一种最大密度检测欠定混合矩阵估计算法
|
摘要
针对源数未知条件下欠定盲源分离混合矩阵估计问题,提出了最大密度检测混合矩阵估计算法。在观测信号稀疏表示的基础上,首先对观测信号进行预处理;然后寻找观测信号的最大密度点;接着在此基础上确定有效样本点集合,再聚类得到辐射源数和混合矩阵。为验证算法的有效性,在时频单源点检测法和小波变换法下开展了仿真实验。结果表明,所提出算法的源数和混合矩阵估计效果优于参考算法,计算复杂度远低于参考算法。进一步实验表明,所提出算法对于正定、超定和欠定盲源分离混合矩阵的估计都具有较好的适用性。
Aiming at mixing matrix estimation when the source number is unknown for underdetermined blind source separation(UBSS),a mixing matrix estimation method based on maximum density point searching is proposed.Based on sparse representation of observed signals,for the proposed algorithm,preprocessing of observed signals is processed first,and then the maximum density point of each observed signal is searched,after which the effective sample points are assembled,and then the source number and mixing matrix are estimated by the clustering method.For validation of the proposed algorithm,the simulations are developed by employing two sparse representation methods,which are single source point detection in the time-frequency domain and wavelet transform.Results show that the source number and the mixing matrix effect of the proposed algorithm are better than those of the reference algorithm,and that the calculation complexity of the proposed algorithm is much less than that of the reference algorithm.Further tests show that the proposed algorithm is applicable for mixing matrix estimation of positive-determined,overdetermined and underdetermined blind source separation models.
Aiming at mixing matrix estimation when the source number is unknown for underdetermined blind source separation(UBSS),a mixing matrix estimation method based on maximum density point searching is proposed.Based on sparse representation of observed signals,for the proposed algorithm,preprocessing of observed signals is processed first,and then the maximum density point of each observed signal is searched,after which the effective sample points are assembled,and then the source number and mixing matrix are estimated by the clustering method.For validation of the proposed algorithm,the simulations are developed by employing two sparse representation methods,which are single source point detection in the time-frequency domain and wavelet transform.Results show that the source number and the mixing matrix effect of the proposed algorithm are better than those of the reference algorithm,and that the calculation complexity of the proposed algorithm is much less than that of the reference algorithm.Further tests show that the proposed algorithm is applicable for mixing matrix estimation of positive-determined,overdetermined and underdetermined blind source separation models.
引文
[1]KIM S G,YOO C D.Underdetermined Blind Source Separation Based on Subspace Representation[J].IEEETransactions on Signal Processing,2009,57(7):2604-2614.
[2]毕晓君,宫汝江.基于混合聚类和网格密度的欠定盲矩阵估计[J].系统工程与电子技术,2012,34(3):614-618.BI Xiaojun,GONG Rujiang.Underdetermined Blind Mixing Matrix Estimation Algorithm Based on Mixing Clustering and Mesh Density[J].Systems Engineering and Electronics,2012,34(3):614-618.
[3]KIRBIZ S,GUNSEL B.Perceptually Enhanced Blind Single-channel Music Source Separation by Non-negative Matrix Factorization[J].Digital Signal Processing:a Review Journal,2013,23(2):646-658.
[4]付卫红,马丽芬,李爱丽.基于改进K-均值聚类的欠定混合矩阵盲估计[J].系统工程与电子技术,2014,36(11):2143-2148.FU Weihong,MA Lifen,LI Aili.Blind Estimation of Underdetermined Mixing Matrix Based on Improved K-means Clustering[J].Systems Engineering and Electronics,2014,36(11):2143-2148.
[5]CONG F Y,LIN Q H,KUANG L D,et al.Tensor Decomposition of EEG Signals:a Brief Review[J].Journal of Neuroscience Methods,2015,248:59-69.
[6]LI Y B,NIE W,YE F,et al.A Mixing Matrix Estimation Algorithm for Underdetermined Blind Source Separation[J].Circuits,Systems,and Signal Processing,2016,35(9):3367-3379.
[7]ZOU L,CHEN X,JI X Y,et al.Underdetermined Joint Blind Source Separation of Multiple Datasets[J].IEEEAccess,2017,5:7474-7487.
