Title:On Decoding Using Codewords of the Dual Code

Abstract:We present novel decoding schemes for hard and soft decision decoding of block codes using the minimal weight codewords of the dual code. The decoding schemes will be described for cyclic codes where polynomials can be used, however, the modification for non-cyclic codes is possible and straight forward. The hard decision decoding calculates syndrome polynomials which are the product of the received polynomial with dual codewords. Proper cyclic shifts of these syndrome polynomials are obtained and the non-zero positions are counted componentwise for these shifts. The values of this counting are a reliability measure and can be used for locating the error and also the non-error positions. This reliability measure is the basis for various variants of hard decision decoding algorithms. Decoding schemes with iterative error reduction are possible as well as information set decoding using the inherent reliability information of the measure even if there is no reliability information from the channel. Further, we will show how reliability information from the channel can be included in order to obtain soft decision decoding schemes. We derive the relation between bit flipping, believe propagation, and majority logic decoding to the novel schemes. As examples to illustrate the functioning we use BCH and Reed-Muller codes as examples for binary codes, and RS codes for non-binary codes. Besides the Plotkin construction we recall a known result that Reed-Muller codes punctured by one position are cyclic and thus, are equivalent to special cases of BCH codes. Simulation results for hard and soft decision decoding will be given for several examples and compared with results from literature. Finally, we analyze the soft decision decoding of the Plotkin construction and derive that one of the two codes uses a $3$ dB better channel (also known as channel polarization).