Search

Classification is one of the most important research topics in hyperspectral image (HSI) analyses and applications. Although convolutional neural networks (CNNs) have been widely introduced into the study of HSI classification with appreciable performance, the misclassification problem of the pixels on the boundary of adjacent land covers is still significant due to the interfering neighboring pixels whose categories are different from the target pixel. To address this challenge, in this article, we propose a center attention network for HSI classification. The proposed method simultaneously captures spectral-spatial features of the target pixel and its neighboring pixels for classification. Specifically, the method adopts a center attention module (CAM) that pays more attention to the features which are more correlated with the target pixel, that is, the central pixel of the sample, and then sums up the weighted features to generate more relevant and discriminative features. In this way, our method has a high potential for improving the performance of HSI classification. In addition, the CAM greatly reduces the number of parameters in the network via weighted sum of the spectral-spatial features, thus improving the computing efficiency while still maintaining classification accuracy. We evaluate the proposed method on three public datasets, and the experimental results demonstrate the superiority of our method on accuracy and efficiency compared with several state-of-the-art methods.

Identifying geochemical patterns from backgrounds and generating associated mineralization remains challenging due to the complex structure of mineral deposits. To learn how to identify geochemical anomalies that are spatially associated with mineralization, we need in-depth knowledge of the dependence process. Quantitative association rules (QARs) are applied to discover remarkable relations and dependencies between attributes in a dataset, but it is difficult to generate relationships from geochemical data. In previous studies, no methodology to find association rules is proposed to deal with geochemical data problem, and the classical methods designed for Boolean and nominal attributes require previous discretization, which makes the whole process limited in processing complex data. In this paper, we proposed a hybrid method of graph clustering and quantitative association rules (GCQAR) as a new way of identifying significant geochemical patterns. Graph Clustering (GC) is used as partitioning paradigm because of its ability to handle large-scale datasets. The GC is based on modularity to effectively generate the groups of the graph, to avoid the over-partitioning, and to cover all the rules. In each partition, a set of geochemical quantitative association rules is produced. The results obtained in the experimental study performed on data collected in the field of Xiaoshan, Henan province, China. Our GCQAR has significant benefits in terms of recognition geochemical patterns compared to the traditional methods used in the field of geochemistry.

When ultrasonic guided wave are used for icing detection, the multimodal and dispersion characteristics of guided wave make the guided wave signal very complicated, and it is difficult to identify time domain signal directly. Wavelet packet decomposition has excellent time-frequency resolution characteristic, and singular value decomposition can extract important features of time-frequency matrix. Therefore, a guided wave signal processing method based on wavelet packet decomposition-singular value (WPD-SV) is proposed, which successfully extracted icing related signal features, and the finite element simulation and experimental results show that the quantitative detection of wing icing is realized.

It is of great significance to detect the weld surface to ensure welding quality. Aiming at the problem of weld surface defect detection and 3D reconstruction algorithm, a method of defect detection and three - dimensional reconstruction based on laser scanning 3D point cloud has been proposed. The method of BP neural network is used to filter the point cloud data obtained by laser displacement sensor scanning, and the result is qualitatively and quantitatively analyzed. The three-dimensional reconstruction model of pit defects on weld surface is established. The reconstructed error is within the allowable range. The experimental results show that applying BP neural network to weld defect detection can effectively reduce the complexity of 3D reconstruction and improve the accuracy of 3D reconstruction of weld surface.

Carbon fiber composite has good material properties, but it is easy to cause defects during the process of preparation and performance. Delamination is one of the main forms of failure, which directly affects the safety of the material. Due to the dispersion and multimodality of ultrasonic guided waves, it is necessary to select sensitive modes for defect detection. In this paper, firstly, the dispersion curve and wave structures are calculated by semi analytical finite element method. Then the finite element simulation model in the frequency domain is established by COMSOL. The mode-frequency points with better detection effect are selected by analyzing the displacement changes, and the frequency sweep experiment is carried out to verify the simulation result. The result shows that mode 4 at 480kHz is suitable for the detection and identification of the delamination defects with different sizes.

The effective removal of aircraft icing is the necessary guarantee for the safe flight of aircraft. The construction of a deicing system with low power consumption, lightweight, and high reliability is the new requirement of deicing technology development. In this paper, a low-power deicing method based on the shear stress generated by Lamb waves excited by the piezoelectric patch is proposed. The ability of Lamb waves with different modes to produce interface shear stress under a given input power is studied. It is found that Lamb waves can produce large shear stress in the (0-0.5) MHz band of A0 mode and (1-1.5) MHz band of S0 mode. At the same time, the influence of piezoelectric patch thickness on the interface input shear stress is analyzed by simulation. It is found that the interface input shear stress decreases monotonically with the increase of piezoelectric patch thickness. Finally, the optimal Lamb wave mode and the size of the piezoelectric patch are determined.

