Search

Face pair matching is the task of deciding whether or not two face images belong to the same person. This has been a very active and challenging topic recently due to the presence of various sources of variation in facial images, especially under unconstrained environment. We investigate cohort normalization that has been widely used in biomet-ric verification as means to improve the robustness of face recognition under challenging environments to the face pair matching problem. Specifically, given a pair of images and an additional fixed cohort set (identities of cohort samples never appear in the test stage), two picture-specific cohort score lists are computed and the correspondent score profiles of which are modeled by polynomial regression. The extracted regression coefficients are subsequently classified using a classifier. We advance the state-of-the-art in cohort normalization by providing a better understanding of the cohort behavior. In particular, we found that the choice of the cohort set had little impact on the generalization performance. Furthermore, the larger the size of the cohort set, the more stable the system performance becomes. Experiments performed on the Labeled Faces in the Wild (LFW) benchmark show that our system achieves performance that is comparable to state-of-the-art methods.

This work is framed into the context of automatic face recognition in electronic identity documents. In particular we study the impact of digital alteration of the face images used for enrollment on the recognition accuracy. Alterations can be produced both unintentionally (e.g., by the acquisition or printing device) or intentionally (e.g., people modify images to appear more attractive). Our results show that state-of-the-art algorithms are sufficiently robust to deal with some alterations whereas other kinds of degradation can significantly affect the accuracy, thus requiring the adoption of proper detection mechanisms.

This paper investigates the problem of exponential input-to-state stability (e-ISS) for hybrid dynamical networks (HDN) via impulsive interconnection. New concepts of input-to-state exponent property (IS-E) and augmented time are proposed for dynamical systems and hybrid systems respectively. By using IS-E estimations of nodes in HDN and methods such as multiple Lyapunov functions and hybrid time, two types of e-ISS criteria for continuous-time/discrete-time HDN are established respectively. The requirements on ISS property of every subsystem and small-gain condition for interconnection in interconnected systems or networks in the literature is relaxed. The obtained e-ISS results are extended to the case of delayed impulsive interconnection. One representative example is given to illustrate the theoretical results.

With the vast amount of information arriving each day, it is necessary to develop automatic techniques for analyzing and handling these huge volumes of information. This problem is addressed by Topic Detection and Tracking (TDT), which organizes news stories by topics, and each topic is viewed as a flat collection of news stories. However, a topic in news is not only a flat collection of news stories but also a set of events. Additionally, there exists a three-layer hierarchy (topic → event → story), which can make people hold the new things that happen in the news easily. Therefore, to recognize the events in topics is significant. Unfortunately, the similarity between two stories, which belong to different events in a topic, is usually high. This is induced by common words occurring in both the two stories. And these common words usually cause events in the same topic to be mutually confusing. To address this problem, we present a novel approach for event identification in this paper. First, we need to remove topic-specific stopwords from each story, then some named-entities are selected as part of features due to their high distinguishable characteristic for identifying events. There is another issue deserving of in-depth consideration. We know weights on different features were empirically determined in the previous work. In our work, we propose a new method to calculate these weights. The experiments are implemented on a Linguistic Data Consortium dataset. The experimental results show that our scheme for event identification has significant improvement over the previous methods.

With the application of numbers of non-linear power load such as arc furnace and transducer, the interharmonics problem emerges in the power system besides of the tranditional harmonics. It is not an integer multiple of fundamental and usually with small amplitude. In terms of these identities, some of the old methods to measure harmonics cannot work well. In this paper, a new approach - RELAX algorithm for sinusoidal parameter estimation is presented to make interharmonics measurement and the advantage of this method in comparison with others is also involved. The simulation with Matlab validates the approach proposed.

