Noisy channel performance of predictive trellis coded quantization

An investigation is conducted of the noisy channel performance of a low-delay, high-quality speech coder at 16 kbits/sec. The system to be studied is based on trellis coded quantization (TCQ). In the work presented, the authors demonstrate that when trellises based on feedback-free convolutional encoding circuits are employed, error propagation in TCQ systems is not a serious problem. Performance comparisons are made using a fixed predictor and four different adaptive predictors for both error-free and noisy channels. Informal listening tests reveal that the encoded speech is of good communications quality even in very noisy channels. In an ideal channel, reconstructed speech of near toll quality can be obtained from a system employing an all-pole least squares lattice predictor with pitch prediction.