protected by a serial number. If a notebook is stolen the key can be discovered, but a USB Token is more difficult to steal and it also has a unique serial number. Several companies including Rainbow Technologies offer them for commercial and government use. USB Token however only provides secure authentication they do not address the issue of confidentiality or integrity.

Table 2: Comparison of various authentication procedures

IV GROWTH OF ACCEPTABILITY: USB TOKEN


81

Md. Showkat Ali et al.

USB devices, being portable, easy to install and plug and play to use, are being well accepted and used by a wide number of computer professionals and users. The IDC market analysis shows a report that shipments of iPods and other MP3 flash memory music players will surge nearly 124 million units in 2009 4. All these USB devices can be used for generalpurpose data storage as well. According to a 2005 FBI Computer Crime Survey, 44% of organizations have reported network intrusions from within their own organizations. Technology analyst Gartner warns that portable devices containing a USB or FireWire connection are a serious new threat to businesses. In their report, Gartner named removable media devices as a significant security risk in the workplace and advised that these can be used both to download confidential data, and also to introduce a virus into the company network. Any user can come into the office, plug in a USB stick the size of the average keychain and take in/out over 32 GB of data. This poses a tremendous threat: Users can take confidential data or they can unknowingly introduce viruses, trojans, illegal software and more – actions that can affect whole network and company severely. An administrator has no way to control these social attacks until now. This means that a malicious insider can use a single USB storage to covertly take out (i.e. ‘steal’) proprietary data and millions of financial, consumer or otherwise sensitive corporate records at one go. This is why the USB token-based identification system can protect unauthorized access of unexpected USB devices to computers in organizations and protect vital business data. But this is to clarify that only USB devices through their unique tokens cannot guarantee identification of perfect owner of the devices because the device itself can be theft easily. In an Identity Fraud Survey, 2006, it is depicted that in 2005, the costs and damages caused by identity theft reached $56.6 billion 4 . However, USB token has come to be a better choice to reinforce the existing security solution mechanism due to its extreme portability and low cost issues. Anyone can be able to represent his digital signature through USB token from any corner of the world. The future probability and public acceptability of USB token-based identification mechanism 5 can be illustrated through the following data collected from online poll presented in figure 3.

Online poll result of number of users vs authentication method Number of total voter: 1750 First vote: Friday, 01 July 2005, 14:43 Last vote: Thursday, 19 April 2007, 21:31

Authentication methods

Soft Token Biometrics

A Combination PKI Keyfob USB Token Smart Card

Username/Password

0

200

400

600

800

1000

1200

Figure 3: Online poll result illustrating usage of various e-

Number of users

authentication methods.

V. DESIGN SPECIFICATION

The application developed as a security agent model is aimed to facilitate by the advantages of USB token. The basic requirements of a USB token-based security system can be identified as follows,6-15 USB standard specification awareness Robustness in data and work flow Time efficiency

82



Layered design approach Central access control mechanism Network support for remote assistance Multitasking Security warning mechanism Event logging Higher degree of Configurability VI. SOFTWARE IMPLEMENTATION

This section describes the implementation strategies of the software based on the design specification described in the last section. The workflow of the major modules of the system is described in the following sections.6-15 USB Token Registration

The system is able to register new USB tokens by detecting new USB devices when plugged in by extracting token from the USB controller and looking up the system database. Administrators can plug in new USB devices for registering it for authenticity and configure the system for the newly registered USB token. The procedure of new USB device detection is illustrated in figure 4.

Figure 4: Flowchart of authenticating a new USB device

Login Access Control

The responsibility of the login access controller is to provide an efficient and secure user identification service. It also provides a reliable service that is able to process the inputs from readers and output with the corresponding action. The login access control system is a micro module of the authentication server that acts a login access control server and runs asynchronously in the background. When a request is sent to the server along with the user USB token it looks up the database to check the authorization information entries corresponding to the given token. Based on the gathered information from the database it sends positive or

User

USB Token Manager

Access Req

Authentication Server (AS)

Login Access

Authentication Request

Journal of Computer Science, Volume1, Number 1, June 2007, ISSN : 1994‐6244 ACK Token (+ve/-ve)

Request for Login Access

83


Figure 5: Login Server negative acknowledgement to the requester. Upon the acknowledgement sent from the server, the client allows the user to login into the system. For workflow of login access controller is shown in figure 5. Process Access Control

The process control software provides user identification service as well as gives user permission to some predefined application as authorized by the administrator. Process access control is another vital micro module of the central authentication server of the implemented system that runs asynchronously in the background. On receiving request from the client it extracts the USB token from the request and communicates with the database to check permission entries corresponding to the given token. Based on the gathered information the process access controller sets permission to the requested process for the requesting user and sends positive or negative acknowledgement to the requester. The workflow of process access controller is shown in figure 6.

