基于p范数漏溢式LMS的稀疏系统辨识-arXiv.PDF

基于ℓ 范数漏溢式LMS的稀疏系统辨识p 1 2 2 冯永 ,曾瑞 ,伍家松 1.东南大学生物科学与医学工程学院,南京,210096 2.东南大学计算机科学与工程学院,南京,210096 摘要: 漏溢式最小均方(leakyLMS,LLMS)算法用于解决标准LMS算法在处理系统输入信号高度相关时性能 衰减的问题，但是它们均未考虑利用系统冲击响应的稀疏性来进一步提高算法性能。因此本文提出一种基于 ℓ 范数约束的LLMS的新算法(ℓ-LLMS)，作为标准LLMS算法的延伸，把ℓ 范数约束引入到LLMS算法的 p p p 代价函数中，从而加强稀疏系统的辨识效果。计算机仿真实验表明，对于稀疏系统辨识，在不同稀疏度条件 下，该算法在稳态误差方面均优于标准LMS, LLMS和ℓ-LMS算法.p 关键词: 自适应滤波；ℓ 范数约束；稀疏系统辨识；漏溢式最小均方；最小均方算法 p pNormConstraintLeakyLMSAlgorithmforSparseSystem Identification 1 2 2 FENGYong,ZENGRui ,WUJiasong 1.SchoolofBiologicalScienceandMedicalEngineering,SoutheastUniversity,Nanjing21 96,China 2.SchoolofComputerScienceandEngineering,SoutheastUniversity,Nanjing21 96,China Abstract:Thispaperproposesanewleakyleastmeansquare(leakyLMS,LLMS)algorithminwhichanormpenaltyis introduced to forcethe solutionto be sparse in the application of system identification. The leaky LMS algorithm is derivedbecausetheperformanceofthestandardLMSalgorithmdeteriorateswhentheinputishighlycorrelated.However, bothofthemdonottakethesparsityinformationintoaccounttoyieldbetterbehaviors.AsamodificationoftheLLMS algorithm,theproposedalgorithm,namedℓ-LLMS,incorporatesap normpenaltyintothecostfunctionoftheLLMSto p obtain a shrinkage in the weight update equation, which then enhances the performance of the filter in system identification settings, especiallywhen the impulseresponse is sparse.The simulationresults verify that the proposed algorithmimprovestheperformanceofthefilterinsparsesystemse