松弛稀疏性条件下的欠定盲分离  

Underdetermined Blind Separation Under Relaxed Sparsity Condition

作　　者：肖明[1,2] 孙功宪[2] 吕俊[2] 

机构地区：[1]茂名学院电子信息工程系,茂名525000 [2]华南理工大学电子与信息学院,广州510641

出　　处：《计算机工程》2010年第12期274-276,共3页Computer Engineering

基　　金：国家"973"计划基金资助项目(2010CB731800);国家自然科学基金资助重点项目(U0635001;U0835003);国家自然科学基金资助项目(0874061;60774094);广东省国际合作研究基金资助项目(2009B050700020)

摘　　要：采用2步法研究松弛稀疏性条件下的欠定盲信号分离。在矩阵恢复上,将时域检索平均法从时域扩展到小波域,得到单源区间矩阵恢复算法。在源信号恢复上,分析最短路径法和l1范数算法,提出基于任意观测信号数的统计稀疏分解准则算法。仿真结果表明,相比l1范数解算法,该算法具有较低的计算复杂度,且可提高恢复信号的信噪比。This paper uses two-step strategy to research underdetermined blind signal separation under relaxed sparsity condition. In matrix recovery, the Searching-and-Averaging Method in Time Domain（SAMTD） algorithm is expanded into the wavelet domain and Matrix Recovery In Single Source Intervals（MRISSI） is obtained. In source recovery, statistically sparse decomposing principle Statistical Sparse Decomposition Principle（SSDP） for arbitrary number of observed signal is proposed after the analysis of the shortest-path method and l1-norm algorithm. Simulation results show that this algorithm has lower computation complexity and it can improve the SNR of the recovered signal compared with l1-norm algorithm.

关 键 词：欠定盲信号分离 单源区间 松弛稀疏性 统计稀疏分解准则 

分 类 号：TN911.12[电子电信—通信与信息系统]