基于比率矩阵聚类的欠定盲分离算法

利用稀疏分量的直线聚类性,提出了欠定盲源分离中估计混合矩阵的一种方法。该方法通过构造比率矩阵对观测信号进行分选,剔除了源信号频谱重叠的部分,然后利用鲁棒竞争的聚类学习算法获得对混合矩阵的精确估计,解决了源信号在频域不充分稀疏的条件下准确估计混合矩阵的问题。在估计出混合矩阵的基础上,利用最短路径法分离出源信号。由仿真结果可以看出,与传统的K均值估计混合矩阵的方法相比,方法具有更好的鲁棒性。     

基于源信号数目估计的欠定盲分离

该文利用欠定盲分离下稀疏源信号的特点,估计源信号的数目且恢复源信号。通常在用两步法来解决欠定盲分离时,首先利用K-均值算法对观测信号聚类估计出混叠矩阵,最后利用最短路径法来恢复源信号,但是在以往的算法中,第1步估计混叠矩阵时,通常假设源信号数目是已知的,从而进行K-均值聚类,而事实上源信号数目根本无法知道,因此对源信号数目的估计对两步法有很重要的影响。因此本文提出了一种新的两步法算法,其中第1步利用稀疏源信号反映在观测信号中的特征来准确地估计出稀疏源信号的数目,且能得到混叠矩阵,从而恢复源信号。最后的仿真结果,以及与通常的K-均值聚类算法对比的仿真结果说明了此算法的可行性和优异的性能。     

一种欠定盲源分离中混合矩阵的估计方法

欠定盲源分离问题中基于源信号稀疏性的两阶段法中,混合矩阵估计的准确与否,直接影响源信号的恢复效果。文中提出了一种在稀疏域估计混合矩阵的新方法。该方法通过搜索稀疏域中同一直线附近的点,利用这些点重构出混合矩阵,避免了远离直线周边的点对估计混合矩阵的干扰,从而大大降低了计算量。仿真表明该算法性能良好。     

基于JADE和平行因子分解的欠定混合盲辨识算法

针对源信号个数未知的欠定混合盲源分离问题,本文提出了一种基于特征矩阵联合近似对角化（Joint Approximate Diagonalization of Eigenmatrices, JADE）和平行因子分解的欠定混合盲辨识算法,该算法不需要源信号满足稀疏性要求,仅在源信号满足相互独立和最多一个高斯信号的条件下,通过将JADE算法中的样本四阶协方差矩阵叠加成三阶张量,再对此三阶张量进行平行因子分解来完成源信号数和混合矩阵的估计,由于平行因子分解的唯一辨识性在欠定条件下仍然满足,该算法能够解决欠定盲源分离问题。并对该欠定混合盲辨识算法进行了深入的分析。通过仿真实验,计算估计矩阵与混合矩阵的平均相关误差,结果表明本文提出的算法在适定和欠定混合时均具有很好的辨识效果,而且实现简单,可满足实际应用的要求。      

基于时频分布的欠定混叠盲分离

针对欠定混合信号的盲分离问题,提出了基于时频分布的欠定盲分离算法,首先计算信号的时频分布矩阵并找出信号的自源时频点,然后把自源点对应的时频分布矩阵表示成三阶张量并通过张量分解估计出混合矩阵,最后通过计算矩阵的伪逆和时频合成来完成源信号的恢复.该算法不需要假设源信号是稀疏的或相互独立的.仿真结果表明与已有算法相比本文方法...     

动态变化混叠模型下盲源分离中的源数估计

在进行欠定盲分离时,特别是对于源信号数目及混合矩阵动态变化的情况,常规的欠定盲分离及源数估计方法不能对源信号数目的变化时刻做出判断,因此很难实现动态变化的源信号数目实时和准确的估计。针对这个问题,提出了一种动态变化混叠模型下欠定盲源分离中的源数估计方法。首先,建立动态变化混叠情形下盲源分离的数学模型及动态标识矩阵。其次,基于构建的动态标识矩阵统计和判断动态源信号数目的变化情况。最后,通过分段时间内多维观测矢量采样点聚类区间局部峰值统计,实现动态变化混叠模型下盲源分离中的源信号数目的有效估计。仿真结果表明,该方法能有效实现动态变化混叠模型下欠定盲源分离中的源数估计,并且信号估计效果良好。     

