基于反向传播神经网络的自适应SCL译码算法

BP Neural Network-based Adaptive SCL Decoding Algorithm for Polar Codes

SCL译码算法是SC译码算法的加强版。凭借这种性能优势,SCL译码器已经成为针对Polar码译码最常用的一种译码器。然而,SCL译码器的复杂度随着列表大小L线性增长,限制了SCL译码器进一步的广泛应用。于是,提出了基于反向传播神经网络的自适应SCL译码方案。该自适应译码器利用Polar码冻结位比特信息作为反向传播神经网络的输入,根据反向传播神经网络输出估计信噪比等级,进而据估计的信噪比等级自适应地选择列表大小L对接收的噪声帧进行SCL译码。仿真结果显示,在AWGN信道条件下,所提的自适应SCL译码器能够在实现相同的误码率性能的条件下,极大地减小了译码算法的复杂度。

As an enhanced version of SC(successive-cancellation)decoding algorithm,SCL(successive cancellation list)decoder,for its advanced performance,becomes one of the most favorable decoders for polar codes.However,the linearly-increased complexity with list size limits the applicability of SCL decoder.Therefore,aback-propagation(BP)neural network-based adaptive SCL decoder is proposed.This adaptive SCL decoder uses frozen bitswhich is the special structure of polar codes,to estimate the list size L by BP neural network.Simulation results indicate that the adaptive SCL decoder can achieve similar decoding performance with significant complexity reduction.