基于Markov链蒙特卡洛的因子图均衡算法  

Factor Graph Equalization Based on Markov Chain Monte Carlo

作　　者：巩克现[1] 董政[1] 葛临东[1] 

机构地区：[1]信息工程大学信息工程学院,河南郑州450002

出　　处：《四川大学学报（工程科学版）》2012年第6期120-126,共7页Journal of Sichuan University (Engineering Science Edition)

基　　金：国家自然科学基金资助项目(61072046);河南省自然科学基金资助项目(102300410008)

摘　　要：针对基于因子图模型的非线性失真信道的迭代均衡计算复杂度高的问题,提出了3种不同的接收信息后验概率的有效算法以及并行实现方法。在基于因子图的均衡算法中,均衡器和译码器以迭代处理的方式联合工作,提高了系统的整体性能,但计算复杂度随信道记忆长度呈指数增加,通过Markov链蒙特卡洛算法实现多维积分的计算,并通过因子图分割实现并行Gibbs采样,降低了计算复杂度,仿真表明,该算法有效克服宽带高阶调制的卫星信道非线性失真,有利于硬件或多核并行实现。Three different effective solutions and parallel implement method of computing the posteriori probability of received signal were proposed to overcome the high calculation complexity of iterative equalization based on factor graph for nonlinear channel distor- tion. Equalizer and decoder work interactively in factor graph equalization and the performance of the system was improved while the calculation complexity grew exponentially with channel memory length. Multidimensional integration was adopted by Markov chain Monte Carlo algorithm and parallel Gibbs sampling was implemented by factor graph partition. The calculation complexity was reduced. Simulation demonstrated that it overcomes the non-linear distortion of high order modulation over satellite channel and it is suitable to hardware or multi-core implement.

关 键 词：均衡 迭代处理 蒙特卡洛 因子图 

分 类 号：TN911[电子电信—通信与信息系统]