Group members Luke Makischuk Abderahmane Sebaa Ameneh Sadat Yazdaninik Asma Faizi Professor: R Habash TA: Wei Yang Introduction The purpose of our project was to implement an efficient voice recognition algorithm and improve upon the idea’s from the IEEE paper: “Speech Recognition and Its Application in Voice-based Robot Control System” by Luo Zhizeng and Zhao Jinghing, August 2004 In the article, they used a technique called patterncomparison to implement the voice recognition which is very complex. We came up with a simpler way to implement the voice recognition using Matlab and we also expanded more on how you would go about controlling the robot using Bluetooth technology along with motor control of the robot. References “Speech Recognition and Its Application in Voice-based Robot Control System” by Luo Zhizeng and Zhao Jinghing, August 2004 “A dynamic-time-warp integrated circuit for a 1000-word speech recognition system” Kavaler, R.A.; Lowy, M.; Murveit, H.; Brodersen, R.W, Volume 22, Issue 1, Feb 1987 “Significance of the Modified Group Delay Feature in Speech Recognition” Rajesh M. Hegde; Hema A. Murthy; Venkata Ramana Rao Gadde; Volume 15, Issue 1, Jan. 2007 Lawrence Rabiner, Biing-Hwangluang, Fundomenrols Of Speech Recognition, Qinghua Press, 1999 http://cslu.cse.ogi.edu/HLTsurvey/ch1node7.html Work distribution Asma and Ameneh -Voice recognition research/algorithm and report writeup Abderahmane and Luke-Matlab coding Voice recognition • All done using Matlab Recognized voice commands: • Stop Turn Left Turn Right Move Forward Move Backwards Only works with one person’s voice but can easily be trained to work with anyone's voice Voice recognition How it works • • • Record voice command (Time domain) Transform into frequency domain using Fourier Transform and get the magnitude spectrum Compare spectrum of voice commands Program block diagram Voice command F.T Frequency Spectrum Compare with stored voice commands No Minimum Do they distancee? match? yes command Fourier transform Speech Signal (time domain) Frequency domain Communication with a robot The efficient way to communicate with robot is using Bluetooth. We connect the Bluetooth to the serial port at the back of the PC to transmit to the robot. For receiving, we connect a receiver module to the robot. Continue The robot receives the input commands from the Bluetooth transmitter which is a 3 bit digital signal. Then, if the input commands are matched to the program, the robot moves based on state given. Otherwise, the robot will not respond to the given command. Continue The programmed commands: Microcontroller processor command 001 010 011 100 101 110 Straight forward fast Straight forward middle Straight forward slow Turn right Turn left Straight backwards 111 Or 000 Perfect stop System diagram Motor control Results Correct Incorrect Stop 10 0 Move Forward 9 1 Move Backwards 9 1 Turn Right 10 0 Turn Left 9 1 94 % accurate Limitations The amount of words that could be recognized by our program was limited, the more words we tried adding, the less accurate it became The voice recognition program only works for the person’s voice who is trained for it Program is less accurate in noisy environments Bluetooth is an expensive technology to implement Advantages Simple algorithm used for voice recognition Easy to train a person’s voice Accurate Bluetooth is very fast and efficient Voice commands are a very easy way to control a robot Conclusion For small vocabulary applications, the method of using the Fourier transform based speech recognition is more suitable then the other complex algorithms for it is easy to implement and has a high accuracy rate. Thank you for your time any Questions?