Iterative convolutional decoders based on neural networks applications

A mathematical model of a K/n rate conventional convolutional encoder/decoder system was developed and the neural networks technique, based on the gradient descent algorithm, was applied to decode the received signal. Using the general expression for the energy function, needed for the recurrent neural networks decoding, the expressions for the gradient decent updating rule are derived and the neural network decoder was designed. The developed theory is demonstrated on an example 2/3-rate code. It was shown that the proposed technique can achieve the coding gain similar or better that that achieved by the Viterbi algorithm while preserving the possibility of parallel processing.