Regression Neural Network (rnn) Convolutional Code Decoder Performance In A Variety Of Communication Systems

For a long time, error correcting codes have been regarded as an efficient method which can reduce the influnce of the noises, natural and man-made interference, signal decline and other bad conditions.Convolution code is an effective and wide-used error correcting code, and the decoder methods are always being studied.The task of a convolutional decoder is to find a message sequence which is the most likely to the noisy codeword received at the receiver. The sequential and the Viterbi algorithm decoder have long been used for convolutional decoding. The Viterbi decoder is the optimal in the maximum likelihood sense among the convolutional decoders. But the complexity of the Viterbi decoder increases exponentially with the constraint length of the encoder. In order to solve this question, we need a simple and efficient decoder, which can approach and even exceed the performance of the Viterbi decoder, and not increase the complexity.Artificial neural networks (ANNs) have been widely used in digital communications because of their non-linear processing and efficient hardware implementations. The majority of these developments concentrated on the application of ANNs to implement existing algorithms. It has proved that we can decode the convolutional code by ANNs instead of the Viterbi decoder, the ANNs decoder is simple and can be easily implemented, which provides a brand-new method for decoding the convolution code.According the past research, in this paper a detailed principle and application of a 1/n rate convolution decoder based on recurrent neural network (RNN) are introduced. Experiment results have confirmed that the RNN decoder is capable of performing very close to and even better than the Viterbi decoder and works very well for some specially structured convolution codes. Particularly, decoding performance of RNN decoders are improved when simulated annealing(SA) technique has been used. Further more, we study the influence of the performance of the RNN decoder when the iteration times and the decoder depth are changed,and intruduce a new kinddecoder------analog decoder compared with the hard and soft decoders. I investigate theperformance of the RNN decoder on mobile communication, Ultra-wide-band communication and optical fiber communication systems, and plot the bit error rate (BER) curves of the RNN decoder and Viterbi decoder, and discuss the performances of the RNN decoder in the three systems.The results of the investigations can provide the theoretical guides for the application of the RNN decoder in different communication systems, which have practical values.