Blind Channel Coding Identification of Convolutional encoder and Reed-Solomon encoder using Neural Networks

Forward error correcting (FEC) codes are used to improve the reliability of digital communication systems. They introduce redundancy in the signal which helps the receiver to correct errors without requesting for re transmission. FEC codes can be classified into two categories: Convolutional codes and linear block codes (LBCs). Reed-Solomon (RS) codes lie in the category of LBCs. For non-cooperative communication applications such as adaptive modulation and coding (AMC), military applications and cognitive radio, the channel encoder has to be identified blindly for decoding the received signal. In this study, we propose a scheme for blind identification of convolutional and RS codes. We have used the pattern recognition properties of a neural network (NN) to identify the encoder from a candidate set. NNs have not been used for this purpose, to the best of our knowledge. Performance of the proposed classifier has been evaluated for both the noiseless and the noisy case. To show the application of the proposed approach we present the performance results for the two most common use cases namely the terrestrial wireless and the satellite communication channels. Experimental results have shown that the proposed classifier can identify the encoder with high accuracy in low signal-to-noise ratio.