Multiple Descriptions With Symbol Based Turbo Codes Over Noisy Channels With Packet Loss: Design and Performance Analysis

Channel optimized multiple description (MD) vector quantization with symbol-based turbo codes are proposed, where the decoder exploits both nonuniformity of descriptions and their dependencies. The joint source channel coding strategy is devised by setting an entropy constraint for quantizer design, which together with the MD index assignment (IA), control the level of redundancy at the MD source coder output. This is in turn exploited by the source and channel decoders for robust transmission in presence of noise and packet loss. At the source coder, the IA is designed for an M-description vector quantizer using an efficient simulated annealing algorithm. Various numerical results and comparisons are provided to investigate the effect of different system and design parameters on the performance. The results specifically indicate improved performance in comparison to the prior art, and also in multiple input multiple output communications.