A VECTOR QUANTIZATION APPROACH TO ISOLATED-WORD AUTOMATIC SPEECH RECOGNITION

Abstract

The aim of this research is to develop an Isolated-Word Automatic Speech Recognition (IWASR) System based on Vector Quantization (VQ) approach. This system receives speech inputs from users, analyzes the speech inputs, searches and matches the input speech with the pre-recorded and stored speeches in the trained database/codebook, and returns the matching result to the users. Developing this system is meant to assist customers calling a university’s telephone operator to respond to their enquiries in a fast and convenient way using their natural speech. Callers are assisted using their own speech inputs to select their language preference, faculty in a university and finally select the staff name they wish to contact. To extract features from the speech signals the Mel-Frequency Cepstral Coefficients (MFCC) algorithm was applied. Subsequently, Vector Quantization (VQ) algorithm based on the principle of block coding was used for all feature vectors generated from the MFCC algorithm. A codebook was resulted from training the VQ initial codebook and experimental results showed that the recognition rate has been improved with the increase of codebook size. Simulation results showed that the codebook size of 81 feature vectors had a recognition rate exceeded 85%.