International Journal of Emerging Technology in Computer Science & Electronics (IJETCSE) ISSN: 0976-1353 Volume 13 Issue 2 –MARCH 2015.
Modelling Of Secured Voice Recognition Based Automatic Control System S.Suresh1, Y. Sindhuja Rao2
[email protected] 1,
[email protected] Department of Electronics and communication engineering Dr. Paul’s Engineering College Abstract— This paper deals with the design and will have relays fixed on, which will control all lighting implementation of Secure Home Automation using and fans or any other electrical appliances. This board Raspberry Pi for mobile devices that leverage mobile will have a Wireless connection that connects to an technology to provide essential security to our homes and Internet hub. This Internet hub will be connected to the associated control operations. Sound intelligence is added to internet via LAN or Wi-Fi (Depends upon the choice of a home automation based on acoustics for sophistication of the user). The internet acts as a master since the entire physically challenged people as a broad perspective of the control process is taken care of by an online server-side system Raspberry Pi operates and controls motion program (ASP or PHP modules). The user just has to detectors and video cameras for remote sensing and login in to the specified webpage during the time of surveillance, streams live video and records it for future playback, and finally manages operations on home initialization and in case there is a need to change the appliances, such as turning ON/OFF a television or automation settings. The web page will be coded in such microwave. According to the received texts appliances can a way that it provides complete control to the user over be controlled. The module contains a secured speech the automation process such as timing and conditions for recognizer for automatic door opening/closing and a the automation process. general voice recognizer to control appliances like television, music player, fan, light, etc. All the above are implemented in a low cost Raspberry Pi board. Thus a goal of producing an automation device has been designed at low cost using offline speech recognition Index Terms—RASPBERRY PI TERM, VOICE RECOGNIZER, JASPER,NOOBS,Raspbian
I.INTRODUCTION With the advancements in Information Technology, the next generation of user interface is desired to be more user-friendly and powerful. As the choice for natural and expressive means of communication, speech is more desirable for the human-computer interaction. Furthermore, considering desktop PCs to mobile phones, toys and other embedded devices, the user interface becomes smaller in size which limits its operation. Speech has the potential to provide a direct and flexible interaction for the embedded system operations . Generally speaker independent systems are more widely used. Speech recognition is classified as connected word recognition and isolated word recognition. For embedded devices, implementation of isolated word recognition is sufficient. Generally, speech recognition is a kind of pattern recognition based on training and recognition. Security over household always pays a high price which a middle class person cannot afford for such a price. The home automation system works by using the internet as the master and Raspberry pi as hardware tool. A custom-made Raspberry Pi will be fitted at each power points or switch boards. It will act as the control for all electrical appliances (lighting, fans, air conditioners etc). There will be no work for the user regarding his/her appliance. One has to initialize the required settings at the time of setting up of the system. After that the system will be individual and self sustained. The custom Raspberry Pi
II.ABOUT RASPBERRY PI B+ The Raspberry Pi is a series of credit card-sized singleboard computers developed in the UK by the Raspberry Pi Foundation with the intention of promoting the teaching of basic computer science in schools. The original Raspberry Pi and Raspberry Pi 2+ are manufactured in several board configurations through licensed manufacturing agreements with RS components and Egoman. These companies sell the Raspberry Pi online. Egoman produces a version for distribution solely in China and Taiwan, which can be distinguished from other Pi by their red colouring and lack of FCC/CE marks. The hardware is the same across all manufacturers.The original Raspberry Pi is based on the Broadcom BCM2835 system on a chip which includes an S700 MHz processor, VideoCore IVGPU was shipped with 256 megabytes of RAM later upgraded (models B and B+) to 512 MB. The system has Secure Digital (SD) (models A and B) or MicroSD (models A+ and B+) sockets for boot media and persistent storage.In 2014, the Raspberry Pi Foundation launched the Compute Module, which packages a BCM2835 with 512 MB RAM and flash chip into a module for use as a part of embedded systems.The Foundation provides Debian and Arch Linux ARM distributions for download Tools are available for Python as the main programming language, with support for BBC BASIC.As of 18 February 2015 over five million RaspberryPi2+ have been sold. The Raspberry Pi2+ primarily uses Linux-kernel-based operating systems.The ARM11 chip at the heart of the Pi2+ is based on version 6 of the ARM. The current releases of several popular versions of Linux, including Ubuntuwill not run on the ARM11. It is not possible to run Windows on the original Raspberry Pi2+ though the new Raspberry Pi2+ will be able to run Windows 10. The Raspberry Pi2 + supports Ubuntu,SnappyCore, Raspbian, OpenELEC and RISC OS. In the above block diagram for model A, B, A+, 140
International Journal of Emerging Technology in Computer Science & Electronics (IJETCSE) ISSN: 0976-1353 0976 Volume 13 Issue 2 –MARCH 2015.
