Efficient reliability based decoding of linear block codes

The use of linear block codes is a well-known error-control technique for reliable communication over noisy channel. Good linear block codes can provide significant coding gain. However the best codes are not used widely in practice for lack of an efficient soft-decision decoding algorithm. How to tradeoff the decoding performance and complexity is an important aspect of a soft-decision decoding algorithm. Maximum likelihood (ML) decoding of a linear block code provides the optimal performance, but it is an NP hard problem. Consequently many sub-optimal soft-decision decoding algorithms have been proposed. Most reliable basis (MRB) based algorithms provide a promising tradeoff between the decoding complexity and performance.; This research investigates approaches to improve the performance of MRB based algorithms. Multiple reliable basis based algorithms are proposed to efficiently enlarge the searching list of the algorithm iteratively. Each reliable basis is constructed by randomly biasing the received vector. Tradeoff between decoding performance and complexity can be achieved by appropriately choosing the biasing amplitude and the iteration number.; To reduce the average complexity, a probabilistic sufficient condition (PSC) is proposed. The miss event and the false alarm event are analyzed and the associated probabilities are calculated. Simulation results show that the proposed PSC can efficiently reduce the average complexity with negligible performance degradation.; We then develop an enhanced box and matching algorithm (EBMA). By constructing a control band which is error free with high probability, we enhance the matching capability of the original BMA. The performance of BMA of order ( i+1) is nearly achieved with a linear increase in complexity and no increase in memory with respect to BMA of order i.; We finally apply MRB based algorithms to certain type of memory channel. With an interleaver used in the transceiver, we jointly decode a number of codewords by constructing a common reliable basis. Simulation results show that error performance is improved at the price of delay due to interleaving.