Interative BEAST-decoding of product codes

This paper presents a low-complexity iterative algorithm for decoding concatenated block codes, in particular, product codes. Component decoders in the iterative scheme are based on BEAST?Bidirectional Efficient Algorithm for Searching code Trees. BEAST finds a list of the most probable codewords and uses them to compute approximate a posteriori symbol probabilities. The proposed decoder has significantly lower complexity than the trellis-based BCJR-type decoder. The decoding complexity depends on the trellis complexity of the component codes, and, with the proper design of the constituent codes, it can be kept sufficiently low. It is shown that the iterative BEAST-APP decoder outperforms the iterative schemes that employ algebraic-type decoders to find the candidate codewords. The bit-error-rate performance and the complexity of the proposed decoder are illustrated by simulation results.