Word Recognition from Speech Signal using Spectrum Analysis and LPC

Abstract

In this thesis, word recognition has been done from the speech signal using spectrum analysis. Word recognition involves features extraction, preprocessing, pattern matching and decision making. To recognize the word we find word parameters from spectrum of speech signal. Here, we collect five samples of a single word and then analysis on their spectrum. To find the word parameters we analysis the word spectrum by statistical methods which are given to the some range values. These parameter values help us to recognition the word. From these values we shown that every word parameters has their own value and every parameter lying in between some bounded or range of values. Out of these range values the parameter of word has changed. Parameters finding of speech signal is the key factor to improve the recognition word. To find these parameters we use some statistical methods. And find another different parameters we use the other method which is Linear Predicted Coding (LPC) coefficients. From this method we get the new signal spectrum which are made and different from the original speech signal spectrum. So by this new spectrum analysis we find another new parameters which are help us to recognition the word. This is developed in MATLAB. So in this work, we collect the twenty five words each have five samples and analysis their signal spectrum and find various parameters. From these parameters values we recognition the word.