Loss Concealments for Low Bit-rate Packet Voice Table of Contents

In recent years, packet voice over the Internet has become an attractive alternative to conventional public telephony. However, the Internet is a best-effort network, with no guarantee on its quality of service. A fundamental issue in real-time interactive voice transmissions over an unreliable Internet Protocol (IP) network is the loss, or late arrival of packets. This problem is especially serious in transmitting low bit-rate coded speech when pervasive dependencies are introduced in a bit stream, leading to errors propagated to subsequent frames when a packet is lost. As a result, the study of loss-concealment schemes is important in ensuring high-quality playback. In this research, we choose an end-to-end loss-concealment approach that requires no special support from the underlying network. In particular, we focus on developing a nonredundant sender-receiver collaborated multiple-description coding (MDC) scheme. We propose a new coder-dependent parameter-based MDC. Based on the correlations of coding parameters, we generate multiple descriptions systematically. In particular, we have observed high inter-frame correlations in linear predictors of low bit-rate coders, but low correlations in excitation parameters. Hence, we generate multiple descriptions at iii the sender side by interleaving linear predictors, and reconstruct lost ones at the receiver side by linear interpolations. In addition, we replicate hard-to-reconstruct excitation parameters to multiple descriptions. We also require our design to be done in such a way that does not require extra transmission bandwidth and that can adapt its number of descriptions to network-loss conditions dynamically. We have compared three linear predictor representations, Reflection Coefficient, Log Area Ratio, and Line Spectral Pair (LSP), by their reconstruction qualities, and have found that the LSP representation performs the best. Finally, we have studied LSP reconstructions and enhancements on excitation quality. Extensive tests on FS CELP, ITU G.723.1, and FS MELP under different loss scenarios have shown good quality and reliability of our proposed scheme.