Distance Correction System for Localization Based on ... - IEEE Xplore

30th Annual International IEEE EMBS Conference Vancouver, British Columbia, Canada, August 20-24, 2008

Distance Correction system for Localization Based on Linear Regression and Smoothing in ambient intelligence display Dae-Hee Kim, Jae-Hun Choi., Myung-Eun Lim Soo-Jun Park Abshoct- This paper suggests the method of correcting distance between an ambient intelligence display and a user based on linear regression and smoothing method, by which distance information of a user who approaches to the display can he accurately output even in an unanticipated condition using a passive infrared VIR) sensor and an ultrasonic device. The developed system consists of an ambient intelligence display and an ultrasonic transmitter, and a sensor gateway. Each module communicates with each other through RF (Radio frequency) communication. The ambient intelligence display includes an ultrasonic receiver and a PIR sensor for motion detection. In particular, this system selects and processes algorithms such as smoothing or linear regression for current input data processing dynamically through judgment process that is determined using the previous reliable data stored in a queue. In addition, we implemented GUI software with JAVA for real time location tracking and an ambient intelligence display.

I. INTRODUCTION It becomes possible to realistically express real world through innovative technology development of IT. Location tracking system associated with context-aware computing is the representative technology relates to this. In this paper, we developed the location tracking system for an ambient intelligence display to improve the quality of aged life. An ambient intelligence system refers to a system for identifying a user who approaches to the system and adaptively presenting image or music in synchronization with the distance from a user. "CareNet display" [I], "digital family portraits" [Z], "Interactive Public Ambient Displays" [3] and etc were developed as the research that it relates to ambient intelligence display and interaction. In this system, it is important to accuratelyrecognize the distance kom a user in a periodic manner. However, since users approach to the system in a variety of patterns, and the patterns are not always constant, it may be difficult to recognize a user. In addition, it may be impossible to identify a user or recognize the distance kom a user when there is any obstacle in the middle of the path a user approaches or noises are abruptly generated. Therefore, in the field of an ambient intelligence system, This work was supported by the lT R&D program of MKWIlTA [2006-S-007-01, Ubiquitous Health Monitoring Madule and System Dwzlopmznt] Dae-Hee Kim Author is with the Electronics and Telecommunications Research Institute LifcInfamatics Team (amail : [email protected]) Jae-Hun Choi Author is with the Electronics and Telecommunications Research Institute LifeIdamatics Team (*mail :[email protected]) Myung-Eun Lim Author is with the Electronics and Telecommunications Research Institute LiCeInfo1111atics Team (e-mail : [email protected]) Soo-Jun Park Author is with the Electronics and Telccmmications Research Institute LifeIdamatics Team (*mail : [email protected])

978-14244-1 815-2/08/$25.00 02008 IEEE.

there have been needs for a means for accurately identifying a user and recognizing the distance form a user even in an unanticipated condition such as obstacles, noises, or various motions of a user. Particularly, in the case of the "Interactive Public Ambient Displays", it is the method that it divides 4 steps into interaction between a user and a display according to the distance and provides services. But there are some problems to identify multi-user. In this paper, we propose the distance correction system of an ambient intelligence using linear regression and smoothing algorithm using a PIR sensor based on ultrasonic system. The ultrasonic system applied to TF (Time of Flight) algorithm for the application of the system was implemented. The method to be proposed determines whether a current data is normal, distrust, and abnormal state or not based on the previous reliable data stored in a queue and applies the compensation method that it fits for each case and moving possible regions of the aged within one cycle were modeled for state judgment of a current data. Moreover, the method for the identificationof multiuser and GUI for real-time location tracking system was developed 11. RELATED WORK The position awareness technology which looked for the location of an object and recording it had been being mentioned kom many systems. In the indoor location tracking system there are the Active Bedge[4] system using the inkared ray cellular proximity, the Active Bat[S] method that is distance measurement system using moving time of ultrasonic wave, the RADAR[6] using 802.11RF scene analysis and triangulation, the EasyLiving[7] system using a vision and triangulation. These systems select i&ared ray, ultrasonic wave, WID, UWB, signal strength, and etc, and are appropriately implemented according to a situation. Moreover, the research about the predictive algorithm application for improving the reliability of a system is continuously in progress. For example, the adaptive window prediction algorithm [8] was used based on the previous action video kame in order to predict the velocity of a car, and the system was developed by applying the adaptive compensation method [9] using the Doppler Effect in the mobile satellite commnnication system. The location tracking technology applied to the ambient display has to provide accurate distance data and robust against the obstacle or the noise betweenamultiuser and the display. Specially, because the system for the ambient intelligence display based on location tracking supports the daily life and will improve the quality of life, the continuous research that it can enhance the system reliability is needed.

