噪声情况下采用稀疏非负矩阵分解与深度吸引子网络的人声分离算法

为实现噪声情况下的人声分离,提出了一种采用稀疏非负矩阵分解与深度吸引子网络的单通道人声分离算法。首先,通过训练得到人声与噪声的字典矩阵,将其作为先验信息从带噪混合语音中分离出人声与噪声的系数矩阵;然后,根据人声系数矩阵中不同的声源成分在嵌入空间中的相似性不同,使用深度吸引子网络将其分离为各声源语音的系数矩阵;最后,使用分离得到的各语音系数矩阵与人声的字典矩阵重构干净的分离语音。在不同噪声情况下的实验结果表明,本文算法能够在抑制背景噪声的同时提高分离语音的整体质量,优于结合声噪人声分离模型的对比算法。      

基于条件熵扩维的多变量混沌时间序列相空间重构

提出一种多变量混沌时间序列相空间重构的条件熵扩维方法.首先使用互信息法求解每个变量的时间延迟,其次按条件熵最大原则逐步扩展相空间的嵌入维数,使得重构坐标从低维到高维的转换保持较强的独立性,最终的重构相空间具有较低的冗余度,为多变量时间序列的预测构造了有效的模型输入向量.通过对几个经典多变量混沌时间序列进行数值实验,结果表明该方法比单变量预测和已有多变量预测方法具有更好的预测效果,说明了该重构方法的有效性. 关键词： 多变量混沌时间序列 相空间重构 条件熵 神经网络预测     

Chaotic time series prediction and additive white Gaussian noise

Taken's delay embedding theorem states that a pseudo state-space can be reconstructed from a time series consisting of observations of a chaotic process. However, experimental observations are inevitably corrupted by measurement noise, which can be modelled as Additive White Gaussian Noise (AWGN). This Letter analyses time series prediction in the presence of AWGN using the triangle inequality and the mean of the Nakagami distribution. It is shown that using more delay coordinates than those used by a typical delay embedding can improve prediction accuracy, when the mean magnitude of the input vector dominates the mean magnitude of AWGN.     

用RQA 法分析不同信噪比下的非线性系统

以Lorenz系统为例，运用重现图形和重现定量分析法分析了不同信噪比下噪声对连续系统混沌区非线性特征量的影响．选取自相关函数第一次降到C(0)／2时所对应的t作为延迟时间τ，运用错误近邻分析法确定序列的嵌入维m，以重构相空间．结果表明：当信噪比大于5时，序列存在的确定性规律RPA仍能显示出来．随着信噪比的降低，RQA的参量：％recur，％determ，ratio，entropy和trends的绝对值呈下降的趋势．当SNR≥1000时，噪声对各参量的值影响不大．     

Application of local polynomial estimation in suppressing strong chaotic noise

In this paper, we propose a new method that combines chaotic series phase space reconstruction and local polynomial estimation to solve the problem of suppressing strong chaotic noise. First, chaotic noise time series are reconstructed to obtain multivariate time series according to Takens delay embedding theorem. Then the chaotic noise is estimated accurately using local polynomial estimation method. After chaotic noise is separated from observation signal, we can get the estimation of the useful signal. This local polynomial estimation method can combine the advantages of local and global law. Finally, it makes the estimation more exactly and we can calculate the formula of mean square error theoretically. The simulation results show that the method is effective for the suppression of strong chaotic noise when the signal to interference ratio is low.     

基于径向基神经网络预测的混沌时间序列嵌入维数估计方法

根据Takens定理,研究了混沌时间序列相空间重构嵌入维数的选取问题.提出了基于径向基函数神经网络预测模型性能的嵌入维数估计方法,即根据嵌入维数与混沌时间序列预测模型性能的变化关系来确定嵌入维数.通过对几种典型混沌动力学系统的数值验证,结果表明该方法能够确定出合适的相空间重构嵌入维数. 关键词： 混沌 相空间重构 嵌入维数 预测     

Shrinkage-divergence-proximity locally linear embedding algorithm for dimensionality reduction of hyperspectral image

Existing manifold learning algorithms use Euclidean distance to measure the proximity of data points. However, in high-dimensional space, Minkowski metrics are no longer stable because the ratio of distance of nearest and farthest neighbors to a given query is almost unit. It will degrade the performance of manifold learning algorithms when applied to dimensionality reduction of high-dimensional data. We introduce a new distance function named shrinkage-divergence-proximity (SDP) to manifold learning, which is meaningful in any high-dimensional space. An improved locally linear embedding (LLE) algorithm named SDP-LLE is proposed in light of the theoretical result. Experiments are conducted on a hyperspectral data set and an image segmentation data set. Experimental results show that the proposed method can efficiently reduce the dimensionality while getting higher classification accuracy.     

