Source-controlled block turbo coding

Abstract

This thesis proposes the use of block turbo codes for transmitting nonuniform binary memoryless sources over AWGN channels. As opposed to standard convolutional turbo codes, block turbo codes are known to achieve good performance for high code rates, up to 98% of the channel capacity for uniform sources. The objective of this thesis is to investigate if such a performance is also achievable for non-uniform sources. The idea is to use a joint source-channel (or source controlled) coding scheme to exploit the statistics of the binary source at the decoder. ☐ Block turbo codes are a class of product codes, which can be decoded by iteratively decoding the component codes. Each component code is decoded using the Chase algorithm, which is a low-complexity suboptimum algorithm [8] for near Maximum-Likelihood decoding of linear block codes. In this thesis, we propose decoding modifications to take into account the non-uniform nature of the source in the decoding of block turbo codes.