1443

111. SYSTEM DESCRIPTION A. Hardware Implementation

Gcncrally, bccausc an ambicnt intclligcncc display has thc characteristic that users interact with it, It is important that to recognize the location of users with accuracy. The proposed system used the ultrasonic sensor of 40 kHz for location tracking. Because ultrasonic wave moves at the speed of 343.5mls having a direction, we can find the distance and angle using the arrival time and speed of ultrasonic wave. Fig. 1 is an implemented hardware illustrating ambient intelligence display with two ultrasonic sensors and a PIR sensor, and the ultrasonic transmitter, and the gateway that is an actuator controlling the system. Ultrasonic sensors of the receiver attached bottom part of both sides in the ambient intelligence display, and the passive infrared sensor for motion detection is located at a center. The receivable range of the ultrasonic sensor is 6m with an angle -40" 40" horn center and an ATMEGA128 MCU(Micro controller unit) and CClOlO chips were used to process sensor data and wireless comm~~nication, and an AN213 1 chip was used for USB communication between the gateway and a computer. The sensing range of a passive inhared sensor is the 10m. FIG. 2 is the hardware block diagram illustrating the module of an ultrasonic receiver and a transmitter.

-

L

FIG. 1 implemented hardware (ambient intelligence display, ultrasonic transmitter, gateway horn counterclockwise)

FIG. 3 implemented GUI & system operation. B. System Operation Sequence Fig. 4 shows how we can obtain the distance and the angle information hom the implemented hardware. First, the gateway receives an order from a computer through USB port and it synchronizes the ultrasonic receiver and transmitter using RF (Radio Frequency) communication. Second, the ultrasonic transmitter emits ultrasonic wave and the receiver part calculate the time difference between consecutive ulti-asonic sensors. Third, the PIR sensor data is obtained to distinguish whether or not the user moves, the PIR sensor value is given "1" when it detect a user, but the PIR sensor value is given "0" when it fails to detect a user. Fourth, the PIR sensor value is delivered to the gateway and the DOA (Direction of Arrival; it means an angle value) and the distance value are computed by TF (Time of Flight) algorithm. We adapted the round robin method to process multiple users because it takes 140ms to process horn 1 to 4 operation of the fig4. In case of a period in the system was set 1 second, It was possible to a process of the location tracking till 7 person. Wireless communlcaton

3 PIR sensor data acqulsltl

2 Calculat~onof time d~ffer$ce ,' between consecut~veu l t r a s o n ~ ~ s e n s ~ ~ s

\

(11.)

FIG. 2 hardwarc block diagram Fig. 3 is a whole system, if the user who had the transmitter move to the ambient intelligence display, (GUI) graphic user interface shows the location of the user and the ambient intelligence display provide services according to the interaction zone.

FIG. 4 flowchafi of localization using ultrasonic and PIR IV. METHOD A. TF (Time of Flight) Algorithm

AyyB

smoothing process may be performed by calculating the distance and the angle based on two pieces of the most previous data stored in the queue and the current data using the equation 3

C

angle

length

C

FIG. 5 triangulation model for computation of distance and angle

Calculated Data = 1 ; (Queue Dataend

Because the algorithm based on the TF using a time diffcrcncc bctwccn consccutivc ultrasonic scnsors, cach sensor need a synchronization process by RF (Radio Frequency) and the TF algorithm has the advantage of identification for multiple users because it is possible to gives a user's ID to a transmitter using the wireless packet of RF signal. Fig5 shows the triangulation model for the computation of distance and angle. Uppercase A and B are ultrasonic sensors of receiver and C is an ultrasonic transmitter. Lowercase c is the distance between ultrasonic sensors of receiver. The angle value of A, B, C can be found if the distance value of a, b, c are known. The distance (length) and the direction of arrival (angle) of a user are calculated by equation 1 and 2.

+ Queue Dataend + Current Data)

,

... '. -

_...

(3)

'

FIG. 6 human moving region

I coord I

distance

I

1 coord 1

angle

distance

1

angle

1

Table 1 distance and angle interval for region

B. Algorithm We applied a queue to implement the proposed method. Fig 6 illustrates a data flow in a queue. When the current data are input, existing data are shifted from left to right, and the rightmost data beyond the size of the queue is discarded. In our system, data stored in each queue is comprised of the PIR sensor value, the distance value, and angle value. Direction + Current Data

Queue Dataad.,

Queue Data,,$z

Queue Data,,$3

Queue Data,,$d

+

FIG. 6 Data flow in a queue The proposed system determines doubt status comparing the previous reliable data with a current data and we regard doubt status as the state that noise exists. Whether or not noise exists is determined based on the human moving region of fig.7. As shown fig 7, current location and relational region are displayed ret dot point and deviant crease lines. We divided a whole region into several parts according to the distance and the angle interval like a table 1. If current location is not located in the previous region or neighboring regions of it then it is determined that there is noise, a smoothing process is carried out for the data stored in the queue and the input data to output the current distance and angle data. ~h~

Then, the calculation result is stored in the queue. If it is determined that the current distance and angle data are in an abnormal condition, it is determined whether the PIR sensor value detected by the PIR sensor is "1" or "0" As a result, if it is determined that the PIR sensor value is "I", that is, if there is motions of a user, the current user's distance and angle values are predicted using a regression process In this case, data sequence pairs (1, Q~eueData,,~.~), (2, Que~eData,,~.~), (3, QueueDataend),and (4, QueueDataend) may be used. Coefficients a, b, and c of . y = c ~ . y ~ + b x +are obtained fiom a set of the four sequence pairs based on the least square method. The coefficients may be obtained by modeling Y = A X , squaring both sides to obtain a matrix A - + for minimizing y- AX, and then differentiating with respect to the matrix A. As a result, equation 3 may be obtained as follows