基于两步法稀疏分量分析的欠定盲源分离

分析了解决欠定盲源分离问题的稀疏分量分析方法。首先讨论了数据矩阵稀疏表示（分解）的方法,其次重点讨论了基于稀疏因式分解方法的盲源分离。该盲源分离技术分两步．一步是估计混合矩阵,第二步是估计源矩阵。如源信号是高度稀疏的,盲分离可直接在时域内实现。否则．对观测的混合矩阵运用小波包变换预处理后才能进行。仿真结果证明了理论分析的正确性。     

小波域欠定盲图像混合矩阵估计研究

欠定盲源分离已经成为当前盲信号处理的研究热点,欠定的盲图像分离技术在现实的科研和生产实践中有重要的研究意义。由于图像信号的本身特征,直接进行处理不能满足欠定盲源分离的条件,文章所做的工作就是将盲图像的混合图像进过一层小波变换,在小波域得到了充分稀疏的高频分量,然后利用超平面法矢量聚类算法在小波域进行混合矩阵的估计。通过仿真实验,对比传统算法,得到了较好的估计结果。     

一种基于广义协方差矩阵的欠定盲辨识方法

针对欠定盲源分离模型中的混合矩阵估计问题,研究了一种基于广义协方差矩阵的欠定盲辨识方法。该方法利用观测信号采样数据的广义协方差矩阵性质,将构建的一系列模型函数堆叠为一个张量模型,进而将欠定的混合矩阵辨识转换为张量分解模型中秩一分量的估计。理论分析和仿真结果表明,基于广义协方差矩阵的欠定盲辨识方法比经典的基于二阶协方差矩阵和基于四阶累积量的欠定盲辨识方法具有优越的性能。     

基于ICA的雷达信号欠定盲分离算法

该文针对源信号时域和频域不充分稀疏的情况,提出了欠定盲源分离中估计混合矩阵的一种新方法。该方法对等间隔分段的观测信号应用独立分量分析(ICA)的盲分离算法获得多个子混合矩阵,然后对其分选剔除了不属于原混合矩阵的元素,最后利用C均值聚类的学习算法获得对混合矩阵的精确估计,解决了源信号在时域和频域不充分稀疏的情况下准确估计混合矩阵的问题。在估计出混合矩阵的基础上,利用基于稀疏分解的统计量算法分离出源信号。由仿真结果,以及与传统的K均值聚类,时域检索平均算法对比的实验结果说明了该文算法的有效性和鲁棒性。     

Underdetermined blind separation of non-disjoint signals in time-frequency domain based on matrix diagonalization

To estimate precisely the mixing matrix and extract the source signals in underdetermined case is a challenging problem, especially when the source signals are non-disjointed in time-frequency (TF) domain. The conventional algorithms such as subspace-based achieve blind source separation exploiting the sparsity of the original signals and the mixtures must satisfy the assumption that the number of sources that contribute their energy at any TF point is strictly less than that of sensors. This paper proposes a new method considering the uncorrelated property of the sources in the practical field which relaxes the sparsity condition of sources in TF domain. The method shows that the number of the sources that exist in any TF neighborhood simultaneously equals to that of sensors. We can identify the active sources and estimate their corresponding TF values in any TF neighborhood by matrix diagonalization. Moreover, this paper proposes a method for estimating the mixing matrix by classifying the eigenvectors corresponded to the single source TF neighborhoods. The simulation results show the proposed algorithm separates the sources with higher signal-to-interference ratio compared to other conventional algorithms.     