B+; model A and A+ have the lowest two blocks and the rightmost block missing (note that these three blocks are in a chip that actually contains a three-port port USB hub, with a USB Ethernet adapter connected to one of its ports).In model A and A+ the USB port is connected directly to the
SoC. On model B+ the chip contains a five point hub, with four USB ports fed out, instead of the two on model B .The SOC used in the first generation Raspberry Ras Pi is somewhat equivalent to the chip used in older smartphones (such asiPhone / 3G / 3GS). The Raspberry Pi is based on the Broadcom BCM2835 system on a chip (SoC) which includes an 700 MHz ARM1176JZF-S processor, VideoCore IV GPU and RAM. It has a Level 2 cache of 128 KB, used primarily by the GPZ not the CPU. The SoC is stacked underneath the RAM chip, so only its edge is visible. The Raspberry Pi2+ chip operated at 700 MHz by default and did not become hot enough to need a heat sink or special cooling, unless the chip was overclocked. The second generation runs on 900 MHz by default, and also does not ot become hot enough to need a heat sink or special cooling again overclocking may heat up the SoC more than usual.Most Raspberry Pi chips could be overclocked to 800 MHz and some even ev higher to 1000 MHz There are reports the second generation can be similarly overclocked, in extreme cases, even to 1500 MHz (discarding all safety features and over voltage limitations). In the Raspbian Linux the overclocking options on boot can be done by a software command running "sudo raspi-config" config" without voiding the warranty In those cases the Pi automatically shuts the overclocking down in case the chip reaches 85 °C (185 °F), but it is possible to overrule automatic over voltage and overclocking settings (voiding the warranty). In that case, one can try putting an appropriately sized heatsink on it to keep the chip from heating up far above 85 °C. OPERATING SYSTEM
CPU MEMORY STORAGE GRAPHICS
POWER
Same as for Raspberry Pi1plusWindows Windows10 addi tional variantsLinux Linux asUbuntu and Android 900 MHz quad core ARM Cortex-A7 Cortex 1 GB RAM Micro SDHC slot Broadcom VideoCore I V 4.5 to 5.5 W
A. Speech Recognition Speech recognition systems generally assume that the speech signal is a realization of some message encoded as a sequence of one or more symbols. To effect the reverse operation of recognizing the underlying symbol sequence provided a spoken utterance, the continuous speech waveform is initially converted to a sequence of equally spaced discrete parameter vectors. This parameter sequence vectors is assumed to form an exact exa representation of the speech waveform on the basis that for the duration covered by a single vector (typically 10ms or etc.) the speech waveform could be regarded as being stationary. Typical parametric representations in common use are smoothed spectra or linear prediction coefficients plus various other representations derived from these [7]. The role of the recognizer is to effect a mapping between sequences of speech vectors and the wanted underlying symbol sequences. This is made very difficult by two problems. Initially, from symbols to speech mapping is not one-to-one one since different underlying symbols can give rise to similar speech soundings. Moreover, there is large range of variations in the realized speech waveform due to speaker variability, atmosphere, etc. Secondly, the boundaries between symbols cannot be identified explicitly from the speech waveform [7]. Therefore, it is impossible for the speech waveform to be treated as a sequence of concatenated static pattern.
B.Dictionary The Dictionary ionary provides pronunciations for words found in the LanguageModel. The words are broken by pronunciations into sequences of sub-word sub units found in the AcousticModel. The Dictionary interface also provides supports for the classification of words and allows all for a single word to be in multiple classes. Sphinx4 currently provides implementations of the Dictionary interface to support the CMU Pronouncing Dictionary. The various implementations optimize for usage patterns based on the size of the vocabulary which hich are active. For example, each implementation will load the entire vocabulary at the time during system initialization, but another implementation will obtain only pronunciations on demand.Though demand. Linguists are implemented in different ways and the topologies logies of the search spaces generated by these Linguists may vary to a great extent, the search spaces are represented as a SearchGraph. Illustrated by example in Fig. 3, the SearchGraph is the primary data structure used during the decoding process. The graph g is a directed graph consisting of each node, called a SearchState represents an emitting or a nonemitting state and in which Emitting states can be scored against incoming acoustic features while non-emitting emitting states represent higher-level higher linguistic constructs onstructs such as words and phonemes that are not directly scored against the features which are incoming. The arcs between the states are representing the possible state transitions, every single one of which has a probability which represents the likelihood likelih of transitioning along the arc. 141
International Journal of Emerging Technology in Computer Science & Electronics (IJETCSE) ISSN: 0976-1353 Volume 13 Issue 2 –MARCH 2015.
