SPEECH RECOGNITION USING ANN

Abstract

In a system of speech recognition containing voice of numerical values from 1 to 0, the recognition requires the comparison between the entered voice signal of the number and the various recorded number of the dictionary. The problem can be solved efficiently by a artificial neural network algorithm whose goal is to put in optimal correspondence the temporal scales of the two numerical value. In this the artificial neural network is trained with the template which are stored in the dictionary. Because dynamic time warping algorithm is very time consuming and provide the complexity for the voice sampled data, so the highly efficient voice recognition technique artificial neural network is used. Before applying artificial neural network algorithm directly to the voice sampled date first we extract the one feature from the templates and the entered data and artificial neural network algorithm is used. The feature which is extracted in our project is known as "Mel frequency cepstral coefficients (MFCC) " Finally the comparison between thesis results and the project dynamic time warping results, other author's results is done.