Integration of Speaker and Speech Recognition Systems

摘要

Speech is the most important and primary mode of communication and also the most natural and efficient form of exchanging information among humans. This paper presents a detailed study of textdependent Speaker and Speech Recognition system. Speaker recognition system uses vector quantization (VQ) as the modeling technique while features of the speech signal are extracted using Mel Frequency Cepstum Coefficients (MFCC). K-means clustering algorithm has been used to obtain the vector quantized codebook. For Speech recognition system, formant frequencies of the word sample are used to determine the unknown word. Speaker recognition system yields highest accuracy with Hanning window and Mel perceptual feature extraction realized with 35 filter bank. Accuracy also improves as the number of vectors in the VQ codebook is increased from 64 to 100 whereas for Speech recognition, highest accuracy obtained is 95%.