基于改进的二阶统计量盲分离研究

利用二阶统计量(不同时延相关矩阵)的非平稳性和信号时序结构特征,能简单估计出线性瞬时混合的盲源信号。但随时延 增大,仅利用某一个时延协方差均衡化,忽略了信号的时间变化特性,很难保证算法的性能。本文通过分析矩阵的平均特征,提出了一种改进的基于二阶统计量盲源分离算法,对一组均衡化的时延相关函数进行等时延分段,并对等间隔段的两个时延矩阵分别求取均值,采用类似联合近似对角化,估计出最优化的酉矩阵,最终得到信源的稳健估计。性能指标分析和仿真实验证明,本文算法很好地解决了原算法的不足,并成功分离出信号。     

一种新的相关机械振源盲分离方法

针对传统盲源分离算法在机械振源不满足统计独立特性时,无法有效分离出振源信号的问题,提出了基于信号稀疏特性的相关机械源盲分离方法。盲源分离算法的关键在于准确地估计出混合矩阵。因此,首先提出了不相关源混合矩阵的估计方法;然后针对相关源,提出了有效剔除相关成分的方法,使得剩余信号可以按照不相关源进行处理。通过理论分析、仿真验证以及实测数据分析,验证了该方法的有效性。     

基于多个时间点联合解相关的卷积盲源分离

实际信号的混合均为卷积混合,且信号是非平稳的。盲源分离的目标就是找到一组分离滤波器,使得源信号的估计信号互相统计独立。结合信号的非平稳性,利用二阶解相关原理,文章阐明了一种在频域实现卷积混合的盲源分离算法,并且考虑了噪声对分离性能的影响。为了避免频点排列次序的不确定性,利用了多阶段盲源分离思想。利用该算法,对两路混合的实录水声信号进行盲分离,得到了两路源信号的估计信号,通过对估计信号的分析,利用信噪比提高率这一标准,验证了该算法的有效性。该算法收敛速度快,精度高,可用于浅海环境下实录水声混合信号的盲分离。     

应用JADE盲分离算法分离统计相关源

在一些如故障诊断等复杂的应用系统中，不相关源与相关源往往以相互混合方式同时存在于传感观测中。由于相关源不满足盲源离理论有关源的统计独立性前提假设，因此现有的盲源分离算法如基于矩阵联合对角化的算法等，无法从混有相关源的传感观测中准确分离源信号(不相关源信号和相关源信号)。本文在基于矩阵联合对角化的盲分离算法的基础上，提出通过对其源估计进行修正的相关源分离算法。理论分析和仿真结果表明：修正后的基于矩阵联合对角化的盲分离算法，能有效地分离包含相关源的混合观测信号。     

基于相似性函数的稀疏成分分析混合矩阵估计方法

稀疏成分分析是信号处理中解决欠定盲源分离问题的新方法,本文研究了稀疏成分分析中的混合矩阵估计问题,提出了无需预知源个数利用一种相似性函数估计混合矩阵的方法。首先,估计相似性函数中的核参数,使得算法适应不同的稀疏信号。然后,给出了估计混合矩阵的不动点算法。最后,实验结果表明提出的算法通过适当地选取参数,能够准确有效地估计出具有不同源个数的混合矩阵,对不太稀疏的源也有令人满意的结果。     

基于平面聚类势函数法的欠定混合信号盲分离

针对不充分稀疏欠定混合信号盲分离,提出了一种基于超平面聚类的势函数法来估计源信号个数和混合矩阵。该方法在源信号个数未知的情况下,利用聚类平面法线向量构成势函数,通过估计势函数的局部最大值来估计聚类平面的法线向量,然后再通过估计聚类平面的交线来实现混合矩阵的估计。为了提高算法对异常值的鲁棒性,不直接估计势函数的局部最大值,而是采用聚类算法来估计势函数的局部最大值。计算机仿真试验证实了该算法的有效性及其较好的性能。     

基于EMMD的欠定旋转机械振动信号盲源分离

传统盲源分离算法通常基于观测信号数不小于源信号数的假设, 当观测信号数小于源信号数时, 盲源分离效果较差, 而欠定的甚至单通道信号的盲源分离问题普遍存在于旋转机械中, 针对该问题, 提出一种基于极值域均值分解的欠定旋转机械振动信号盲源分离算法。首先把欠定的观测信号进行极值域均值分解得到一系列本征模函数, 将欠定观测信号和其本征模函数组成多维信号, 作为新的观测信号实现升维;然后利用奇异值分解和贝叶斯准则进行源数估计, 最后利用基于四阶累积量的特征矩阵联合对角化方法实现信号的盲分离。仿真结果表明, 该方法能够有效地解决欠定观测信号的盲源分离问题。     

