基于小波变换、二维主元分析和EBF神经网络的指纹识别方法

指纹识别是模式识别中的一个十分重要课题。结合小波变换(WT)、二维主元分析(2DPCA)和椭球基函数（EBF）特点,本文提出了一种基于WT和2DPCA的EBF神经网络指纹识别方法。首先,利用小波变换将原始图像分解为高频分量和低频分量,并忽略水平高频与垂直高频分量,获得原始图像的基本特征。然后,通过2DPCA算法对该图像进行降维,获取降维特征;最后结合椭球基函数神经网络(Ellipsoidal Basis Function Neural Network, EBFNN)完成指纹识别。本算法将2DPCA优化的特征提取与EBFNN的自适应性相结合,在FVC2000（国际指纹竞赛数据库）上作了测试。并与WT-PNN算法和WT-2DPCA-RBF算法进行比较。实验结果表明,本文提出的算法在平移、旋转及光照变化的指纹数据库上的识别效果优于WT－PNN算法和WT-2DPCA-RBF算法。     

An enhanced feature extraction model using lifting-based wavelet packet transform scheme and sampling-importance-resampling analysis

The efficiency of data processing is critical for the on-line monitoring applications of industrial components and systems, both from the viewpoints of the rapid adaptation to the non-stationary signals and the cost of information storage and transmission. In this paper, we propose an enhanced feature extraction model for machinery performance assessment, which is based on the lifting-based wavelet packet transform (WPT) and sampling-importance-resampling methods. The lifting-based WPT decomposes the signals. Then the sampling-importance-resampling procedure is applied in the wavelet domain to extract the distribution information and compose the feature vectors. Finally, a support vector machine is used to assess the normal or abnormal condition based on these extracted features. To validate the proposed new model, an endurance test of pressure regulators has been carried out. Compared to the traditional wavelet packet method, the new model can not only keep the precision level but also improve the efficiency by over 60%.     

Bearing fault detection using wavelet packet transform of induction motor stator current

Induction motor vibrations, caused by bearing defects, result in the modulation of the stator current. In this research, bearing defect is detected using the stator current analysis via Meyer wavelet in the wavelet packet structure, with energy comparison as the fault index. The advantage of this method is in the detection of incipient faults. The presented method is evaluated using experimental signals. Sets of data are gathered before and after using defective bearings. Compared to conventional methods, the superiority of the proposed method is shown in the success of fault detection.     

基于小波变换和核主成分分析的人脸识别算法

人脸识别是当前模式识别和图像处理领域的研究热点。属于生物鉴别技术的一部分。一个完整的人脸识别系统主要由以下几个基本环节构成：图像预处理、人脸检测与定位、特征提取分类识别。本文主要针对图像的特征提取分类识别环节进行分析和试验：首先应用哈尔小波变换初步提取人脸图像的特征;再对小波系数运用核主成分分析进行最终的人脸特征提取。     

脑电信号的混沌分析和小波包变换特征提取算法

针对脑电(EEG)信号的手部动作模式信息处理,提出一种混沌分析和小波包变换相结合的特征提取方法.用眼动辅助来采集手部动作时的脑电信号,对采集的C3、C4 、P3和P4脑电信号消噪后分别用混沌分析和小波包变换的方法进行特征提取,前者提取混沌特征的最大Lyapunov指数和关联维数,组成8维向量;后者提取脑电信号的4种特征节律波,分别计算其相对能量,组成16维向量;最后把两种方法提取的向量组成24维特征向量,输入SVM分类器,实现基于EEG信号的手部动作模式的识别.对不同个体上翻、下翻、展拳、握拳4种手部动作的识别实验表明,平均识别率均在80%以上,明显优于其他方法识别的结果.     

