基于PD-CNN的Polar码译码算法

A Decoding Algorithm for Polar Codes Based on PD-CNN

针对传统Polar码译码算法在相关噪声信道下性能严重下降的问题,提出了一种基于前置预判-卷积神经网络(Prior Decision-Convolutional Neural Networks,PD-CNN)的译码算法。通过前置预判深度优化CNN,使其准确地估计信道噪声并使残余噪声尽可能遵循高斯分布,再根据残余噪声分布统计更新出可靠的似然比信息。分析了不同译码算法对不同码率Polar码在不同噪声相关强度下的译码性能,并与本文所提出的译码算法进行对比。仿真结果表明:在相关噪声信道下,当误码率为10^-5时,本文所提出的译码算法与标准置信度传播算法相比可获得约2.5 dB的增益。此外,在高信噪比时,与置信度传播-卷积神经网络算法相比,本文提出的译码算法具有相同的性能,但复杂度更低,译码延迟最大可减少42%。

Since the traditional decoding algorithms of Polar code are seriously degraded under the correlated noise channel,a decoding algorithm based on the Prior Decision-Convolutional Neural Networks(PD-CNN)is proposed.The CNN is optimized depth by the pre-predicted in order to accurately estimate the channel noise and make the residual noise follow the Gaussian distribution as much as possible,and make the reliable likelihood ratio information be updated according to the residual noise distribution statistics.The decoding performances of different decoding algorithms for different code rate of Polar codes under different noise correlation intensities are analyzed and compared with the decoding algorithm proposed in this paper.The simulation results show that the proposed decoding algorithm can obtain a gain of about 2.5 dB compared with the standard Belief Propagation algorithm at the bit error rate(BER)of 10^-5 in the correlated noise channel.In addition,the proposed decoding algorithm has the same performance compared with the Belief Propagation-Convolution Neural Network algorithm at high signal-to-noise ratio(SNR),but the complexity is lower,and the decoding delay can be reduced by 42%.