Generalised and Channel Independent SVM Based Robust Decoders for Wireless Applications

Emerging applications in wireless communications and Software Defined Radio require robust and generalized decoders with a very good efficiency. This paper aims at introducing a novel and powerful method of implementing a decoder using Support Vector Machines (SVM) to exhibit good performance irrespective of the channel model. The method proposed also ensures a generalization in the design of decoder, which can be easily adaptable for any type of coding technique used. In addition, this method overcomes the demerits of the traditional decoders like Viterbi and other decoders using Neural Networks. The error correction codes like Hamming and Convolutional codes are considered for experimentation. Using SVM, which is a class of machine learning algorithm, this process is viewed as a multi-class classification problem and error correction is achieved in a simpler way. An extensive analysis with regard to the effect of channel and modulation techniques is also made and presented. The proposed SVM model is sufficiently cross validated and found to be an effective replacement for the existing counterparts.