Wrist pulse signals can reflect the pathological changes of a person's body condition due to the richness and importance of the contained information. In recent years, the computerized pulse signal analysis has shown a great potential to the modernization of traditional pulse diagnosis. In this paper, we attempted to use the wrist pulse signals collected by a Doppler ultrasonic blood analyzer to perform wrist pulse signal diagnosis. We first cropped the wrist pulse signal to obtain the single-period waveform, and then employed KPCA to extract features from the waveform. Finally, we used a nearest neighborhood classifier to classify the extracted features. We adopted a wrist pulse signal dataset, which includes pulse signals from both healthy persons and patients. Several experiments on the dataset were carried out and the results show that our developed approach is feasible for computerized wrist pulse diagnosis.

Aiming at problems of multiple solutions in the existing optimizing models for power fault-section estimation, the improved model was established considering the influence of the reclosure, the different protection and joint influence between main and backup protections. Besides, quantum immune algorithm was presented for fault model in order to overcome many populations and slow convergence of genetic algorithm. Quantum immune algorithm codes the chromosome by quantum bit probability, and makes the populations evolve by clonal selection and quantum rotation gate, so rapid convergence, small populations and good global search capability characterize the performance of quantum immune algorithm. Test results show that, the improved model is logical and quantum immune algorithm has better comprehensive performance than genetic algorithm, which proves that the algorithm is feasible.

The method based on wavetlet singular entropy(WSE) and artificial immune recognition system (AIRS) for transmission line fault classification is presented in this paper. Wavelet singular entropy is used to quantify uncertainty of fault high frequency transient voltages so as to reflect and identify various failure states of power system. On this basis, AIRS for fault claasification is prresented to overcome the shortcomings of artificial neural network (ANN) and support vector machines (SVMs). The classifer can also decrease number of input parameters, relieve the dependence on prior knowledge of decision maker and improve generalization ability. The simulation results show the method is effective and correct.

Broadband power-line communication (BPLC) is recognized as an attractive method for access networks. And the orthogonal frequency division multiplexing (OFDM) with dynamic resource allocation is the key technology in combating with the multi-path fading and time-varying characteristics of power line channel. With the goal of fairness in bandwidth and maximum total throughput for all the users, a novel algorithm based on genetic algorithm is proposed in this paper. In order to accelerate the convergence speed, some sophisticated genes are adding in the initial population. And we simulation this method with a typical power-line access networks and demonstrated the effectiveness under various fairness ratios.

Distributed generation (DG) is a hot spot nowadays, at the same time the harmonic pollution can't be ignored when DG is connected with the power grid through power converters. First discuss several harmonic suppression methods have been adopted in practice, such as increasing the number of converter pulse, using pulse width modulation (PWM) converter, passive power filter (PPF), active power filter (APF) and so on, among these methods, Hybrid type combining PPF and APF is the most economical and effective method, so a Hybrid filter constructed of a shunt active filter and distributed passive filters is presented and the parameters of the passive filter are discussed. Genetic algorithm (GA) is an important method of multi-objective optimization. So using GA optimizes the PPF parameters, namely R, L, C, from three respects of the cost, the DG penetration power and the total harmonic distortion (THD), optimization results make the PPF cost is the most economical, the DG penetration power is the largest and the THD is the smallest. At last, simulation shows that PPF compensates a majority of harmonic current, APF compensates a few harmonic current and prevents series/parallel resonance, at the same time, the level of APF power is very small because of avoiding to fundamental voltage.

Increasing demand for premium electric power, in terms of both quality and reliability, and emerging new energy technologies have led to the development of DG. In the process of connected DG with the power grid, harmonic/inter-harmonic is very serious problem. A new method based on modified MM filter and HHT is proposed to detect harmonic/inter-harmonic caused by the power electric elements. Researched intensively the characteristics of MM transform and MM filter, MM filter is modified aimed at the drawback of selection of structure elements and statistic deflection. HHT is perfectly self-adaptive analysis method. To the signal with harmonic/inter-harmonic, white pulse and positive & negative impulse, firstly, using modified MM filter to remove the white noise, positive & negative impulse and high times harmonic because of its low-passing characteristic, that method's key element is to choosing appropriate SE and its filter effect is better than the wavelet transform. Based on this, using EMD to get a set of IMFs for the denoised signal, the IMFs are almost the single frequency signal. At last, carrying out Hilbert transform to IMFs can get the instantaneous frequency and amplitude of every harmonic/inter-harmonic accurately.

