Development of Intelligent Speech-Recognition System Using Wavelet Transform and Neural Network

The pre-processing of speech signals is considered an essential stage in the development of a robust speech recognition system. First part of the paper will report on a comparison of two pre-processing techniques with end-point detection in order to investigate their contribution to the accuracy and speed of the recognition system; these techniques are low-pass filter and wavelet transform respectively. The second part of the paper reports on the implementation of melfrequency cepstral coefficient algorithm on the preprocessed speech signal in order to extract essential features suitable for training using neural network. The last part of the paper reports on how neural network is developed and trained based on the extracted features for recognition purpose.