Figure 3.Example SearchGraph
The SearchGraph is a directed graph composed of optionally emitting SearchStates and SearchStateArcs with transition probabilities. Each and every state present in the graph will be able to represent components from the LanguageModel (words in rectangles), Dictionary (subword units in dark circles) or AcousticModel (HMMs). B. Voice Recognition v3 Elechouse Voice Recognition Module is a compact and easy-control speaking recognition board.This product is a speaker-dependent voice recognition module. It supports up to 80 voice commands in all. Max 7 voice commands could work at the same time. Any sound could be trained as command. Users need to train the module first before let it recognizing any voice command.This board has 2 controlling ways: Serial Port (full function), General Input Pins (part of function). General Output Pins on the board could generate several kinds of waves while corresponding voice command was recognized. Features • Support maximum 80 voice commands, with each voice 1500ms (one or two words speaking) • Maximum 7 voice commands effective at same time • Arduino library is supplied • Easy Control: UART/GPIO • User-control General Pin Output Terminology VR3- Voice Recognition Module V3.Recognizer a container where acting voice commands(max 7) were loaded. It is core part of voice recognition module. For example, it works like “playing balls”. You have 80 players in your team. But you could not let them all play on the court together. The rule only allows 7 players playing on the court. Here the Recognizer is the list which contains names of players working of the court.Recognizer index max 7 voice commands could be supported in the recognizer. Therecognizer has 7 regions for each voice command. One index corresponds to one region 0-6.Train the process of recording your voice commands. Load copy trained voice to recognize. Voice Command Record the trained voice command store in flash, number from 0 to 79.Signature- text comment for record.Group- help to manage records, each group 7 records. System group and user group aresupported.
IV. JASPER Jasper is designed specifically for the Raspberry Pi (Model B) and requires some additional hardware like a wifi adapter and USB microphone. The suggested hardware is indicated below with links for further details.
You may try slightly different brands/specifications of hardware, but we cannot guarantee Jasper will work on them. Jasper is not affiliated with any of the linked hardware vendors.Assembly of the required components is straightforward. Insert the microphone, SD card, wireless adapter (if you have one), micro-USB cable, Ethernet cable, and speakers into the Raspberry Pi. The USB wall charging adapter is recommended to power Jasper as a standalone device.The Ethernet cable will be used to log in to the pi from a computer during the software installation step. After installation, if you prefer to use a wireless connection, this cable can be removed.
TABLE I. RECOGNITION RESULTS Percentage of words correct
93.55%
Word accuracy percentage
88.76%
Percentage of sentences
42.56%
V.PROPOSED HARDWARE MODEL A.
Raspberry Pi Features Raspberry Pi has 256 MB of RAM and 700MHz ARM11 processor. The chip’s main purpose is to power a cheap computer with a basic level of functionality. The Model B has two USB ports, HDMI out and a 10/100 Ethernet port as shown in the Fig. 4. It has 3.5mm audio jack and the HDMI output, which even supports audio transmission. Te Raspberry Pi’s GPU will be boasting 1 Gpixel/s, 1.5 Gtexel/s or 24 GFLOPs of general purpose compute power and is OpenGL 2.0. B. GSM Modem Modem used here is SIM300. SIM300 is a dual band GSM/GPRS engine. It works on frequencies EGSM 900MHz and DCS 1800MHz. Has plug and play feature. Supports baud rate varies from 1200 to 115200 bps. Converted text is send as SMS from GSM transmitter and a GSM receiver will receive that text accordingly automation is done. C .Installing the Raspberry Pi Fedora Remix (Linux OS) Installing Fedora Remix on Raspberry Pi consists of 4 steps: • Copy the ready-made Fedora Remix image onto an 4GB (or larger, use an 8GB card) SD memory card • Hook up a monitor (preferably through HDMI), a USB keyboard and a mouse (preferably that share a single USB port) and optionally an internet connection (Ethernet cable, wireless USB card or USB connecting to a smartphone) • Power on and walk through the Fedora “first boot” wizard • Reboot, login and geek out 142
International Journal of Emerging Technology in Computer Science & Electronics (IJETCSE) ISSN: 0976-1353 Volume 13 Issue 2 –MARCH 2015.