Proc (USB Input)

BUS (Type of Device)

Device Info

No String Search

Yes File (Devices) = Null

Generate Error

No

Yes String = Serial Number

Goto Old DB

Old DB of USB Serial Number

Compare new serial number with old DB

No New Serial Number=True

Yes

Store New Serial Number to Auth. DB

Exit

Figure 6 Flowchart of process control

84



Automated security warning mail

Security warning mail system is an essential part of the security system implemented. It runs in the background at the server layer as a mailing daemon and is another micro module of the authentication server. The logic or program access controller triggers the security warning mail system after certain number of attempts of illegal access trial as set by the administrator.

Figure 7: Mail Daemon On being triggered the mail system sends predefined mail to the administrator or system authority. The workflow of the security warning system is shown in figure 7. The security warning mail system can be used with any other authentication services that can be added to the security system according to requirement, for example, the security system can be used to monitor presence of a number of people during a certain period of time. Appropriate monitoring service module can be implemented to monitor the USB tokens requested for login access during a certain period of time and thus can identify people who have not requested any login in time and can trigger the mail daemon to send notification messages to the appropriate authority mentioning about the absent ones. The system overview is shown in the block diagram presented in figure 8.

Figure 8: System Overview


85


VII. CONCLUSION

The USB token-based security system is growing up with its potential to be effective digital certification of identification and its easy to use feature along with its portability and comparatively low cost. Though USB token cannot be the candidate of stand-alone solution for the choice of digital identification mechanism due to the possibility of device stealth but it can be a better choice for reinforcing the security system. The model of USB token-based security system that is implemented and represented in this paper thus provides a very effective and generally reachable security solution that can be extended or customized according to the requirement of any large organization in order to arrange itself with the new security technology in an affordable manner.

VIII. REFERENCES

[01] [02] [03] [04] [05] [06] [07] [08] [09] [10] [11] [12] [13] [14] [15]

Van Eck, W. Electromagnetic radiation from video display units: an eavesdropping risk? Leidschendam, The Netherlands, (1985) SafeNet, Curtailing the Piracy Epidemic, A Case for Hardware Security Key, white paper, ©2005 SafeNet, Inc. Radcliff, D. Busted! (Online) Computer World. Dec. (1998). De Trompet 2960 1967 DD HEEMSKERK, ID Control – Identity and Access Management Page 2 of 5, THE NETHERLANDS (EUROPE) Contu R. and Girard J. Put Security Policies in Place for Portable Storage Devices, Gartner. (2004) John Leyden. “Gummi bears defeat fingerprint sensors.” The Register. 16 May 2002. Jan Wolter. “A Guide to Web Authentication Alternatives.” Dec 1997. Cathy Bowen, Government Smart Card IDs: Lessons Learned, Card Technology Magazine, November (2002). Donald Davis (2002), Early Steps Toward Biometric Standards, Card Technology agazine, November. Fancher C. H., In your Pocket: Smartcards, IEEE Spectrum (February), pp. 47-53(1997). Guillou L. C., et al., The smart Card: A Standardized security Device Dedicated to Public cryptology, in G.J. Simmons (Ed.), Contemporary Crypto-logy. The Science of Information Integrity, IEEE Press, pp. 561-613(1992). Jain A, Bolle R. and Pankanti S. BIOMETRICS: Personal Identification in Networked Society. Kluwer Academic Publishers, 1999. Paul B., and Raymund E., The Smart Diskette – A universal token and personal crypto-engine, Advances in Cryptology-CRYPTO’89, Lecture Note in Comp. Sc. pp. 74-79, G.Goos and J. Hartmanis (Eds). (1989) Quisquater J-J. The adolescence of smart card, Future Generation Computer Systems, pp.13 -37. (1997) Urien P., Internet Card, a Smart card as a true Internet node, Computer Communication, 23, pp. 1655-1666(2000). © 2007 Journal of Computer Science Vol 1, No 1, June 2007, pp 78-82 Faculty of Computer Science ISSN 1994-6244

Minkowski Space Time in Theory of Relativity Md. Showkat Ali1 and Md. Aman Mahbub2

ABSTRACT

In physics and mathematics, Minkowski space (or, Minkowski space time) is the mathematical setting in which Einstein’s theory of special relativity is most conveniently formulated. In this setting the three normal dimensions of space are combined with a single dimension of time to form a four-dimensional manifold for representing a Spacetime. Minkowski space is a four dimensional real vector space equipped with a non degenerate, symmetric bilinear form. Elements of Minkowski space are called events of four vectors. Minkowski space is often denoted R1,3 to emphasize the signature, although it is denoted by M4 or simply M. It is perhaps the simplest example of a pseudo – Riemannian manifold. In this paper we discover how we can find Minkowski space time in null space and thus use it to describe the universe. Keywords: Minkowski space, theory of relativity and Riemannian manifold.

