Decodage a seuil iteratif sans entrelacement des codes convolutionnels doublement orthogonaux (French text)

The object of this thesis is to explore a new coding and iterative decoding technique without the use of interleaving and propose some practical solutions that circumvent the complexity and the latency shortcomings of the usual coding and Turbo decoding techniques.; Using threshold decoding at each iteration step, the independence constraint on the observables must be satisfied. This problem is solved by defining the new convolutional self doubly orthogonal codes which are threshold decodable. The parity check equations generated at each iteration step of the decoding procedure are then orthogonal over two consecutive iteration steps.; From the analysis of the iterative threshold decoding for convolutional self doubly orthogonal codes defined in the wide sense, some modifications of this decoding procedure improving the error performances are proposed. One of these modifications consists in the use of weighting each parity check equation by appropriate coefficients which minimize the bit error probability at the last iteration step. The simulation results show that for high signal to noise ratio, all coefficients can take the same value without degradation of error performances as compared to the use of different coefficient values. For many convolutional self doubly orthogonal codes defined in wide sense, coding gain improvements of about 0.5 to 1.0 dB are obtained using this modification. Furthermore, for convolutional self doubly orthogonal codes defined in the strict sense, at high signal to noise ratio, the use of weighting does not improve the error performances but, for lower signal to noise ratio values, substantial improvements in error performances are also possible with the use of appropriate weighting coefficients.; The error performance analysis of iterative threshold decoding without interleaving using the convolutional self multi-orthogonal codes defined in the strict sense is presented. This analysis shows that beyond the double orthogonality, only marginal improvements in error performances are obtained.; The recursive convolutional self doubly orthogonal codes are also defined. An appropriate iterative decoding method is developed and explored. Some simulation results show that the use of the recursive convolutional self doubly orthogonal codes defined in the strict sense provides very good error performances. (Abstract shortened by UMI.)...