Small-time scale network traffic prediction based on a local support vector machine regression model

In this paper we apply the nonlinear time series analysis method to small-time scale traffic measurement data. The prediction-based method is used to determine the embedding dimension of the traffic data. Based on the reconstructed phase space, the local support vector machine prediction method is used to predict the traffic measurement data, and the BIC-based neighbouring point selection method is used to choose the number of the nearest neighbouring points for the local support vector machine regression model. The experimental results show that the local support vector machine prediction method whose neighbouring points are optimized can effectively predict the small-time scale traffic measurement data and can reproduce the statistical features of real traffic measurements.     

Chaotic time series prediction: From one to another

In this Letter, a new local linear prediction model is proposed to predict a chaotic time series of a component x(t) by using the chaotic time series of another component y(t) in the same system with x(t). Our approach is based on the phase space reconstruction coming from the Takens embedding theorem. To illustrate our results, we present an example of Lorenz system and compare with the performance of the original local linear prediction model.     

Targeting of Kolmogorov-Arnold-Moser Orbits by the Bailout Embedding Method in Two Coupled Standard Maps

A bailout embedding method for controlling chaos can make the chaotic orbits targeting into Kolmogorov- Arnold-Moser orbits. We apply this method to a high-dimensional system with two coupled standard maps. The numerical simulation shows that this method could obtain target islands in order and hence could be used to control chaos. Moreover, it is robust in the presence of weak external noise.     

一种混沌海杂波背景下的微弱信号检测方法

基于复杂非线性系统的相空间重构理论,提出了一种基于遗传算法的支持向量机预测方法.利用改进的自相关法和饱和关联维数法确定混沌信号的时间延迟和嵌入维,从而实现相空间重构.通过遗传算法优化支持向量机中的惩罚系数和核函数参数,并结合支持向量机建立混沌序列的单步预测模型,从预测误差中检测出淹没在混沌背景中的微弱信号(包括瞬态信号和周期信号).以Lorenz系统和加拿大McMaster大学利用IPIX雷达实测得到的海杂波数据作为混沌背景噪声进行仿真实验,结果表明该方法能够有效地从混沌背景噪声中检测出微弱目标信号,所得的均方根误差为0.00049521(信噪比为-89.7704 dB),这比传统支持向量机方法的均方根误差(0.049,信噪比为-54.60 dB)降低了两个数量级.     

Test for low-dimensional determinism in electroencephalograms

We tested low-dimensional determinism in an electroencephalogram (EEG), based on the fact that smoothness (continuity) on an embedded phase space is enough to imply determinism within time series. A modified version of the method developed by Salvino and Cawley [Phys. Rev. Lett. 73, 1091 (1994)] was used. In our method, we chose a box randomly and then estimated the mean directional element in the box containing the d+1 data points, where d is the embedding dimension. The global average for the mean local directional elements over the boxes, W, is a measure for smoothness. The nonlinear noise reduction method developed by Sauer [Physica D 58, 193 (1992)] is then applied to the EEG. We also compared the results for the EEG with those for its surrogate data. We found that the W values for the noise-reduced EEG had stable values around 0.35, which means that the EEG is not a low-dimensional deterministic signal. However, this method may not be applicable to the time series generated from high-dimensional deterministic systems. We cannot exclude the possibility that the determinism in the EEG may be too high-dimensional to be detected with current methods.     

Denoising human speech signals using chaoslike features

A local projective noise reduction scheme, originally developed for low-dimensional stationary deterministic chaotic signals, is successfully applied to human speech. This is possible by exploiting properties of the speech signal which resemble structure exhibited by deterministic dynamical systems. In high-dimensional embedding spaces, the strong inherent nonstationarity is resolved as a sequence of many different dynamical regimes of moderate complexity.     

A Nonlinear Maximum Correntropy Information Filter for High-Dimensional Neural Decoding

Neural signal decoding is a critical technology in brain machine interface (BMI) to interpret movement intention from multi-neural activity collected from paralyzed patients. As a commonly-used decoding algorithm, the Kalman filter is often applied to derive the movement states from high-dimensional neural firing observation. However, its performance is limited and less effective for noisy nonlinear neural systems with high-dimensional measurements. In this paper, we propose a nonlinear maximum correntropy information filter, aiming at better state estimation in the filtering process for a noisy high-dimensional measurement system. We reconstruct the measurement model between the high-dimensional measurements and low-dimensional states using the neural network, and derive the state estimation using the correntropy criterion to cope with the non-Gaussian noise and eliminate large initial uncertainty. Moreover, analyses of convergence and robustness are given. The effectiveness of the proposed algorithm is evaluated by applying it on multiple segments of neural spiking data from two rats to interpret the movement states when the subjects perform a two-lever discrimination task. Our results demonstrate better and more robust state estimation performance when compared with other filters.     

