Convolutional Codes are used in a variety of areas from computers to communications. Ideally one simply looks at a received message, which may contain errors, and decodes it into the error-free message. Unfortunately, this decoding process can be quite complicated and might not exploit the maximum error correction capabilities of the code. For these reasons neural networks have been widely used as decoders. A neural network approach for decoding of convolutional codes is studied. Here sample neural network uses simple perceptron model with one hidden layer. The training of the neural network is done using Back-propagation. A sequentially programmed Viterbi decoding algorithm is used to generate training patterns for training the neural network decoder. The performance of the trained neural network is compared with Viterbi decoding solution. The comparisons indicate that the neural network approach perform with comparable error-correcting accuracy as the Viterbi decoding algorithm.

1. Viterbi, A. J. and Omura, J. K. 1979. Principles of Digital Communication and Coding. McGraw-Hill, Inc.
2. Pless, V. 2003. Introduction to the theory of Error-Correcting Codes. John wiley and sons, Inc.
3. Bouloutas, A., Hart, G.W. and Schwartz, M. 1991. Two extensions of the Viterbi algorithm. IEEE Transactions on Information Theory. 37
4. , 430-436.
5. Fettweis, G. and Meyr, H. 1989. Parallel Viterbi algorithm implementation: breaking the ACS-bottleneck. IEEE Transactions on Communications. 37
6. , 785-790.
7. Yegnanarayana, B. 2003. Artificial Neural networks. PHI, New-Delhi.
8. Huang, S.-C. and Huang, Y.-F. 1991. Bounds on the number of hidden neurons in multilayer perceptrons. IEEE Transactions on Neural Networks. 2
9. , 47-54.

Convololutional Codes Back Propogation Viterbi Decoding