Based On Matching Pursuits Sinusoidal Modeling Of Speech Coding

Speech coding use as few bits as possible while retaining high quality, which is a tradeoff. Now the middle and high rate speech coding is developed as state of art, but the low rate with high quality speech coding is still a open problem which attracts much researchers, the sinusoidal modeling is the main method employed. Matching pursuit is developed as a adaptive signal decomposing techniques in 1993 and is found very effective and valuable in various application, recently attracts speech coding researchers's focus.This thesis try to deal with the low rate with high quality speech coding by employing matching pursuit based on sinusoid modeling. We decompose the speech signal into few sinusoid modeling bases with matching pursuit and output the parameters, which condense the information of speech signal. Furthermore the parameters is compressed losslessly with adaptive Huffman coding to obtain higher ratio. A fast matching pursuit method based on speech pitch is presented, which gets the same compress ratio and much low computing complexity. The validity of this method is verified by the experiment.