Unbalanced Multiple Description Coding by adding Enhancement Layer for AMR-WB Speech Codec in VoIP

摘要

To minimize the degradation of the speech quality caused by the packet loss of the real-time speech communication like the Voice over Internet Protocol (VoIP) over the internet environment, we introduce the Unbalanced Multiple Description Coding (UMDC) of Adaptive Multi-rate -^sWideband (AMR-WB) speech coder by adding Enhancement Layer. The UMDC asymmetrically splits the bitstream of the AMR-WB coder into two sub-streams, a high bit-rate substream and a low bit-rate sub-stream. The sub-streams are transmitted through different network paths. If only one substream arrives at the decoder, a little degraded speech quality is obtained. The enhancement layer which consists of the pulse position, the pulse sign and the gain extracted from the speech codec improves the speech quality. The enhancement layer is transmitted along with a low bit-rate sub-steam. It prevents the degradation of the speech quality slightly. The performance is tested by the use of WPESQ, the voice quality indicator, in the Gilbert two state Markov model. We see that the proposed method improves the speech quality a little than and the UMDC which doesn't add the enhancement layer.