Direct Polarization for q-ary Source and Channel Coding

TL;DR

This paper extends polar coding to q-ary sources and channels, providing a recursive likelihood ratio method for decoding, demonstrating improved performance over existing strategies, especially with large alphabets.

Contribution

It introduces a recursive LR vector-based decoding method for q-ary polar codes, simplifying implementation and improving performance compared to previous multilevel approaches.

Findings

Successive cancellation decoding is straightforward with LR vectors.

Complexity is quadratic but manageable with parallelization.

Performance approaches theoretical limits for large alphabets.

Abstract

It has been shown that an extension of the basic binary polar transformation also polarizes over finite fields. With it the direct encoding of q-ary sources and channels is a process that can be implemented with simple and efficient algorithms. However, direct polar decoding of q-ary sources and channels is more involved. In this paper we obtain a recursive equation for the likelihood ratio expressed as a LR vector. With it successive cancellation (SC) decoding is applied in a straightforward way. The complexity is quadratic in the order of the field, but the use of the LR vector introduces factors that soften that complexity. We also show that operations can be parallelized in the decoder. The Bhattacharyya parameters are expressed as a function of the LR vectors, as in the binary case, simplifying the construction of the codes. We have applied direct polar coding to several sources and channels and we have compared it with other multilevel strategies. The direct q-ary polar coding is closer to the theoretical limit than other techniques when the alphabet size is large. Our results suggest that direct q-ary polar coding could be used in real scenarios.