Speech Recognition using Neural Network

Abstract

Speech recognition is a subjective phenomenon. Despite being a huge research in this field, this process still faces a lot of problem. Different techniques are used for different purposes. This paper gives an overview of speech recognition process. Various progresses have been done in this field. In this work of project, it is shown that how the speech signals are recognized using back propagation algorithm in neural network. Voices of different persons of various ages in a silent and noise free environment by a good quality microphone are recorded. Same sentence of duration 10-12 seconds is spoken by these persons. These spoken sentences are then converted into wave formats. Then features of the recorded samples are extracted by training these signals using LPC. Learning is required whenever we don’t have the complete information about the input or output signal. At the input stage, 128 samples of each sentence are applied, then through hidden layers these are passed to output layer. These networks are trained to perform tasks such as pattern recognition, decision making and motoric control. Key words Neural network, speech recognition, back propagation, training algorithm. 1.