Search

Consider of influences of noise in sampling signals comprehensively, a method of fault diagnosis which combines matching pursuit (MP) and biomimetic pattern recognition (BPR) is put forward in this paper. Firstly, the matching pursuit (MP) algorithm is used to select optimum wavelets in different SNR situations from the Laplace wavelet dictionary. Then the feature vector is extracted according to the operation result of waveform MP and super high dimensional detection feature spaces of biomimetic pattern recognition (BPR) are constructed. After that, the real-time detected partial discharge (PD) signals are cut, and the feature for each discharge pulse is extracted respectively, which realized the multiple fault recognition revolutionarily. Simulations show that the robustness and accuracy of fault pattern recognition is improved.

A human face recognition algorithm based on greedy kernel principal component analysis (GKPCA) is presented to meet the requirement of quick face recognition on line. In the algorithm, typical human face are decomposed by fast wavelet transform(FWT), then the greedy algorithm is used to reduce training set and the features of the low frequency sub-images are extracted by kernel principal component analysis(KPCA). Consequently, the features extracted are recognized by support vector machine (SVM). Simulations of the algorithm proposed on the basis of ORL (Olivetti Research Lab) face database and NORL face databases show that the algorithm is capable of reducing training time with high recognition rate.

According to the characteristics of partial discharge (PD) signals of gas insulated switchgear (GIS), a fault type recognition method based on wavelet packet transform is put forward. First the wavelet packet decomposition tree of PD signals is constructed. Then in feature extraction of PD signals, different data fusion algorithms are used in feature vector dimension reduction to enhance sensitivity of feature vector to PD fault signals characteristics. A fault recognizer based on back propagation neural network (BPNN) is designed. The results of simulation show that the wavelet transform based GIS fault signal recognition method is effective.

The trajectory of moving people is always random and flexible, which requires the motion model should be more complicated than general motion model, and a dynamical model suitable for maneuvering object is introduced in this paper. The dynamic model named maneuvering model is expressed in terms of the intrinsic coordinate frame for the target and is a good example of a nonlinear model, so we apply particle filter which has good performance in target tracking to track mobile people in WSN (wireless sensor network). According to our computer simulation, we consider that our tracking approach using maneuvering model is more robust and performs more consistently than that with general motion model.

In the paper, we will introduce a new approach of pattern recognition theory. In the control of power systems, the prerequisite of protection device's accurate, fast and reliable performance is its corresponding fault type and fault location can be detected quickly and defined exactly. In our researches, global information will be introduced into the backup protection system, basing on linear discriminant analysis of pattern recognition theory, we are going to seek after for the data characteristics of electrical quantities' marked changes by analyzing and computing real-time PMU measurements, so that we carry out fast and exact discrimination of fault components and fault sections.

Learning is the process of constructing a model from complex world. And machine learning is concerned with constructing computer programs that automatically improve with experience. Machine learning draws on concepts and results from many fields. Obviously, no matter what we adopt new analytical method or technical means, we must have a distinct recognition of system itself and its complexity, and increase continuously analysis, operation and control level. The forecasting analysis of nonlinear complex system will be discussed carefully in this paper. We are mainly using grey forecasting theory to forecast data sequences, and the usual forecast precision has exceeded 90%. In the symbolic forecast of 2-letters nonlinear complex system, the precision of grey forecasting certainly will decrease as the length of symbolic sequence is increasing. But in this place we have found a generating rule that may realize chaotic synchronization at least in short and medium term, and we can analysis and forecast in this way.

This paper proposes an approach called dasiastructure-based spectral clusteringpsila to identify clusters in motion time series for sequential pattern discovery. The proposed approach deploys a dasiastatistical feature-based distance computationpsila for spectral clustering algorithm. Compared to traditional spectral clustering approaches, in which the similarity matrix is constructed from the original data points by applying some similarity functions, the proposed approach builds the matrix based on a finite set of feature vectors. When the proposed approach uses less data points and simpler similarity function to computing the similarity matrix input for spectral clustering, it can improve the computational efficiency in constructing the similarity graph in spectral clustering compared to conventional approach. Promising experimental results with high accuracy on real world data sets demonstrate the capability and effectiveness of the proposed approach for pattern discovery in motion video sequences.

We propose a new method for clustering multivariate time series. A univariate time series can be represented by a fixed-length vector whose components are statistical features of the time series, capturing the global structure. These descriptive vectors, one for each component of the multivariate time series, are concatenated, before being clustered using a standard fast clustering algorithm such as k-means or hierarchical clustering. Such statistical feature extraction also serves as a dimension-reduction procedure for multivariate time series. We demonstrate the effectiveness and simplicity of our proposed method by clustering human motion sequences: dynamic and high-dimensional multivariate time series. The proposed method based on univariate time series structure and statistical metrics provides a novel, yet simple and flexible way to cluster multivariate time series data efficiently with promising accuracy. The success of our method on the case study suggests that clustering may be a valuable addition to the tools available for human motion pattern recognition research.

Discovering knowledge from video data has recently at- tracted growing interest from vision researchers. In this pa- per, we describe a tensor space representation for analyzing human activity patterns in monocular videos. Given a set of moving silhouettes derived from raw video data, the pro- posed methodology first learns a tensor subspace model to embed the silhouettes into low-dimensional projection tra- jectories with preserved temporal order. Symmetric mean Hausdorff distance is then used to measure dissimilarity be- tween the embedded motion trajectories in the tensor sub- space, as the basis for supervised or unsupervised learn- ing. The experimental results on two recent video data sets have shown that the proposed method can effectively ana- lyze human activities with intra- and inter-person variations on both temporal and spatial scales.