Multiple Description Coding and Path Diversity for Voice Communication over MANETs

We compare a single description coder (G.729) over a single path (SD), a new multiple description coder based on G.729 (MD-G.729) with path diversity (MD-PD), and a duplicated full-rate single description coder (G.729) with path diversity (DSD-PD) under various packet loss conditions for voice communication over wireless mobile ad-hoc networks (MANETs). We show that for low bitrate speech codecs, using a multiple description coder is not very advantageous because the large packet overheads overshadow the small bandwidth savings. Instead, we can use simple path diversity wherein the full rate single description codec is duplicated over independent paths (DSD-PD). Such a method requires only a slightly higher bandwidth than MD-PD but the quality of speech delivered is significantly better when compared to MD-PD. We compare the three different communication methods under random and bursty packet loss conditions on the basis of the quality of the delivered speech. We evaluate the delivered speech quality using the objective speech quality measurement algorithm, PESQ