基于小波包分解的意识脑电特征提取

针对2种不同意识任务(想象左手运动和想象右手运动)的脑-机接口(brain-computer interface,BCI)设计,提出了基于小波包分解的特征提取方法。首先深入研究了小波包变换,结合事件相关去同步化(event-related desynchronization,ERD)/事件相关同步化(event-related synchronization,ERS)现象,提出以小波包分解系数来考虑特征,然后对C3、C4导联脑电信号进行小波包分解系数方差和相对能量2种特征的提取,最后采用最简线性分类器进行分类。结果表明,2种特征对应的最大分类正确率均达到了85%,对应时间分别为4.34 s和4.39 s。因此,在保证分类正确率的前提下,所提方法更加简单和有效,为大脑意识任务分类提供了新思路。     

Recognition of wear mode using multi-variable synthesis approach based on wavelet packet and improved three-line method

Condition monitoring is an indispensable means of ensuring smooth running of key equipment, because it can improve machinery availability and performance, and also reduce damage and maintenance cost. One kind of condition monitoring is oil monitoring and it is applied extensively because of its capability to provide warning and to predict faults at early stages, with stronger pertinence. But the extraction and selection of features from oil data have always been the bottleneck of its effective application. In this study, prior to extraction and selection of features, denoising was implemented on the oil spectrometric data using 1D-DPT. For the purpose of mining more effective boundary features, we designed amelioration on classical three-line method based on statistics, and thus improved the three-line method. After the denoised signal was decomposed with WT, the three features, boundary, correlation degree and centroid were extracted, respectively, using the improved three-line method, correlation coefficients and K-means clustering. On the basis of these features, multi-variable synthesis analysis was advanced and the distance criterion parameter of synthesis analysis was proposed to classify and identify wear mode. Finally, through the comparison with examples applying the classical three-line method, we demonstrate the better the ability of the improved method to classify and recognize wear patterns with higher accuracy and precision.     

Virtual prototype and experimental research on gear multi-fault diagnosis using wavelet-autoregressive model and principal component analysis method

Gear systems are an essential element widely used in a variety of industrial applications. Since approximately 80% of the breakdowns in transmission machinery are caused by gear failure, the efficiency of early fault detection and accurate fault diagnosis are therefore critical to normal machinery operations. Reviewed literature indicates that only limited research has considered the gear multi-fault diagnosis, especially for single, coupled distributed and localized faults. Through virtual prototype simulation analysis and experimental study, a novel method for gear multi-fault diagnosis has been presented in this paper. This new method was developed based on the integration of Wavelet transform (WT) technique, Autoregressive (AR) model and Principal Component Analysis (PCA) for fault detection. The WT method was used in the study as the de-noising technique for processing raw vibration signals. Compared with the noise removing method based on the time synchronous average (TSA), the WT technique can be performed directly on the raw vibration signals without the need to calculate any ensemble average of the tested gear vibration signals. More importantly, the WT can deal with coupled faults of a gear pair in one operation while the TSA must be carried out several times for multiple fault detection. The analysis results of the virtual prototype simulation prove that the proposed method is a more time efficient and effective way to detect coupled fault than TSA, and the fault classification rate is superior to the TSA based approaches. In the experimental tests, the proposed method was compared with the Mahalanobis distance approach. However, the latter turns out to be inefficient for the gear multi-fault diagnosis. Its defect detection rate is below 60%, which is much less than that of the proposed method. Furthermore, the ability of the AR model to cope with localized as well as distributed gear faults is verified by both the virtual prototype simulation and experimental studies.     

主成分分析在光全散射特征波长选择中的应用

为了能在用光全散射法测量颗粒粒径时选择对粒径影响较显著的特征波长进行测量,通过在可见及可见-红外波段内对粒径服从单峰R-R分布颗粒系的消光光谱,一阶微分以及二阶微分消光光谱进行主成分变换,提出一种特征波长选择方法。该方法首先对颗粒系的一阶微分消光光谱进行主成分变换,然后将每个波长下的一阶微分消光谱对主成分贡献率的大小作为特征波长选择的主要依据,并将光谱范围的边界波长也作为特征波长。分别对粒径服从单峰及双峰R-R分布的颗粒系进行数值仿真,并采用标准颗粒的实测数据进行验证。验证结果显示,采用基于主成分分析的波长选择方法计算方便、易于实现,得到的标准颗粒粒径反演误差均小于3%,表明采用提出的波长选择方法能够保证选出的光谱消光值具有较高的信息量。     