基于WOA-VMD改进的信号盲源分离研究及应用EI北大核心CSCD

准确有效识别出水电站厂房振动信号的各个振源,对于水电站长期安全稳定运行有重要指导意义,盲源分离(blind source separation,BSS)是进行信号分解与振源识别的一种有效方法。为了实现水电站厂房复杂环境下振动信号的盲源分离,建立一种基于鲸鱼算法变分模态分解(whale optimization algorithm and variational mode decomposition,WOA-VMD)降噪改进的信号盲源分离模型。采用WOA-VMD和相关法对观测信号进行降噪处理,确保盲源分离结果的准确性;求解多维降噪信号的协方差矩阵并进行奇异值分解,采用优势特征值法进行源数估计;最后对降噪信号进行中心化、白化前处理,通过联合近似对角化算法得到分离矩阵,实现观测信号的盲源分离。仿真结果表明:相较于传统盲源分离模型,改进模型将分离信号与源信号的相关系数分别提升了9.1%,7.1%,8.3%,分离信号主频误差也均有所降低。将该方法运用到水电站厂房振动工程实例,也取得了较好的分离效果。     

一种改进的盲解卷积算法在轴承声学诊断中的应用

针对时域盲解卷积算法滤波器长度估计困难的缺点,提出一种基于遗传算法优化的改进算法。该算法利用遗传算法搜索最佳时延,解决了盲解卷积结果不确定问题,并改进了信号分量的聚类指标,采用峭度作为独立分量间距离测度,提高了信号分量聚类的准确性,获得了可靠的估计信号。计算机仿真和实际环境中故障轴承声信号提取实验验证了该算法的有效性。     

一种盲源提取方法及其在滚动轴承故障特征提取中的应用

利用滚动轴承发生故障时其故障信号往往呈现出一定周期性的特点,首先计算出故障信号的理论基本周期τ,将τ作为待提取源信号的基本周期用所述方法的相关步骤计算出大致目标源信号及权重分离矩阵^W。然后将^W作为初始权重分离矩阵,将基于高阶统计量的固定点算法用于原始观测信号提取出更为精确的目标故障信号。通过仿真信号和实验信号验证了所述方法相对于约束独立成分分析(Constrained independent component analysis,CICA)方法具有以下优点:不需要精确估计目标源信号的周期及不需要构建精确的参考信号。此外,通过仿真还验证了所述方法相对于其他较新的盲源提取方法具较高的提取精度等优点。     

时滞对典型二阶振荡系统稳定性的影响

针对一类典型二阶振荡系统进行了稳定性分析,分别讨论了闭环系统为单位负反馈和单位正反馈时信号传输时滞对控制系统性能的影响。通过绘制和分析相对阻尼系数在不同取值区间的Nyquist曲线,得出了时滞与闭环系统稳定性的关系,并对各种情况进行了单位阶跃响应的实例仿真。仿真结果表明了分析的正确性。     

一种基于累积量的ESPRIT型盲波束形成算法

讨论一种用于非高斯信号源方位估计的盲波束形成算法。假设背景噪声是二阶统计未知的有色(空间相关)高斯噪声,基于“阵元对”模型的ESPRIT方位估计算法可以通过用累积量矩阵取代自相关矩阵完成信号空间的重建。经过这种处理,加性有色高斯噪声可被滤除,因此算法不需要相关噪声矩阵的知识。当加性噪声源是空间相关矩阵未知的有色高斯噪声时,计算机仿真比较了基于累积量的ESPRIT算法与基于二阶统计的ESPRIT算法的性能。     

Sweep signal time-delay estimation with time-frequency cross-correlation algorithm

Sweep signal is usually employed as a source signal in active detection such as radar and sonar. Since the frequency spectrum of the sweep signal varies against time, a novel algorithm, namely a time-frequency cross-correlation (TFCC) algorithm based on wavelet packet transform (WPT), is proposed to estimate the time delay of sweep signal. In this algorithm, the source sweep and the received signals are decomposed with WPT to obtain their time-frequency representations and the TFCC between the source sweep and the received signals is performed. Each reflected sweep in the received signal is converted into a time-frequency correlation peak whose position can indicate its time delay. The TFCC algorithm can suppress ambient noise effectively and improve the performance of sweep extraction and can match more precisely the source and the reflected sweeps to their known time-frequency characters. Numerical experiments were performed to compare the performance of the TFCC algorithm with that of the conventional cross- correlation and phase-data algorithms. The results proved that the TFCC algorithm can extract the reflected sweeps effectively and its performance is better than that of the conventional algorithms.     

A spline-based algorithm for continuous time-delay estimation using sampled data

Time delay estimation (TDE) lies at the heart of signal processing algorithms in a broad range of application areas, including communications, coherent imaging, speech processing, and acoustics. In medical ultrasound for example, TDE is used in blood flow estimation, tissue motion measurement, tissue elasticity estimation, phase aberration correction, and a number of other algorithms. Because of its central significance, TDE accuracy, precision, and computational cost are of critical importance. Furthermore, because TDE is typically performed on sampled signals-and delay estimates are usually desired over a continuous domain-time delay estimator performance should be considered in conjunction with associated interpolation. In this paper we present a new time-delay estimator that directly determines continuous time-delay estimates from sampled data. The technique forms a spline-based, piecewise continuous representation of the reference signal then solves for the minimum of the sum squared error between the reference and the delayed signals to determine their relative time delay. Computer simulation results clearly show that the proposed algorithm significantly outperforms other algorithms in terms of jitter and bias over a broad range of conditions. We also describe a modified version of the algorithm that includes companding with only a minor increase in computational cost.     

