Multi-Layer Optimization for Voice Coding over Ad Hoc Networks

In this paper, we explore the capabilities of a multi-layer collaboration to enhance real-time voice transmission over perturbed wireless ad hoc networks when multiple paths can be considered. To take advantage of the path diversity provided by the MSBR multi-path routing protocol, Multiple Description Coding separates the bitstream of a speech coder in two or more complementary descriptions, each sent on different paths. We propose a cross-layer design which allows the network and application layers to co-adjust their behaviour in order to gain in robustness against packet losses. We examine the voice transmission in a network where the speech codec and the network layer dynamically collaborate versus a classical approach in which each layer operates independently. Simulation results under different network conditions show that the collaboration between the network and application layers allows the receiver to maintain superior speech quality in packet loss conditions.