Nowadays, with the rapid development of distributed generation, it has an increasing rate of permeation, the islanding phenomenon in grid-connected run mode brings hazards to network, electrical equipment and life safety. So, it is necessary to effectively detect the islanding condition and swiftly stop the run mode of grid-connect. In this paper, three-phase inverter and network systems was taken as an example, several existing methods were introduced and the advantages and disadvantages were compared of several existing methods by using MATLAB simulation software for the realization of the simulation, the proposed combination of a variety of detection methods for the detection program.

Genetic algorithm (GA) is an important method of multi-objective optimization. The method is introduced that using GA optimizes PPF parameters in this thesis. Distributed generation (DG) is a hot spot nowadays, but the harmonic pollution isn't ignored when DG is connected with the power grid through power converters. First discuss several harmonic suppression methods have been adopted in practice, such as increasing the pulse number of converter, using pulse width modulation (PWM) converter, passive power filter (PPF) equipment, active power filter (APF) equipment and so on, among these hybrid type combining PPF and APF is the most economical and effective method, so hybrid type equipment is applied to suppress harmonics that are brought by grid- connected DG. From three respects of the cost, the DG penetration power and the total harmonic distortion (THD), using GA optimizes the PPF parameters, namely R, L, C, optimization results make the PPF cost is the most economical, the DG penetration power is the largest and the THD is the smallest, simulation shows this method is effective.

A new method based on improved multiple signal classification (MUSIC) algorithm and independent component analysis (ICA) method is applied in this paper. In contrast with the original MUSIC algorithm, the proposed version can give good performance at lower signal/noise ratio (SNR). After estimating the frequencies accurately, the amplitudes and phases of interharmonics are calculated with ICA method. Compared with some other algorithms, the improved MUSIC algorithm obtained higher resolution in frequency domain and ICA method separated fixed signal into several independent signals. The simulation indicates the validity of this method.

The grid-connected operating mode of distributed generation (DG) is now the trend of the development of DG. In this paper, the topology and mathematical model of a grid-connected system of DG are studied deeply, instantaneous reactive power theory and hysteretic comparison control strategy have been chosen, and a grid-connected system is got which can send the active power to the grid, compensate harmonic and absorb the reactive power that the local loads generated. Then the simulation models of main circuit and control circuit of the grid-connected system of DG have been established using MATLAB, and the two different filter modes of the grid-connected system are analyzed and simulated which verified the grid-connected system reliable and valid.

Indoor power line is considered as one of the most attractive transmission media for home network, because it pervasive exists in almost every residential. In this paper, a novel home network integrated with power line and wireless communication is presented. However, the power line has the hostile channel for transmission high frequency signal. To design home networks better, the channel characteristics including attenuation and noise performance are analyzed. According to Shannonpsilas law the theoretical channel capacity is calculated. And the calculation result shows that the power line channel capacity is able to provide the Internet Protocol TV (IPTV) and other home network requirements.

This paper presents a channel model of the indoor power line for broadband power line communication (BPLC). Considering transmission theory, the two-port network model of power line has been established in this paper. And according to this model, characteristics of power line channel under the different network configuration and terminate loads have been analyzed. The model has been verified with practical measurements on actual power network. It is demonstrated that the model based on two-port network theory can accurately reflect the situation of power line channel. And the result shows that the characteristic of channel is variation with frequency and network topology.