滑动轴承油膜失稳振动的二阶定向循环平稳性分析

定义了二阶定向循环平稳统计量,提出了基于二阶定向循环平稳的信号分析方法。该方法将用于单通道实信号分析的二阶循环统计量扩展到复信号,并定义了正交双通道融合复信号的定向循环自相关、定向循环谱相关密度。为了验证该方法,利用转子试验台模拟滑动轴承旋转机械的油膜失稳故障以获取数据,分析结果表明该方法不仅能够提取系统在某一截面内的周期性振动特征,而且能够揭示转子在指定循环频率处的旋向。此外二阶定向循环谱相关密度在指定频率处的切片能显示更加丰富的转子频率耦合调制信息,试验结果验证了该方法的有效性。       

基于最小二乘支持向量机的结构地震响应时滞控制算法

针对时滞会导致控制系统效果降低、控制性能恶化甚至系统不稳定,将线性二次型控制（Linear Quadratic Regulator,LQR）与最小二乘支持向量机（Least Squares Support Vector Machines,LSSVM）进行集成,提出时滞LSSVM-LQR智能控制算法。该算法采集结构状态响应数据后,将用LQR算法计算出的结构最优控制力输入到LSSVM中,以训练并回归预测出时滞后时刻的最优控制力;由作动器对结构提供控制。基于MATLAB平台编写计算程序,并用一幢三层框架结构进行数值验证。结果表明,LSSVM-LQR算法能有效降低时滞对结构控制系统的不利影响。     

泄漏检测应用中突发干扰对LMS时延估计性能的影响

文章中把LMS时延估计算法用于地下管网的泄漏检测和漏点定位，避免了经典时延估计算法一广义互相关法(GCC方法)需要知道信号和噪声统计特性等先验知识的不足，但是泄漏检测应用中，管道外在环境造成的突发性强干扰，导致了传感器接收到的信号是非平稳的，而非平稳信号对LMS的时延估计性能有不利影响，文章中分析了这种由突发性强干扰导致的非平稳信号对LMS时延估计收敛性、收敛速度和时延估计值的影响，提出了消除突发干扰的方法。实验表明，在地下管网泄漏检测应用中，该方法能够有效地消除强突发干扰噪声，使得估计性能得到显著的改善。     

基于卷积盲源分离的噪声鲁棒性语音识别的研究

研究了一种基于卷积盲分离算法与MFCC（Mel-Frequency Cepstral Coefficient）特征相结合的噪声鲁棒语音识别方法。该方法在预处理阶段,首先计算预白化观测数据的多阶自相关协方差矩阵,以获得多时延处理的二阶解相关统计信息。然后利用得到的二阶统计信息构建两个对称正定矩阵,通过Cholesky因式分解等一系列变换获得唯一存在的矩阵,根据此矩阵估算语音信号并提取MFCC特征用于后续识别。实验结果表明,在低信噪比条件下,该方法对于数字语音的识别性能优于基本的MFCC识别器和文献中已有的卷积分离算法。     

Correlation-sensitive adaptive sequence detection

In high density magnetic recording, noise samples corresponding to adjacent signal samples are heavily correlated as a result of front-end equalizers, media noise, and signal nonlinearities combined with nonlinear filters to cancel them. This correlation significantly deteriorates the performance of detectors at high densities. In this paper, we propose a novel sequence detector that is correlation sensitive and adaptive to the nonstationary signal sample statistics. We derive the correlation-sensitive maximum likelihood detector. It can be used with any Viterbi-like receiver (e.g., partial response maximum likelihood, fixed delay tree search, multilevel decision feedback equalization) that relies on a tree/trellis structure. Our detector adjusts the metric computation to the noise correlation statistics. Because these statistics are nonstationary, we develop an adaptive algorithm that tracks the data correlation matrices. Simulation results are presented that show the applicability of the new correlation-sensitive adaptive sequence detector     

Dynamics of adaptive optical systems

Dynamic (time) characteristics of adaptive systems are analyzed. A common adaptive system with a finite frequency band (or a finite response time) is described as a dynamic constant time-delay system, where time delay is to be much shorter than the time of coherence radius transfer through an optical beam by a mean wind speed. The questions of coherent beam formation are considered with use of the reference source. The analytical calculation of the Strehl parameter is made on basis of the generalized Huygens-Kirchhoff principle. An adaptive system is considered where the correcting phase is calculated with the use of both its derivatives and the signal, as well as adaptive systems using different time-predicting algorithms of the correcting signal for future time points. The use of a predicted phase front of the correcting wave allows much longer time delays. The stronger the phase distortions in the optical wave, the higher the time gain in comparison with common (with constant time delay) adaptive systems.