Magnitude-phase coding of base-band speech signals

Base-band vocoders have important potential applications for digital speech transmission because they allow the encoding of good and natural quality speech at bit rates in the range 4.8-9.6 kbps. In this paper, we propose a new coding scheme for base-band vocoders which is aimed at reducing the bit rate, while preserving the quality of compressed speech. This approach improves the compression ratio over existing techniques by taking advantage implicitly of the quasi-harmonic structure of voiced speech. In the first part of the paper, we sum up the main features of a particular base-band vocoder (voice excited predictive coder-VEPC), and we briefly restate the sub-band coding technique used to encode the base-band signal. In the second part of the paper we show how to improve the coding efficiency of the base-band signal thanks to an original magnitude-phase coding technique. We then discuss the implementation of this technique with complex quadrature mirror filters (CQMF), and the separate quantization of the magnitude and phase signals by use of differential coders. Finally, we report simulation results on a complete vocoder implemented with the proposed technique.