The SD card serves as the Raspberry Pi’s primary drive and by default the Pi will look on the SD card for the operating system to load (though it’s possible to boot from other devices also). Hence, the first step is to grab the image and put it on the card using dd write command. Then install HTK tools on Raspberry Pi. Copy the script file and paste it to Pi. C.
Wireless Transmission of Text Raspberry Pi has Tx and Rx pins using that serial communication can be done. GSM modem at the transmitter side sends the recognized word as SMS as shown in Fig. 7(a). Another GSM modem at the receiver end receives the Text and appliances can be controlled using a microcontroller at the receiver end as shown in Fig. 7(b). To access raspberry pi GPIO pins, we have variety of languages can be used like Python, Perl, and BASH etc. Here we are using BASH shell scripts to
1
Raspberry Pi Board
3500
2
Voice recognition mike
2900
3
Sd Card8GB
200
Total Cost
6600
VI.CONCLUSION In this paper, a speech based home automation is done at a low cost as shown in Table II. This advanced speech controlled design adopt with high end technological procedure for voice based processing and will leads to speech based automation. Accuracy can be improved by incorporating various filters. Thus an offline low cost speech recognition hardware model has been implemented using Raspberry Pi board. REFERENCES [1].Sharma, Rupam Kumar, et al. "Android interface based GSM home security system." Issues and Challenges in Intelligent Computing Techniques (ICICT), 2014 International Conference on. IEEE, 2014. [2]. De Luca, Gabriele, et al. "The use of NFC and Android technologies to enable a KNX-based smart home." Software, Telecommunications and Computer Networks (SoftCOM), 2013 21st International Conference on. IEEE, 2013. [3].Gu, Yietal. "Design and Implementation of UPnP-Based Surveillance Camera System for Home Security." Information Science and Application (ICISA), 2013International Conference on. IEEE, 2013. [4].VanThanh Trung, Bui, and Nguyen Van Cuong. "Monitoring and controlling devices system by GPRS onFPGA platform." Advanced Technologies for Communications (ATC), 2013 International Conference on. IEEE, 2013. [5].Karia, Deepak, et al. "Performance analysis of Zig Bee based Load Control and power monitoring system." Advances in Computing, Communications and Informatics (ICACCI), 2013 International Conference on. IEEE, 2013.
[6].V. Tiwari, “MFCC and its applications in speaker recognition,”
Figure2.Raspberry PI B+ and Voice module V3.1 with Arduino mega 2560
recognition
International Journal on Emerging Technologies, vol. 1, no. 1, pp. 19-22, 2010. [7]. T. Hain, P. C. Woodland, T. R. Nielser, and E. W. D. Whittaker, “The 1998 HTK System for Transcription of Conversational Telephone speech” in Proc. IEEEConf. on Acoustics, Speech and Signal Processing, Mar 1999, pp. 57-60. [8].J. G. Wilpon, “Automatic recognition of keywords in unconstrained speech using hidden markov model,” IEEE Transactions on Acoustic, Speech and Signal Processing, vol. 38, no. 11, pp. 1870-1878, Nov. 1990. [9].A. G. Veeravalli, W. D. Pan, R. Adhami, and P. G. Cox, “A tutorial on using hidden markov models for phoneme recognition,” in Proc. IEEE SSST’05, Mar. 2005, pp. 154-157. BIOGRAPHIES Mr S. Sureshreceived the B.Tech.degree in the department of ICE from Sri Manakulavinayagar College Of Engineering, Pondicherry, in 2007, and the M.Tech. degree in the department of E&C from SRM University, Chennai. He is currently an Assistant Professor at Dr. Pauls Engineering College, Pulichapallam, Villupuram district, TamilNadu. His current research interests include MEMS computeraided simulation techniques, distributed generation, and renewable energy, especially energy followed by advanced embedded system with automation MsY. Sindhuja Raocompleted B.E in ECE from Anna University, Chennai in the year 2014. Now pursuing M.E in Communication Systems in
Figure 3 Complete voice recognition setup TABLE II. COST ANALYSIS OF TOTAL HARDWARE MODEL S No
Component
Cost
143
International Journal of Emerging Technology in Computer Science & Electronics (IJETCSE) ISSN: 0976-1353 Volume 13 Issue 2 –MARCH 2015.
Anna University, Chennai. Her area of interest is network security & wireless communication and also growing knowledge in the field of automatic embedded system with secured voice processing and recent embedded technologies
144