Mixing matrix estimation method for dual‐channel time‐frequency overlapped signals based on interval probability

For dual‐channel time‐frequency (TF) overlapped signals with low sparsity in underdetermined blind source separation (UBSS), this paper proposes an effective method based on interval probability to estimate and expand the types of mixing matrices. First, the detection of TF single‐source points (TF‐SSP) is used to improve the TF sparsity of each source. For more distinguishability, as the ratios of the coefficients from different columns of the mixing matrix are close, a local peak‐detection mechanism based on interval probability (LPIP) is proposed. LPIP utilizes uniform subintervals to optimize and classify the TF coefficient ratios of the detected TF‐SSP effectively in the case of a high level of TF overlap among sources and reduces the TF interference points and redundant signal features greatly to enhance the estimation accuracy. The simulation results show that under both noiseless and noisy cases, the proposed method performs better than the selected mainstream traditional methods, has good robustness, and has low algorithm complexity.     

利用DBSCAN和概率密度估计的欠定盲源分离混合矩阵估计

针对欠定盲源分离中混合矩阵估计精度不佳的问题,本文提出了一种结合带噪声的基于密度的空间聚类（combining density-based spatial clustering of application with noise, DBSCAN）和概率密度估计的混合矩阵估计算法。首先,通过向量转换方式获得单声源时频点检测准则,并基于此准则从混合信号中检测出单声源点。其次,利用基于密度的空间聚类算法对单声源点进行聚类,由此估计出声源个数以及各类别所属的单声源点。再次,利用概率密度估计获得各类别的聚类中心,并构成混合矩阵。所提混合矩阵估计方法不需要提前设定声源个数,并且避免了由于数据分布不均所造成的聚类效果差的问题。最后,采用压缩感知技术实现源信号恢复,从而从混合信号中分离出各个声源信号。实验结果表明,本文所提的混合矩阵估计方法在声源个数未知的情况下,能够准确估计出混合矩阵;并且分离出的信号具有较高的质量。     

Underdetermined Blind Separation of Nondisjoint Sources in the Time-Frequency Domain

This paper considers the blind separation of nonstationary sources in the underdetermined case, when there are more sources than sensors. A general framework for this problem is to work on sources that are sparse in some signal representation domain. Recently, two methods have been proposed with respect to the time-frequency (TF) domain. The first uses quadratic time-frequency distributions (TFDs) and a clustering approach, and the second uses a linear TFD. Both of these methods assume that the sources are disjoint in the TF domain; i.e., there is, at most, one source present at a point in the TF domain. In this paper, we relax this assumption by allowing the sources to be TF-nondisjoint to a certain extent. In particular, the number of sources present at a point is strictly less than the number of sensors. The separation can still be achieved due to subspace projection that allows us to identify the sources present and to estimate their corresponding TFD values. In particular, we propose two subspace-based algorithms for TF-nondisjoint sources: one uses quadratic TFDs and the other a linear TFD. Another contribution of this paper is a new estimation procedure for the mixing matrix. Finally, then numerical performance of the proposed methods are provided highlighting their performance gain compared to existing ones     

Audio-Visual Underdetermined Blind Source Separation Algorithm Based on Gaussian Potential Function

Most existing algorithms for the underdetermined blind source separation （UBSS） problem are two-stage algorithm, i.e., mixing parameters estimation and sources estimation. In the mixing parameters estimation, the previously proposed traditional clustering algorithms are sensitive to the initializations of the mixing parameters. To reduce the sensitiveness to the initialization, we propose a new algorithm for the UBSS problem based on anechoic speech mixtures by employing the visual information, i.e., the interaural time difference （ITD） and the interaural level difference （ILD）, as the initializations of the mixing parameters. In our algorithm, the video signals are utilized to estimate the distances between microphones and sources, and then the estimations of the ITD and ILD can be obtained. With the sparsity assumption in the time-frequency domain, the Gaussian potential function algorithm is utilized to estimate the mixing parameters by using the ITDs and ILDs as the initializations of the mixing parameters. And the time-frequency masking is used to recover the sources by evaluating the various ITDs and ILDs. Experimental results demonstrate the competitive performance of the proposed algorithm compared with the baseline algorithms.     

