Packet speech transmission on ATM networks using a variable rate embedded ADPCM coding scheme

摘要

The authors propose a packet speech coding scheme to vary the code bits of an embedded ADPCM coder between 6 bits per sample, for high prediction gain speech blocks, to 2 bits for blocks representing silence. Subjective quality of 3.2 MOS on a scale of 1 to 5 at an enhancement-bit loss rate of 30% was obtained.