[8]WANG C C,XU J Q,ZENG Y H.Research on Influence of Source Number Estimation on Application of Blind Source Separation Algorithms[C]//Procedia Computer Science:107.Amsterdam:Elsevier B V,2017:379-384.
[9]SU Q,SHEN Y H,WEI Y M,et al.Underdetermined Blind Source Separation by a Novel Time-frequency Method[J].International Journal of Electronics and Communications(AEü),2017,77:43-49.
[10]刘明骞,陈钱,吴芸,等.Underlay认知下时频重叠信号的分量个数估计[J].西安电子科技大学学报,2017,44(6):43-47.LIU Mingqian,CHEN Qian,WU Yun,et al.Component Number Estimation of Time-frequency Overlapped Signals in Underlay Cognitive Radio[J].Journal of Xidian University,2017,44(6):43-47.
[11]YANG Z,CHEN W T,HUANG J.Enhancing Recommendation on Extremely Sparse Data with Blocks-coupled Nonnegative Matrix Factorization[J].Neurocomputing,2018,278:126-133.
[12]TASESKA M,HABETS E A P.Blind Source Separation of Moving Sources Using Sparsity-based Source Detection and Tracking[J].IEEE/ACM Transactions on Audio,Speech,and Language Processing,2018,26(3):657-670.
[2]毕晓君,宫汝江.基于混合聚类和网格密度的欠定盲矩阵估计[J].系统工程与电子技术,2012,34(3):614-618.BI Xiaojun,GONG Rujiang.Underdetermined Blind Mixing Matrix Estimation Algorithm Based on Mixing Clustering and Mesh Density[J].Systems Engineering and Electronics,2012,34(3):614-618.
[3]KIRBIZ S,GUNSEL B.Perceptually Enhanced Blind Single-channel Music Source Separation by Non-negative Matrix Factorization[J].Digital Signal Processing:a Review Journal,2013,23(2):646-658.
[4]付卫红,马丽芬,李爱丽.基于改进K-均值聚类的欠定混合矩阵盲估计[J].系统工程与电子技术,2014,36(11):2143-2148.FU Weihong,MA Lifen,LI Aili.Blind Estimation of Underdetermined Mixing Matrix Based on Improved K-means Clustering[J].Systems Engineering and Electronics,2014,36(11):2143-2148.
[5]CONG F Y,LIN Q H,KUANG L D,et al.Tensor Decomposition of EEG Signals:a Brief Review[J].Journal of Neuroscience Methods,2015,248:59-69.
[6]LI Y B,NIE W,YE F,et al.A Mixing Matrix Estimation Algorithm for Underdetermined Blind Source Separation[J].Circuits,Systems,and Signal Processing,2016,35(9):3367-3379.
[7]ZOU L,CHEN X,JI X Y,et al.Underdetermined Joint Blind Source Separation of Multiple Datasets[J].IEEEAccess,2017,5:7474-7487.
[8]WANG C C,XU J Q,ZENG Y H.Research on Influence of Source Number Estimation on Application of Blind Source Separation Algorithms[C]//Procedia Computer Science:107.Amsterdam:Elsevier B V,2017:379-384.
[9]SU Q,SHEN Y H,WEI Y M,et al.Underdetermined Blind Source Separation by a Novel Time-frequency Method[J].International Journal of Electronics and Communications(AEü),2017,77:43-49.
[10]刘明骞,陈钱,吴芸,等.Underlay认知下时频重叠信号的分量个数估计[J].西安电子科技大学学报,2017,44(6):43-47.LIU Mingqian,CHEN Qian,WU Yun,et al.Component Number Estimation of Time-frequency Overlapped Signals in Underlay Cognitive Radio[J].Journal of Xidian University,2017,44(6):43-47.
[11]YANG Z,CHEN W T,HUANG J.Enhancing Recommendation on Extremely Sparse Data with Blocks-coupled Nonnegative Matrix Factorization[J].Neurocomputing,2018,278:126-133.
[12]TASESKA M,HABETS E A P.Blind Source Separation of Moving Sources Using Sparsity-based Source Detection and Tracking[J].IEEE/ACM Transactions on Audio,Speech,and Language Processing,2018,26(3):657-670.