Adaptive iterative decoding for expediting the convergence of unary error correction codes

摘要

Multimedia encoders typically generate symbols having a wide range of legitimate values. In practical mobile wireless scenarios, the transmission of these symbols is required to be bandwidth efficient and error resilient, motivating both source coding and channel coding. However, Separate Source and Channel Coding (SSCC) schemes are typically unable to exploit the residual redundancy in the source symbols, which cannot be totally reduced by finite-delay, finite-complexity schemes, hence resulting in a capacity loss. Until recently, none of the existing Joint Source and Channel Codes (JSCCs) were suitable for this application, since their decoding complexity increases rapidly with the size of the symbol alphabet. Motivated by this, we proposed a novel JSCC referred to as the Unary Error Correction (UEC) code, which is capable of exploiting all residual redundancy and eliminating any capacity loss, while imposing only a moderate decoding complexity. In this paper, we show that the operation of the UEC decoder can be dynamically adapted, in order to strike an attractive trade-off between its decoding complexity and its error correction capability. Furthermore, we conceive the corresponding Three Dimensional (3D) EXtrinsic Information Transfer (EXIT) charts for controlling this dynamic adaptation, as well as the decoder activation order, when the UEC code is serially concatenated with a turbo code. In this way, we expedite iterative decoding convergence, facilitating a gain of up to 1:2 dB compared to both SSCC and to its non-adaptive UEC benchmarkers, while maintaining the same transmission bandwidth, duration, energy and decoding complexity.