86


I. INTRODUCTION

Minkowski, Hermann (1864 - 1909), developed the concept of the space–time continuum as a consequence of special relativity. To the three dimension of space, he added the concept of a forth dimension, time. The concept developed from Albert Einstein relativity theory and become in turn, the framework for Einstein’s 1916 general theory of relativity. The characteristic feature of Galileo's Spacetime was the set of horizontal slices representing "planes of simultaneity". On a given plane, all of its events are simultaneous. This is the notion of Absolute Time, in which all observers agree on the elapsed-time between two given events. In the particular case of "zero elapsed-time", all observers agree that the events on a given horizontal plane are simultaneous. Einstein's extension of the Principle of Relativity to all physical laws requires us to abandon Galileo's Space time. In its place, we have the Einstein-Minkowski Spacetime. Hermann Minkowski developed a new view of space and time and laid the mathematical foundation of the theory of relativity. By 1907 Minkowski realized that the work of Lorentz and Einstein could be best understood in a non-Euclidean space. He considered space and time, which were formerly thought to be independent, to be coupled together in a four-dimensional 'space-time continuum'. Minkowski worked out a four-dimensional treatment of electrodynamics. In a paper published in 1908 Minkowski reformulated Einstein's paper by introducing the four-dimensional (space-time) non-Euclidean geometry, a step which Einstein did not think much of at the time. But more important is the attitude or philosophy that Minkowski, Hilbert - with whom Minkowski worked for a few years - Felix Klein and Hermann Weyl pursued, namely, that purely mathematical considerations, including harmony and elegance of ideas, should dominate in embracing new physical facts. In this view Minkowski followed Poincaré whose philosophy was that mathematical physics, as opposed to theoretical physics, can furnish new physical principles [11]. This philosophy would seem to be a carry-over (modified of course) from the Eighteenth Century view that the world is designed mathematically and hence that the world must obey principles and laws which mathematicians uncover, such as the principle of least action of Maupertuis, Lagrange and Hamilton. Einstein was a theoretical physicist and for him mathematics must be suited to the physics [10]. This space-time continuum provided a framework for all later mathematical work in relativity. These ideas were used by Einstein in developing the general theory of relativity. In fact Minkowski had a major influence on Einstein considered mathematics to be a mere tool in the service of physical intuition. With the help of the mathematician Herman Minkowski (who gave us the idea to think in terms of "Spacetime", not just space and time separately), Einstein proposed a new model for Spacetime to replace Galileo's Spacetime. "Henceforth, space by itself, and time by itself, are doomed to fade away into mere shadows, and only a kind of union of the two will preserve an independent reality." II. MINKOWSKI SPACE TIME

A main motive behind this change was the influence of two prominent German mathematicians: David Hilbert and Hermann Minkowski. Time has certainly proved Minkowski correct [3]. Murkowski’s original mathematical interests were in pure mathematics and he spent much of his time investigating quadratic forms and continued fractions. His most original achievement, however, was his 'geometry of numbers' which he initiated in 1890. Geometrie der Zahlen was first published in 1910 but the first 240 pages (of the 256) appeared as the first section in 1896. Geometrie der Zahlen was reprinted in 1953 by Chelsea, New York, and reprinted again in 1968. Minkowski published Diophantische Approximationen: Eine Einführung in die Zahlentheorie in 1907. It gave an elementary account of his work on the geometry of numbers and of its applications to the theories of Diophantine approximation and of algebraic numbers. Work on the geometry of numbers led on to work on convex bodies and to questions about packing problems, the ways in which figures of a given shape can be placed within another given figure. Minkowski developed a new view of space and time and laid the mathematical foundation of the theory of relativity. In 1905, Albert Einstein published "On the Electrodynamics of Moving Bodies". In it, he emphasized the physical experimental fact that electromagnetic phenomena do not appear to be dependent on the absolute motion of the laboratory, but just on the relative motion of the various components in the experiment. In other words, it appears that "by performing electromagnetic experiments, you cannot distinguish between absolute rest and uniform motion." However, the Maxwell Laws of Electromagnetism (which is asserted to be correct) requires that the speed of light be the same constant, c, for every inertial observer. But this (as we have seen) cannot be maintained in Galileo's Spacetime. Einstein realized this inconsistency and could have chosen either: Keep Maxwell's Laws of Electromagnetism, and abandons Galileo's Spacetime or, keep Galileo's Spacetime, and abandon the Maxwell Laws. Recall that Galileo's Spacetime embodies Galileo's Principle of Relativity and Newton's Mechanical Laws of Motion. To give up Galileo's Spacetime would mean that there was something wrong with Galileo's Relativity or with Newton's Mechanics or with both. Einstein felt so strongly about Galileo's Principle of Relativity (which applied to the mechanical laws) that he extended it to include electromagnetic and optical laws. In other words, "no physical experiment (mechanical, electromagnetic, optical---or any physical law whatsoever) can distinguish between a state of absolute rest and a state of constant velocity." According, to Minkowski, the external world is not Euclidean space of three dimension, i.e. it is not composed of points whose coordinates are (x, y, z), where x, y, z are real numbers; but it is composed of events whose coordinates are ( x1 , x 2 , x3 ) are space coordinate and the forth one x 4 involves 1. Faculty of Computer Science, IBAIS University, Dhaka-1209. 2. Dept. of Mathematics, Jahangirnagar University. The authors like to thank Mr. Shahidul Alam for his contribution.

