Speech Compression Based On Best Wavelet Packet Analysis

Vast amounts of redundant information contained in speech signal. It is a very impotent to research this topic, how to improve compress rate more and the distortion is less. This paper researches speech compression algorithm based on wavelet packet.It not only decompositions low parts of the signal, but high parts of signal also be decomposed. Realize integrity analysis to signal. It overcome toneless in high parts reconstruction. Respect to Shannon entropy type criterion, the best wavelet packet can change with speech change, so this algorithm can raise the adaptability to signal. Another work was studied: how to select wavelet filter? The result is: when we compress speech signal with best wavelet packet, wavelet filter DB6 is enough: this algorithm's wavelet filter, is not long than the wavelet filter DB10. So this algorithm lowering the calculation complications, Analysis the experiment conclusion: this algorithm is compromising compression rates and computation complicate;and the reconstruction speech compressed 16 times. Reserves person's characteristic, and so much, been understood, clearly and naturely.