Publications-Detail

Joint Softbit Source and Channel Decoding

Authors:
Vary, P.
Book Title:
Proceedings of Aachen Symposium on Signal Theory
Pages:
p.p. 25-33
Address:
Aachen, Germany
Date:
Sep. 2001
Language:
English

Abstract

Source codecs and channel codecs for speech-, audio-, and video-transmission are not perfect. After source encoding there is still residual redundancy in the transmitted bitstream due to complexity and delay constraints, and at the receiving end channel decoding may produce residual bit errors in adverse channel conditions. Therefore, error concealment is needed to reduce the subjectively annoying effects. In this tutorial contribution a generalizing approach to error concealment is described which is based on joint softbit source and channel decoding. At the receiver the residual redundancy is exploited to support parameter estimation in terms of Softbit Source Decoding (SD) instead of conventional parameter decoding by hardbit table lookup. The proposed method requires bit reliability information provided by the demodulator or by the equalizer or specifically by the channel decoder in terms of softbits and can exploit additionally a priori knowledge about codec parameters. Softbit Source Decoding can be improved by adding artificial redundancy at the transmitter using channel block codes such as parity check codes. Here these codes are not decoded explicitly but they support the SD parameter estimation process by increasing the a priori knowledge. Further improvements can be achieved by replacing the conventional block codes by a new class of (nonlinear) channel codes, which has been called Source Optimized Channel Codes (SOCCs). These codes are designed by optimizing a quality measure in the parameter domain instead of maximizing the Hamming distance, taking into account the source and channel statistics. The subjective quality can significantly be enhanced, with a desired inherent muting mechanism or graceful degradation behavior in case of a degrading channel.

Download

BibTeX