ON THE DESIGN AND ANALYSIS OF CORRELATED VECTOR QUANTIZATION FOR SPEECH SIGNAL

摘要

In this Paper, we provide a correlated vector quantize scheme for compression encoding the speech signal by exploiting the redundance existing in the two correlated analysis parameters of speech frame. We discuss the correlated vcctor quantization of parameters of RPELP(Regular Pulses Excited Linear Predictive) model of the speech signal. We study the statistical correlation of predictive coefficients and the exciting pulse sequence, and provide some algorithms for designing a CVO.Our simulation result shows that CVQ can compress the bit rates further without degrading the reconstructed speech quality obviously.