Vector Quantization (VQ) Based Speech Recognition System on TMS320C6713DSK

The details of implementation of speech recognition system on the TMS320C6713 processor based DSK is discussed. This recognition system is based on the vector quantization technique at the acoustical level. There is a need to find out a new method of speech recognition system in which it should have less number of computations in computing the Minimum Euclidean Distance. It is also required to discover a new algorithm for forming a codebook in order to achieve a good Success rate. The efficiency of the implementation of this system lies in choosing an efficient method for MFCC vector generation and VQ codebook generation. The number of MEL filters used in the implementation is optimal and the method of design of MEL filters is efficient to keep MFCC feature vector length small to reduce the overall computational effort during training and testing phases.