Matiur Rahaman Mian et al.

time. If and event occurs in the space, then the position of the point where it occurs and the instant when it occurs both are represented by the location of the event in the forth dimensional continuum. All the four dimensions are not equivalent. For, an axis which measures the distance in x – direction, can be rotated to measure the distance in y and z – direction; but the same axis can not be rotated to measure the time interval. Therefore the direction of time interval is not unique. This distinction is expressed by saying that the space time continuum is 3+1 dimensional rather than as four dimensional. The Minkowski space – time ( µ ,η ) is the simplest empty spacetime in general relativity, and is infecting the space-time of special relativity. Mathematically, it is the manifold R4 with a flat Lorentz metric η . In terms of the natural coordinates

( x 1 , x 2 , x 3 , x 4 ) on R4, the metric η can be expressed in the form

ds 2 = − (dx 4 ) 2 + (dx1 ) 2 + (dx 2 ) 2 + (dx 3 ) 2 (1) These coordinates must be independent of transformation from one system to another system. If we use spherical coordinates (t , r , θ , φ ) where

x 4 = t , x 3 = r cos θ , x 2 = r sin θ cos φ 1 3 4 and x = r sin θ sin φ , then on calculations, dx = dt , dx = − r sin θ dθ + cos θ dr 2 1 dx = sin θ cos φ dr + r cos θ cos φ dθ − r sin θ sin φ dφ dx = sin θ sin φ dr + r cos θ sin φ dθ + r sin θ cos φ dφ Substituting these values in metric (1), it reduces to

ds 2 = − dt 2 + dr 2 + r 2 (dθ 2 + sin 2 θ dφ 2 ) ds 2 = − dt 2 + dr 2 + r 2 dΩ 2 where, dΩ = dθ + sin 2

2

2

(2)

θ dφ 2 .

Nothing unusual will happen to the coordinates, but we will want to keep careful track of the ranges of the other two coordinates. In this case, of course we have − ∞ < t < + ∞ , 0 ≤ r < + ∞ . Technically, the world line r = 0 represents a coordinate singularity and should be covered by a different patch, but we all know what is going on so we will just act like r = 0 is well behaved. Our task is made somewhat easier if we switch to null coordinates:

1 ⎫ u = (t + r )⎪ ⎪ 2 ⎬ (3) 1 v = (t − r ) ⎪⎪ 2 ⎭ With corresponding ranges given by,

−∞< u

Capacity and Coverage Calculation Model for the UMTS

Capacity and Coverage Calculation Model for the UMTS

Suggest Documents

Lahdekorpi replacing terrestrial umts coverage

Multihop networks for capacity and coverage ... - CiteSeerX

Mobility and Capacity Offload for 3G UMTS ... - Semantic Scholar

Capacity Calculation Methodology.pdf - Google Drive

Coverage and Capacity Issues in Enterprise

Coverage and Capacity Issues in Enterprise

A simple model for the calculation of the axial load-carrying capacity of ...

Reverse link capacity and coverage improvement for CDMA cellular ...

Impact of HSUPA Implementation on UMTS Capacity and Cell ...

Basic capacity calculation methods and benchmarking for MF ... - MIT

System Coverage and Capacity Analysis on Millimeter-Wave Band for ...

Coverage, Capacity and Interference Analysis for ... - Re.Public@Polimi

software for calculation of reservoir active capacity with the ... - SciELO

software for calculation of reservoir active capacity with the ... - SciELO

A Semi-Analytic Model of the UMTS Downlink Capacity with WWW ...

Simplified Calculation Model and Experimental

Optimizing UMTS radio coverage via base station configuration

UMTS

UMTS

Motorcycle Model Coverage (pdf)

Model Notice Creditable Coverage

calculation model

UMTS Capacity simulation study AndrÃ©s Felipe ...

Coverage Estimation for Symbolic Model Checking - CiteSeerX