焊接缺陷的磁光成像小波多尺度识别及分类

针对焊缝微小凹陷、未熔合和焊偏等焊接缺陷,提出了基于磁光成像无损探伤的小波多尺度边缘提取算法及主成分分析-误差反向传播神经网络(PCA-BP)缺陷分类模型;研究了焊件表面及近表面缺陷的可视化无损检测及分类方法。首先,通过对焊件施加感应磁场,利用法拉第磁致旋光原理构成磁光传感器,获取焊接缺陷磁光图像。然后,针对焊接缺陷磁光图像存在噪声干扰、对比度低且成像背景复杂等特征,基于小波模极大值的多尺度边缘信息融合方法,设计了具有高抗噪性的缺陷边缘检测算法。最后,通过PCA法对磁光图像列方向灰度变量进行预处理,得到能表征95%磁光图像列方向灰度变量信息的256个特征点作为输入特征量,构建了三层BP神经网络模型,对焊接缺陷样本进行分类。试验结果表明,所提方法能准确识别微小凹陷、未熔合和焊偏等焊接缺陷,模型分类准确率可达90.80%。     

基于小波变换和ICA特征提取的开关电流电路故障诊断

提出了采用小波变换和独立成分分析(ICA)作为预处理器来进行特征提取的神经网络开关电流电路故障诊断方法。该方法对采集到的故障响应信号进行Haar小波正交滤波器分解,获得低频近似信息和高频细节信息;然后利用独立成分分析方法进行ICA故障特征提取;最后将所得到的最优故障特征输入到BP神经网络中进行故障分类。对六阶切比雪夫低通滤波器和六阶椭圆带通滤波器电路进行了仿真实验验证,获得了100%的故障诊断准确率,与其他方法进行比较,实验结果显示了该方法的优越性。     

基于小波包特征提取及支持向量回归机的光纤布拉格光栅冲击定位系统

以光纤布拉格光栅(FBG)为传感网络,构建了复合材料冲击载荷实时在线监测系统,研究了基于小波包特征提取及支持向量回归机的光纤-碳纤维复合材料结构冲击定位方法.针对同一冲击点,分析不同传感信号,获得了冲击响应信号小波包能量谱,分析结果表明小波包能量谱中特定阶数对冲击敏感.改变冲击点位置研究小波包能量谱与冲击位置之间的关系,提出将第6阶小波包能量值作为冲击定位的特征向量.采用支持向量回归机建立样本数据的回归模型,预测冲击载荷位置,并对支持向量机的相关调整参数进行了优化.实验表明,支持向量机的网络测试误差为4.81％.研究结果可为碳纤维复合材料(CFRP)层状结构的冲击性能评估提供可行的实验方法.     

小波分析在汽车发动机机械故障诊断中的应用

文中阐述了汽车发动机机械故障诊断的理论方法,讨论了小波变换的分析方法.由于小波变换具有传统频谱分析方法所没有的时一频分析特征,特别适合于非平稳信号的分析与处理,应用该方法对实测信号进行了有效的时频分析.     

基于核熵成分分析的光谱重建算法研究

对基于核熵成分分析的光谱反射率重建方法进行了研究,分别采用主成分分析方法和核主成分分析方法构建光谱反射率重建算法进行颜色重建研究,并与基于核熵成分分析算法的光谱反射率进行比较。实验结果表明,基于核熵成分分析的光谱重建算法在色度精度和光谱精度上均优于主成分分析和核主成分分析,对物体表面颜色的真实重建具有一定的应用价值。     

Gear fault diagnosis using active noise cancellation and adaptive wavelet transform

Acoustic signal from a gear mesh with faulty gears is in general non-stationary and noisy in nature. Present work demonstrates improvement of Signal to Noise Ratio (SNR) by using an active noise cancellation (ANC) method for removing the noise. The active noise cancellation technique is designed with the help of a Finite Impulse Response (FIR) based Least Mean Square (LMS) adaptive filter. The acoustic signal from the healthy gear mesh has been used as the reference signal in the adaptive filter. Inadequacy of the continuous wavelet transform to provide good time–frequency information to identify and localize the defect has been removed by processing the denoised signal using an adaptive wavelet technique. The adaptive wavelet is designed from the signal pattern and used as mother wavelet in the continuous wavelet transform (CWT). The CWT coefficients so generated are compared with the standard wavelet based scalograms and are shown to be apposite in analyzing the acoustic signal. A synthetic signal is simulated to conceptualize and evaluate the effectiveness of the proposed method. Synthetic signal analysis also offers vital clues about the suitability of the ANC as a denoising tool, where the error signal is the denoised signal. The experimental validation of the proposed method is presented using a customized gear drive test setup by introducing gears with seeded defects in one or more of their teeth. Measurement of the angles between two or more damaged teeth with a high level of accuracy is shown to be possible using the proposed algorithm. Experiments reveal that acoustic signal analysis can be used as a suitable contactless alternative for precise gear defect identification and gear health monitoring.     

