Message Passing-Based Decoding of Convolutional Codes: Performance and Complexity Analysis

Abstract

In this letter, we propose to apply message passing algorithms to decode standard convolutional codes and assess the resulting performance and the required complexity compared to conventional decoding algorithms for convolutional codes by concentrating on the Viterbi algorithm (VA). We show that, in contrast to the VA for which the decoding complexity increases exponentially with m, the number of memory blocks for the proposed framework, such an increase, is only linear in m. This suggests that applying message passing algorithms can provide considerable savings in the required computational power if it can also exhibit a comparable bit-error-rate performance to that of the VA. In this letter, we show via simulations that this is in fact the case for convolutional codes.