混沌时序相空间重构参数确定的信息论方法

根据信息论基本原理，研究了混沌时间序列相空间重构参数延迟时间和嵌入维数的选取.提出了用符号分析的方法计算互信息函数，确定出延迟时间，在此基础上，提出了一种估计嵌入维数的信息论方法，即根据重构向量条件熵随向量维数的变化关系来确定嵌入维数，通过对几种典型混沌动力学系统的数值验证，结果表明该方法能够确定出合适的相空间重构嵌入维数. 关键词： 混沌 相空间重构 互信息 条件熵 符号分析     

基于核主成分分析和支持向量回归机的红外光谱多组分混合气体定量分析

提出了一种核主成分分析(KPCA)特征提取结合支持向量回归机(SVR)的红外光谱混合气体组分定量分析新方法。首先将特征吸收谱线严重重叠的混合气体光谱通过非线性变换映射到高维特征空间,然后在特征空间中再利用主成分分析法提取主成分,提取出的主成分作为SVR的输入建立校正模型,实现了甲烷、乙烷、丙烷、异丁烷、正丁烷、异戊烷以及正戊烷七种组组分特征吸收光谱严重重叠的混合气体的定量分析。用KPCA-SVR所建模型对未知浓度混合气体的七种组分预测的RMSE (φ×10－60较仅用SVR模型预测的RMSE (φ×10－6)降低了一个数量级。结果表明,核主成分分析法具有很强的非线性特征提取能力,可以充分利用全光谱数据并有效地消除光谱数据噪声,降低数据维数,与支持向量回归机结合可以提高红外光谱分析的精度,缩短模型计算时间,是一种有效的红外光谱分析新方法。     

Noise- and inertia-induced inhomogeneity in the distribution of small particles in fluid flows

The dynamics of small spherical neutrally buoyant particulate impurities immersed in a two-dimensional fluid flow are known to lead to particle accumulation in the regions of the flow in which vorticity dominates over strain, provided that the Stokes number of the particles is sufficiently small. If the flow is viewed as a Hamiltonian dynamical system, it can be seen that the accumulations occur in the nonchaotic parts of the phase space: the Kolmogorov-Arnold-Moser tori. This has suggested a generalization of these dynamics to Hamiltonian maps, dubbed a bailout embedding. In this paper we use a bailout embedding of the standard map to mimic the dynamics of neutrally buoyant impurities subject not only to drag but also to fluctuating forces modeled as white noise. We find that the generation of inhomogeneities associated with the separation of particle from fluid trajectories is enhanced by the presence of noise, so that they appear in much broader ranges of the Stokes number than those allowing spontaneous separation. (c) 2002 American Institute of Physics.     

一种基于聚类的门控卷积网络语声分离方法*

基于深度聚类的语音分离方法已被证明能有效地解决混合语音中说话人输出标签排列的问题,然而,现有关于聚类进行说话人分离方法,大多数是优化嵌入使每个源的重建误差最小化。本文以时域卷积网络（ConvTasNet）为基础网络,设计了一种改进基于聚类的门控卷积（Gate-conv Cluster）语音分离方法,在时域上通过堆叠的门控卷积网络,实现端到端深度聚类的源分离。该框架将非线性门控激活用于时域卷积网络中,提取语音信号的深层次特征;同时在高维特征空间中聚类对语音信号的特征进行表示和划分,为恢复不同信号源提供了一个长期的说话者表示信息。该框架解决了说话人输出标签排列问题并对语音信号的长期依赖性进行建模。通过华尔街日报数据集进行实验得出,该方法在SDRi（信源失真比）和Si-SNR（尺度不变信源噪声比）指标上分别达到了16.72 dB和16.33 dB的效果。     

基于最大Lyapunov指数不变性的混沌时间序列噪声水平估计

提出了一种基于最大Lyapunov指数不变性的计算混沌时间序列噪声水平的新方法. 首先分析了噪声对相空间中两点距离的影响, 然后基于最大Lyapunov指数在不同维数的嵌入相空间不变的性质, 建立了估计噪声水平的方法. 仿真计算结果表明, 当噪声水平小于10% 时, 估计值与真实值符合良好. 该方法对噪声分布类型不敏感, 是一种有效的混沌时间序列噪声估计方法.     

Multi-scale continuum mechanics: from global bifurcations to noise induced high-dimensional chaos

Many mechanical systems consist of continuum mechanical structures, having either linear or nonlinear elasticity or geometry, coupled to nonlinear oscillators. In this paper, we consider the class of linear continua coupled to mechanical pendula. In such mechanical systems, there often exist several natural time scales determined by the physics of the problem. Using a time scale splitting, we analyze a prototypical structural-mechanical system consisting of a planar nonlinear pendulum coupled to a flexible rod made of linear viscoelastic material. In this system both low-dimensional and high-dimensional chaos is observed. The low-dimensional chaos appears in the limit of small coupling between the continua and oscillator, where the natural frequency of the primary mode of the rod is much greater than the natural frequency of the pendulum. In this case, the motion resides on a slow manifold. As the coupling is increased, global motion moves off of the slow manifold and high-dimensional chaos is observed. We present a numerical bifurcation analysis of the resulting system illustrating the mechanism for the onset of high-dimensional chaos. Constrained invariant sets are computed to reveal a process from low-dimensional to high-dimensional transitions. Applications will be to both deterministic and stochastic bifurcations. Practical implications of the bifurcation from low-dimensional to high-dimensional chaos for detection of damage as well as global effects of noise will also be discussed.