Duality of Channel Encoding and Decoding - Part II: Rate-1 Non-binary Convolutional Codes

TL;DR

This paper introduces a low-complexity, linear-approach decoder for rate-1 non-binary convolutional codes that matches the error performance of traditional BCJR MAP decoding, utilizing dual encoders and FFT for efficiency.

Contribution

It presents a novel, linear complexity decoding method for rate-1 non-binary convolutional codes based on dual encoders, reducing computational complexity significantly.

Findings

Achieves error performance comparable to traditional BCJR MAP decoding.

Reduces computational complexity from exponential to linear.

Utilizes FFT to further enhance decoding efficiency.

Abstract

This is the second part of a series of papers on a revisit to the bidirectional Bahl-Cocke-Jelinek-Raviv (BCJR) soft-in-soft-out (SISO) maximum a posteriori probability (MAP) decoding algorithm. Part I revisited the BCJR MAP decoding algorithm for rate-1 binary convolutional codes and proposed a linear complexity decoder using shift registers in the complex number field. Part II proposes a low complexity decoder for rate-1 non-binary convolutional codes that achieves the same error performance as the bidirectional BCJR SISO MAP decoding algorithm. We observe an explicit relationship between the encoding and decoding of rate-1 convolutional codes in $GF(q)$. Based on this relationship, the BCJR forward and backward decoding are implemented by dual encoders using shift registers whose contents are vectors of complex numbers. The input to the dual encoders is the probability mass function (pmf) of the received symbols and the output of the dual encoders is the pmf of the information symbols. The bidirectional BCJR MAP decoding is implemented by linearly combining the shift register contents of the dual encoders for forward and backward decoding. The proposed decoder significantly reduces the computational complexity of the bidirectional BCJR MAP algorithm from exponential to linear with constraint length of convolutional codes. To further reduce complexity, fast Fourier transform (FFT) is applied. Mathematical proofs and simulation results are provided to validate our proposed decoder.