The electric power load is inherently nonlinear and nonstationary time series, which has periodicity and randomicity. It is difficult to accurately forecast with a single method. So a hybrid method based on empirical mode decomposition (EMD) and support vector machines (SVMs) was presented in this paper. First, load time series was adaptively decomposed into a series of smooth intrinsic mode functions (IMFs) with different scales via EMD; second, tendency of each IMF was forecasted with different SVM respectively according to its change and weather factors were taken account in forecasting; finally, final results were obtained by summing up these forecasting results of each IMF together. The simulation results show that the hybrid method has higher precision and greater generalization ability.

A new face recognition method based on the simple gabor feature space(SGFS) and support vector machine (SVM) is presented here. Firstly, the key points of face image were filtered by the Gabor filter bank to obtain the SGFS feature matrices. Secondly the SGFS feature matrices were combined together to establish the face simple Gabor feature, representing the face. Finally, the SVM classifier was built to classify the face simple Gabor feature to realize the face recognition. The Experiment results based on ORL database show that, this method has small mount of key feature point, the max recognition rate is 95% and it is feasibale.

Because the problem of power quality is increasing complex, the research is very important. In the text, it introduces firstly the Power Quality disturbance and the ways of Power Quality Monitoring. There are FFT (Fast Fourier Transform), STFT (short time discrete Fourier transform), Wigner and wavelet. The paper indicates some flaws in those ways. Secondly, it introduces briefly Independent Component Analysis (ICA) and Empirical Mode Decomposition (EMD) arithmetic. With that, the paper put forward a new method that based on the ICA and EMD arithmetic to monitor and analyze power quality disturbance. It makes a simulated experiment. In the part one of the experiment, it simulates voltage sags, voltage swells, transient surges and voltage fluctuations. In the part two of the experiment, it simulates power quality monitoring in the condition of noise and interference. The result of experiment indicates the new way is effective in the Power Quality Monitoring and Analyzing System. (5 pages)

Independent component analysis (ICA) is a recently developed method used to separate unknown source signals from their mixtures. This paper describes some representational algorithms of ICA in brief, such as the fast-ICA, the EGLD-ICA, the kernel-ICA and the MS-ICA. The simulative tests show that the application of ICA in the fields of signal processing and communication, such as extraction weak signal, separation in audio signals and digital image watermarking, is widely and effectively. At the same time, the performances of these kinds of ICA algorithm are contrasted and the results are shown by the form of figures and tables in this paper. And the conclusion tries to discuss the selection of type of ICA algorithm in different practical application.

Smart antenna technology, as one of the key technologies of the wireless communication, increases the traffic capacity of a system, enlarges the coverage range of a cell, and improves the quality of the signal. It is therefore of practical interest to research smart antenna technology applied to wireless communication. In this study, we focus on FASTICA (fast independent component analysis) algorithms with smart antenna at the wireless communication. It can reduce interference and suppress multi-path effect in the up (reverse) link. At first, we introduce a general overview of smart antennas. Second, we narrate the FASTICA algorithms. At last, we utilize FASTICA to separate the demodulated mixed signals. The simulation study makes know FASTICA arithmetic is a new valuable way and is able to distinguish communication signal from many different noises effectively.

In order to accurately collect the electricity usage information from the smart meter which uses the power line for communication, this paper proposes the method of dynamic networking to enhance the reliability of the smart meter communication. We shall firstly establish a logical topology between the smart meter and the concentrator with reference to their communication paths within the power supply range of the same transformer, and then grade smart meters, and choose the relay for each level network based on the selection methods of relay, and finally use the improved ant colony algorithm to choose the optimal communication path for smart meters in the communicref1ref1ation range of multiple relays. Although the optimal path can be discovered quickly in this way, it cannot establish other communication paths in time when the collected data are abnormal. Therefore, on the basis of the characteristics of data, this paper uses the probabilistic neural network to determine the integrity of the data collected. If the collected data is abnormal, we will reconstruct communication network by choosing a new communication path between concentrator and smart meters. By means of the MATLAB simulation, we can automatically organize the network to improve the reliability of communication between the concentrator and smart meters, which is of great significance for the concentrator to collect the users' electricity usage information.