Reduced-complexity Soft-output Decoding of Linear Block Codes for Parallel Architectures

F. Labeau (Canada)


Communication Systems, Coding andModulation, Wireless Communications, Error-CorrectingCodes.


In this paper, we investigate a new trellis-based reduced complexity decoding strategy for linear block codes. It relies on the use of sectionalization of the trellis repre senting the code and on the application of the soft-output Viterbi Algorithm (SOVA) [1] on the resulting trellis. We investigate a modified version of SOVA, namely the Non binary SOVA (NSOVA). This algorithm has been proposed recently for nonbinary codes [2], which inherently have a more complex trellis structure. We show in this paper how a generalization of the NSOVA can be applied to the sec tionalized trellis of a binary code, by considering each sec tion of the trellis as consisting of nonbinary symbols. The corresponding symbol set can be time-varying for nonuni form sectionalizations. We also study in detail the com plexity entailed by applying this algorithm to a sectional ized trellis, and derive optimal sectionalizations for some example codes, that minimize computational complexity. We also show that the NSOVA on a sectionalized trellis lends itself naturally to a parallel implementation, and we study the computational complexity in the case of a parallel implementation.

Important Links:

Go Back