VBR speech coding in ATM networks

ATM is one of the most important techniques for next-generation communication networks. There are some new demands on a speech codec in ATM networks. This paper proposes a new scheme of speech compression that has variable bit rates for speech communication over ATM networks. Based on a new defined parameter named distribution entropy, we classify input speech into three different types and use wavelet filters to split input signals. Subband speech is then coded by VQ. The optimal length of analysis frame and parameters of our codec in the ATM environment are also discussed.