基于PCA的时间分辨油荧光光谱分析及优化

激光诱导荧光技术可广泛应用于油污染的监测中,然而普通的油荧光光谱技术只能实现油污染监测的粗分类,无法区分原油与燃料油的荧光特征。本文基于主成分分析方法(PCA)的时间分辨油荧光分类方法,实验测量了20种油样本的时间分辨荧光光谱特征,给出了对应的荧光寿命和时间分辨油荧光光谱的时序特征。在此基础上,利用前三个主成分构成的三维特征矢量空间,通过分析不同采集时刻下油样本矢量间相关距离的变化,对油样本的时间分辨荧光光谱进行聚类分析。为了体现油荧光变化的时序性,引入矢量距离的离散度参量,提出基于PCA进行时间分辨油荧光光谱分析的优化方法。实验结果表明,基于时间分辨油荧光光谱识别可实现原油与燃料油的光谱时序特征区分,具备良好的油荧光分类效果。     

Gear fault identification and localization using analytic wavelet transform of vibration signal

The aim of this present work is to identify and localize the defect in gear and measure the angle between two damaged teeth in the time domain of the vibration signal. The vibration signals are captured from the experiments and the burst in the vibration signal is focused in the analysis. The enveloping technique is revisited for defect identification but is found unsatisfactory in measuring the angle between two faulty teeth. A signal processing scheme is proposed to filter the noise and to measure the angle between two damaged teeth. The proposed technique consists of undecimated wavelet transform (UWT), which is used to denoise the signal. The analytic wavelet transform (AWT) has been implemented on approximation signal followed by a time marginal integration (TMI) of the AWT scalogram. The TMI graph time-axis is mapped onto the angular displacement of the driver gear. The measurement is shown to identify the first and the second defective teeth impact on gear meshing, which is visible as sharp spikes in the TMI graph. An attempt is also made to replace the approximation from UWT with Intrinsic Mode Function (IMF) derived from the Empirical Mode Decomposition (EMD). The present experimental work establishes the proposed method of measuring and localizing multiple gear teeth defect using vibration signal in the time domain.     

Study of small size wavelet transform processor and wavelet inverse-transform processor using SAW devices

The objective of this research was to investigate the possibility of reducing the sizes of dyadic wavelet transform processor and dyadic wavelet inverse-transform processor using surface acoustic wave (SAW) devices. The motivation for this work was prompted by these processors which are of large sizes. Although these processors are being used, many fields need small size processors, so this work proposes two novel methods of reducing the sizes of these processors: firstly, the architecture which is that various interdigital transducers (IDTs) stand in a line is used to reduce the sizes of these processors. In this architecture, when the electrode-pairs numbers of various IDTs are larger than 20, the bulk acoustic wave (BAW) is eliminated, so this architecture cannot only reduce the sizes of these processors, but also eliminate BAW; secondly, as long as the electrode-overlap envelopes of IDTs will weaken fast with time t, the sizes of these processors are also reduced.     

基于小波包能量谱和NPE的模拟电路故障诊断

提出采用小波包能量谱和邻域保持嵌入作为预处理实行特征提取的模拟电路故障诊断方法。该方法对采集到的故障响应信号进行小波包分解,将不同频带内的能量作为故障特征值,然后利用邻域保持嵌入算法进一步提取故障特征,最后将所得到的最优故障特征输入支持向量机进行故障诊断。仿真结果表明,提出的故障特征提取方法能很好地反映故障响应信号的本质特征,不仅表现出了比其他特征提取方法更好的性能,而且最后的故障诊断中也获得了令人满意的结果。     

基于小波包能量谱和ELM的光伏逆变器多故障在线诊断

以三电平光伏逆变器为研究对象,提出一种多故障模式快速诊断新方法。首先,利用小波包分解提取出三电平逆变器的桥臂电压和上、下管电压信号的能量谱特征向量,并利用主成分分析降维后获取故障特征向量;然后,基于极端学习机诊断模型分离出单器件及多器件开路等多种故障模式。实验结果表明,相比于传统BP神经网络、最小二乘支持向量机故障诊断方法,该方法检测信号易获取,抗干扰性强,诊断速度快、精度高,减小了诊断成本和复杂性,适用于在线诊断。