A joint source-channel coding algorithm for wireless low-bit-rate speech communications

We present a new joint source-channel coding (JSCC) algorithm for improving the quality of wireless low-bit-rate speech communications over coded systems. A discrete-input analog-output additive white Gaussian noise (AWGN) channel model is proposed for deriving the algorithm. The speech parameter coder and decoder (codec) are designed by vector quantization (VQ) and optimized for the coded systems. The BCH code is applied for channel coding. An iterative soft output Viterbi algorithm (SOVA) is used for channel decoding. The soft decoding outputs are delivered for soft speech decoding directly. The proposed algorithm is applied for the low-bit-rate speech communications based on a VQ based code-excited linear prediction (VQ-CELP) speech coding scheme. Simulation results show that the proposed algorithm improves the decoded speech quality significantly by gaining 2-3 mean-opinion scores (MOS) enhancement relative to the traditional system with separate speech-channel coding with hard decoding outputs, especially over the low SNR channels