Adaptive Decoding of Binary Linear Block Codes Using Information Sets and Erasures

Abstract

This paper proposes an information-set based soft-decision decoding algorithm for binary linear block codes. In this new algorithm a list of symbol positions that are able to generate a valid information set is obtained, in advance, by using pre-processing. During the decoding process, the algorithm reconstructs a code-word by means of the information sets from the list. Early termination of the decoding process is based on the acceptance criterion of Barros, Godoy and Wille (BGW). If the BGW criterion was not satisfied, the candidate code-word is changed via bit exchanges (erasures) and reconstructed in order to identify the wrong bits. The performance of the proposed algorithm is quite favorable, being comparable to the maximum-likelihood decoding performance, as shown by results obtained from computer simulation. This paper also demonstrates the equivalency between the BGW criterion and the Taipale and Pursley criterion, proposed independently in 1993 and in 1989 respectively.