This paper proposes a modified belief propagation ( MBP) decoder for polar codes, in which the log-likelihood ratios ( LLRs) associated with the frozen nodes would be revised by a selective correction parameter when the LLRs of frozen nodes are wrongly calculated. By using the Gaussian approximation, an appropriate parameter is obtained creatively. The principles of log-domain BP decoder and corresponding modified decoding algorithm are described in this paper. Finally, computational complexity analysis and performance simulation are given. Data results indicate that compared with the existing BP decoder, the modified BP decoder can obtain a better bit error performance of about 0 . 2 dB at the cost of a little compu-tational complexity.%在极化码置信( BP)译码的因子图中，当承载确定信息的节点的对数似然信息计算错误时，可以被检测到。此时，对于因子图中参与该似然信息计算的节点，引入一个修正参数，以修正该节点承载的信息的对数似然信息。修正参数可以由密度进化的高斯近似算法得到。给出了置信译码原理及相应的改进算法，最后给出了复杂度分析和性能仿真。数据结果表明，在牺牲很小的复杂度的条件下，相比原算法，修正算法能够获得0.2 dB左右的比特信噪比增益。
展开▼
