Design considerations for iteratively-decoded source-channel coding schemes

We address outer variable-length encoding of a first-order Markov source in a serially concatenated coding scheme with an inner recursive convolutional code. Decoding is carried out iteratively between the constituent decoders for variable-length and convolutional code. While variable-length codes are commonly known to be sensitive to transmission errors, we show that they can lead to significant performance improvements compared to fixed-length source encoding with optimized mappings. Specifically, we propose a simple variablelength code construction with a free distance of two and good compression properties at the same time. Numerical results show that the performance gain of the proposed approach also holds for precoded ISI channels where iterative joint source channel equalization and decoding is employed at the receiver.