Multiple description speech coding for robust communication over lossy packet networks

Robust speech communication over unreliable channels is one of the key research areas in the development of voice-over IP (VoIP) technology. The authors propose a multiple description coding (MDC) based speech packetization scheme to combat packet losses. The basic idea is to encode each input speech frame into multiple packets, each of which can be independently decoded. Explicit redundancy is added such that each packet can render an acceptable signal reconstruction of the original frame. Unlike previous approaches using explicit redundancy for loss recovery (D.E. Tsoukalas et al., 1997), we propose to improve the redundancy coding efficiency using context adaptive techniques. Simulation results on independent packet losses show that the proposed scheme gives better average reconstruction audio quality at low loss rates (/spl les/20%) compared to that of previous works.