Enhanced Loop-weakened Belief Propagation Algorithm for Performance Enhanced Polar Code Decoders

A polar code decoder based on the belief propagation algorithm is desirable because of the potentially low latency and its suitability for parallel execution on multicore and SIMD processors. However, current state-of-the-art algorithms require many iterations to achieve comparable bit and frame error rate compared to successive cancellation algorithms. Also, the current state-of-the-art belief propagation algorithms have a high computational complexity compared to successive cancellation. In this paper we present an enhanced belief propagation algorithm, in which parts of the computations are altered to reduce the negative effect of the short cycles in the polar code factor graph. Our proposed algorithm has a gain of
$\approx+0.4\mathbf{dB}$
in both frame and bit error rate compared to successive cancellation and a gain of
$\approx+0.16\mathbf{dB}$
and
$\approx+0.13\mathbf{dB}$
at a frame and bit error rate of 10
⁻³
respectively, compared to belief propagation. Also, the maximum number of iterations of our algorithm is reduce to
$0.6\cdot I_{max}$
. As a result, the latency is up to
$\approx 11$
times lower compared to successive cancellation and up to
$\approx 1.8$
times lower compared to the current state-of-the-art belief propagation polar code algorithm. Furthermore, the reduction in the maximum iteration count results in a lower power consumption after implementation.