Simplified trellis decoding of block codes by selective pruning

Error correcting codes are of paramount importance for reliable communications. By adding redundancy to the transmitted data they allow the decoder to detect and correct errors. However in favorable channel conditions, a part of this redundancy can be removed in order to increase throughput. Unfortunately most coding schemes are poorly adapted to these higher coding rates. For example, the decoding of block codes grows exponentially with code length. In this thesis we propose a novel solution to this problem: selective trellis pruning.Selective trellis pruning reduces decoding complexity by removing certain codewords from the trellis. This reduction is accomplished by making hard decisions on the values of bits in the received sequence above the certainty threshold. This method can produce near-optimal results with only a fraction of the operation required by full decoding thanks to the reduced trellis size. In this work we also introduce an innovative way of obtaining the pruned trellis directly from a simplified version of the generator matrix. By using this method we avoid the long process of constructing and then pruning the full trellises, thus making the selective trellis pruning algorithm an efficient decoding tool. Finally we apply this algorithm to the parallel concatenated turbo block code decoder in order to reduce its complexity