MULTISTAGE VECTOR QUANTIZATION FOR SPEECH ENCODING

摘要

A speech coding method and a speech coder operating according to the code excited linear prediction paradigm. A random codevector is chosen using an orthogonalization search. Said search determines an optimum random excitation gain minimizing a distortion function. The denominator of the gain function is determined using preprocessed matrices determined in advance and depending on the adaptive codevector and synthesis filter matrix in order to reduce the amount of filtering operations during the search of the random codebook.