|
In speech coding algorithm , mixed excitation linear prediction (MELP) algorithm at low rates because a better simulation of natural language features to synthesize high -quality voice , a modern low-rate speech coding algorithms most potential . In this paper, MELP algorithm as the basic model, the pitch period , gain, Fourier amplitude voicing decision and the line spectrum frequency (LSF) , and several other speech coding parameters for a detailed analysis and experimental , in-depth study of the low rate encoding algorithm the choice of parameters , extraction and quantification method to design and implement a 1.2kb / s low -rate speech coding algorithm is a multi- frame joint , three vector quantization parameter interpolation methods to reduce the encoding rate , etc. , the quality of the synthesized speech conducted a performance evaluation . Informal subjective listening tests show that the speech quality speech coding algorithms have higher intelligibility and clarity , and has a certain degree of naturalness .
|