- -

ae2=- E "* Z E [ TAT] t -

E [ YX"] A

-

[

~ = o]

- *--

-yxT1 - ( E E xx ~]

AE[XX~]

= E.'[

1-I

where, A denotes a matrix, and X and Y are vectors. If equation 3 is applied to four pieces of data, equation 4 may be obtained as follows: A =

[A e(ya] [+*${=a] N, (4) =1

when equation 4 is applied to four pieces of data, it is defined

M

N-4 and . The size of the matrix A is 1 x 3, and their values denote the coefficients a, b, and c, respectively, of equatioi~.~'~~.~-2A + bnew ~ + ~value C' of y is predicted by applying a given value of x, (i.e., 4) to equation.y=fl.~2+bx+c The distance and angle data are predicted in this way, and they are output as the current user's distance and angle data. In addition, the output user's distance and angle data are stored in the queue. On the other hand, if the current data are abnormal, and zeros are repeated successively n or more times in the PIR sensor value, it is determined that there is nobody in front of the display. Subsequently, both the current user's distance and angle data are set to 0, and that value is stored in the queue. In this case, the value of n is a variable that can be set and unit to the regressi0n before a user perfectly disappears>and is to predict the current distance and angle. Fig. 7 is a flowchart illustrating the method of correcting the distance in an ambient intelligence display using a linear regression and smoothing method.

(srAm)

suitable for applying in the range within 5m. Figure 9 shows data value which is an algorithm predicted as the application result about the situation where it is unable to receive data (distance value: 0) and suddenly changing data (noise case).

Figure 9. TF algorithm & proposed method (distance) VI. CONCLUSION In this paper we developed the Distance Correction system for Localization Based on Linear Regression and Smoothing method in an ambient intelligence display. We implemented hardware system using an ultrasonic device through RF communication and developed software for location tracking. Especially, this system could process 7 users pel- second. References

di8tanos & angls data, and PIR sensor data

[I]

[2]

\ /

distance dangle through regresaon YES

[3] Ston cunent dlstance a angle data and PIR 6mSOr value in queue

dlstance &angle data to zem

TERMINATE

\-)

Fig. 7 flowchart of the proposed method V. EXPERIMENTAL RESULT

[4]

[5]

We measured distance and angle fiom center of the ambient display to user who had ultrasonic transmitter for location tracking experiment. The average error showed up as 0.27m and 6.28" when applying the TF algorithm. [6]

[7]

[8]

Figure 8. Experimental data of distance and angle (unit : meter & degree) As shown in Figure 8, it was seen that an error increases as a distance increases. Therefore, we can know the fact that it is

[9]

Consolvo, S., Soessler, P., and Shelton, B., 2004, The CareNet Display: Lessons Learned from and In Home Evaluation of and Ambient Display, Ubicomp 2004, pp. 1--17, Springer-Verlag Berlin Heidelberg Elizabeth D. Mynatt , Jim Rowan , Sarah Craighill , Annie Jacobs, "Digital family portraits: supporting peace of mind for extended family members", Proceedings of the SIGCHI conference on Human factors in computing systems, p.333-340, March 2001, Seattle, Washington, United States Daniel Vogel ,Ravin Balakrishnan, Interactive public ambient displays: transitioning from implicit to explicit, public to personal, interaction with multiple users, Proceedings of the 17th annual ACM symposium on User interface software and technology, October 24-27,2004, Santa Fe, NM, USA Roy Want ,Andy Hopper , Veronica FalcZo , Jonathan Gibbons, The active badge location system, ACM Transactions on Information Systems (TOIS), v. 10 n. 1, p.91-102, Jan. 1992 Andy Harter, Andy Hopper, Pete Steggles, Any Ward, and Paul Webster, The anatomy of a context-aware application. In Proceedings of the 5th Annual ACMIIEEE International Conference on Mobile Computing and Networking (Mobicom 1999), pages 59 (68, Seattle, WA, August 1999. ACM Press. Paramvir Bahl and Venkata Padmanabhan. RADAR: An in-building RFbased user location and tracking system. In Proceedings of IEEE INFOCOM, volume 2, pages 775-784, March 2000. S. Shafer, et al, The New EasyLiviilg Project at Microsoft Research, Proceedings o r the 1998 DARPA I NIST Smart Spaces Workshop, July 1998, pp.127-130 Tun-Wen Pai, Wen-Jung Juang and Lee-Jyi Wanh, An Adaptive Windowing Prediction Algorithm for Vehicle Speed Estimation, IEEE Intelligent 'lransportation Systems Conference Proceedmgs, 2001 Moon-Hee You, Seong-Pal Lee and Youngyearl Han, Adaptive Compensation Method Using the Prediction Algorithm for the Doppler Frequency Shift in the LEO Mobile Satellite Communication System, ETRI Journal, December 2000