基于孤立点检测的欠定混合矩阵盲辨识

本文将孤立点检测的思想引入到欠定混合矩阵的盲辨识问题,提出了一种基于孤立点检测的混合矩阵盲辨识方法.首先计算混合信号的空间时频分布并检测出单源时频点,然后检测出单源时频点中的孤立点并将其从中去除,再通过聚类的方法估计混合矩阵.该方法降低了对信号稀疏性的要求,通过去除数据中的孤立点,提高了矩阵的估计精度,同时也有助于对源信号数目的估计.仿真实验表明,与已有算法相比,本文方法进一步提高了混合矩阵的估计精度,并且有更强的鲁棒性.     

A complex mixing matrix estimation algorithm in under-determined blind source separation problems

This paper considers the complex mixing matrix estimation in under-determined blind source separation problems. The proposed estimation algorithm is based on single source points contributed by only one source. First, the problem of complex matrix estimation is transformed to that of real matrix estimation to lay the foundation for detecting single source points. Secondly, a detection algorithm is adopted to detect single source points. Then, a potential function clustering method is proposed to process single source points in order to get better performance. Finally, we can get the complex mixing matrix after derivation and calculation. The algorithm can estimate the complex mixing matrix when the number of sources is more than that of sensors, which proves it can solve the problem of under-determined blind source separation. The experimental results validate the efficiency of the proposed algorithm.     

Underdetermined mixing matrix estimation bycomprehensive application of cluster validity indexes

To solve the problem of mixing matrix estimation for underdetermined blind source separation (UBSS) when thenumber of sources is unknown, this paper proposed a novel mixing matrix estimation method based on averageinformation entropy and cluster validity index (CVI). Firstly, the initial cluster center is selected by using fuzzy C-means (FCM) algorithm and the corresponding membership matrix is obtained, and then the number of clusters isobtained by using the joint decision of CVI and average information entropy index of membership matrix, thenmultiple cluster number estimation results can be obtained by using multiple CVIs. Then, according to the results ofthe number of multiple clusters estimation, the number of radiation sources is determined according to the principleof the subordination of the minority to the majority. The cluster center vectors obtained from the clustering operationof the estimated number of radiation sources are fused, that is the mixing matrix is estimated based on the degree ofsimilarity of the cluster center vectors. When the source signal is not sufficiently sparse, the time-frequency singlesource detection processing can be combined with the proposed method to estimate the mixing matrix. Theeffectiveness of the proposed method is validated by experiments.     

基于欠定盲源分离的同步跳频信号网台分选

针对同步跳频(FH)网台分选问题,该文提出一种基于时频域单源点检测的欠定盲源分离(UBSS)分选算法。该算法首先对观测信号时频变换,利用自适应阈值去噪算法消除时频矩阵背景噪声,增加算法抗噪性能,然后根据信号绝对方位差算法进行单源点检测,有效保证单源点的充分稀疏性,并通过改进的模糊值聚类算法完成混合矩阵和2维波达方向估计,降低噪声和样本集分布差异对聚类结果的影响,提高估计精度。最后采用变步长的稀疏自适应子空间追踪(SASP)算法对源信号进行重构恢复。仿真实验表明,该算法在低信噪比(SNR)条件下,跳频信号波达方向估计和恢复精度较高,能够有效完成同步跳频信号的盲分离。     

Spatial evolutionary spectrum for DOA estimation and blind signalseparation

We combine the concepts of evolutionary spectrum and array processing. We present a cross-power stationary periodogram for both direction-of-arrival (DOA) estimation and blind separation of nonstationary signals. We model the nonstationary signals received by each array sensor as a sum of complex sinusoids with time-varying amplitudes. These magnitudes carry information about the DOA that may also be time-varying. We first estimate the time-varying amplitudes using estimators obtained by minimizing the mean-squared error. Then using the estimated time-varying amplitudes, we estimate the evolutionary cross-power distributions of the sensor. Next, using cross-power estimates at time-frequency points interest, we estimate the DOAs using one of the existing methods. If the directions are time varying, we choose time-frequency points around the time of interest to estimate spontaneous source locations. If the sources are stationary, time-frequency points of interest can be combined for the estimation of fixed directions. Whitening and subspace methods used to find the mixing matrix and separate nonstationary signals received by the array. We present examples illustrating the performance of the proposed algorithms