Classification of Speech Based-DyWT and Its Application
Qing Feng Wu · 2005
A new algorithm for the classification of speech based on Daubechies8 DyWT was proposed. The parameters for speech classification by using the wavelet coefficients of the scales were calculated, and each frame of speech signals was classified into five classes by a finite-state model which is compatible with SMV and can be used in SMV. The result of simulation shows that the SMV based on this speech classification can achieve variable rate coding with high quality of speech, and resolves the problem of tradeoff in speech quality and frequency bandwidth.