Abstract: Communication is the fundamental purpose of speech, i.e., the transmission of messages. Speech is the most common form of interaction between human to human which is very natural and efficient. It conveys the information about words, speaker identity, expression, emotion, gender of the speaker. Now-a-days, both speech and speaker recognition are the two most important research areas. In this paper we try to recognize Assamese spoken words. This paper describes feature extraction technique Linear Predictive Coding (LPC) , Mel Frequency Cepstral Co-efficient (MFCC) and for classification Artificial Neural Network(ANN) are used to recognize Assamese speech samples (words).We create a small database for seven Assamese isolated words which consist of 16 speakers of equal numbers of male and female. Each word is uttered by seventy five times by each speaker.

Keywords: Speech Recognition, LPC, MFCC, DTW, HMM, Neural Network.