Vol-2 Issue-4 2016 IJARIIE-ISSN(O)-2395-4396 2845 www.ijariie.com 101 A RESEARCH PAPER ON ISOLATED SPOKEN WORD RECOGNITION USING HIDDEN MARKOV MODEL 1 Ratna Priya Kanchan, 2 Dr.Manoj Soni 1 Student, Dept. of MAE, IGDTUW, 2 Associate Professor, Dept. of MAE, IGDTUW ABSTRACT The main aim of the project is to develop an isolated spoken word recognition system using Hidden Markov Model (HMM) and that will be implemented on a 3R robotic arm with a good accuracy at all the possible frequency range of human voice. Here different command words like left, right, ready, up, down etc. are recorded by the speaker which can be male or female and results are compared with different feature extraction methods. The spoken word recognition system is mainly divided into two major blocks. First part includes recording data base of the command signals and feature extraction of those recorded signals. Here we use Mel frequency cepstral coefficients and fundamental frequency as feature extraction methods. To obtain Mel frequency cepstral coefficients signal we have to go through the following: framing, applying window function, Fast Fourier transform, filter bank and then discrete cosine transform. Keywords: HMM, Feature Extraction, Word recognition . 1. INTRODUCTION Natural form of communication for human beings is human voice. It is a speech signal which contains a sequence of sounds. Speech is the ability to express thoughts and feelings by articulate sound. Voice signals are generated by nature. Since they are naturally occurring hence are random signals. There are several models put forth by researchers based on their perception of voice signal. Speech recognition is a very challenging problem on which a lot of work has been done. Most of the successful results have been calculated and obtained using Hidden Markov Models explained by Rabiner in 1989 [1]. A speech recognizer would efficiently enable more efficient and accurate communication for everybody, but especially for children, analphabets and people with disabilities. A speech recognizer also acts as a subsystem in a speech-to- speech translator. The isolated speech recognition system implemented during the project trains Hidden Markov Model for each to be recognized. The models are trained well with labeled training data, and the classification is performed by passing the features of each of the model and then selecting the best match. Further the result will be implemented on a 3R robotic arm, which performs the work as directed. It is given commands such as “left, right, forward” and many more throug h a Bluetooth module and Arduino. The commands are spoken to an android application through which the robot listens and acts accordingly.