With rapid development of road and traffic construction, China achieved tremendous growth in social and economic progress, yet also caused a mass of road traffic accidents and suffered great losses in people’s lives and property. Through opening the “black box” the road traffic system was decomposed into two subsystems: accident occurrence and accident loss. A serial two-stage additive DEA model was employed to analyze the performance of road traffic safety of 31 provinces in China. The results effectively show regional differences in road traffic safety, identify the key link affected the whole performance, and provide some feasible suggestions for government traffic management department

Several electric vehicle (EV) charging algorithms have been recently suggested to meet different optimization objectives in power grids. In this paper, a holistic mechanism is proposed to manage the operation of EV charging stations (EVCSs). The proposed framework offers adaptive operation strategies for the EVCS operators to effectively manage the EVCS under different penetration levels of EVs, considering both normal operating conditions and restoration processes during interruptions and emergencies. The performance of the proposed approach is tested with real data and numerically analyzed on the system operational costs under different EV and PV penetration scenarios. We also demonstrate that the proposed adaptive operation mechanism could bring significant advantages to the operation and control of power grids when facing different operating conditions.

Hyperspectral data contains abundant information in spectral domain, which is very useful for mineral classification and geological body mapping. But, due to the lack of labeled data, it is difficult to get an acceptable result by just using the small number of labeled data. We adopt a semi-supervised method called CNN, which can effectively extract inner features of hyperspectral image to classify hyperspectral data. However, with constraint to the size of receptive field, it can hardly get higher level features. We propose dilated CNN for mineral classification of hyperspectral data. At the same size of kernels, dilated CNN has bigger receptive field. At the meanwhile, it can get higher accuracy of classification. We test our model on hyperspectral data at Tianshan area, where is rich in minerals. From the result, we can find that our method can get a great result on the mineral classification task, which can be used for making geological map.

Classifying Hyperspectral images with few training samples is a challenging problem. The generative adversarial networks (GAN) are promising techniques to address the problems. GAN constructs an adversarial game between a discriminator and a generator. The generator generates samples that are not distinguishable by the discriminator, and the discriminator determines whether or not a sample is composed of real data. In this paper, by introducing multilayer features fusion in GAN and a dynamic neighborhood voting mechanism, a novel algorithm for HSIs classification based on 1-D GAN was proposed. Extracting and fusing multiple layers features in discriminator, and using a little labeled samples, we fine-tuned a new sample 1-D CNN spectral classifier for HSIs. In order to improve the accuracy of the classification, we proposed a dynamic neighborhood voting mechanism to classify the HSIs with spatial features. The obtained results show that the proposed models provide competitive results compared to the state-of-the-art methods.

Delamination defects are one of the typical defects of composites materials. Due to the complex distributions of delamination defects, it is hard to use one mode of guided waves to detect delamination defects with different distributions. Thus a multimode testing method is highly demanded. Apply semi-analysis finite element (SAFE) method to calculate dispersion curves and wave structure. Several mode-frequency points from dispersion curves will be selected to do finite element simulation for delamination defects with different sizes but the same depth of composite materials plates. Then analyzes the sensitivity of guided waves for different delamination defects. The sensitive modes are selected to do defects testing experiments for composite materials plate with different delamination defects. The results show that using two or more mode-frequency points can detect out delamination defects with different sizes.

Crime scene investigation (CSI) image retrieval is used to search for crime evidences and is critical in helping in solving various crimes. In recent years, using Convolutional Neural Network (CNN) has demonstrated outstanding performances in large-scale image database retrieval. However, to prevent over-fitting in the training of CNN model due to limited number of CSI images, this paper proposes to cascade two CNN models obtained based on transfer learning and combine CNN features with low-level image feature to better describe CSI images. First, two pre-trained CNN models are fine-tuned using the target image set. CNN features are extracted from fully connected layer of each model and are concatenated as high-level features for the image. These concatenated CNN features are then fused with the low-level image features of the target image set. The final fused image features are used in the image retrieval. Experimental results on CSI image database proved the effectiveness of the proposed algorithm for limited number of training sets. In addition, experiments carried out on the GHIM-10K database proved the generalizability of the proposed algorithm.

Based on a large scale crime scene investigation (CSI) image database, an effective and efficient CSI image retrieval system has been proposed to empower the investigative work of the police force. The main contribution of this paper includes: (1) a DCT domain texture feature extraction algorithm is proposed for CSI images, which is shown to be simple and effective. (2) the use of GIST descriptor on CSI images for the first time and combined with color histogram and the DCT domain texture feature as a fused feature, which describes CSI images from different aspects including color, texture, and scene content. Experimental results prove that the proposed method is effective for CSI image retrieval.

Crime scene investigation (CSI) images are important clues for the police to solve cases. With the available of large scale CSI image database, effective and efficient CSI image retrieval becomes more and more important. The main contribution of this paper includes: (1) a DCT domain texture feature extraction algorithm is constructed for effective description of CSI images; (2) GIST descriptor is first exploited in the description of CSI images and integrated with color histogram and the DCT domain texture feature as a fused feature, which describes CSI images from different view including color, texture, and scene structure; (3) SVM Classification technique is used in CSI image retrieval. Experimental results on real CSI image data show that the fusion feature proposed in this paper can well describe the content of CSI images, with an average 15.3% increment in retrieval precision compared with all the single-feature-based algorithm. Using this fusion feature to further train the SVM classifier in the retrieval process, the precision is further improved by 3.1%.

In the field of biorobotics, a simple, high accuracy and programmable human walking simulation model are critically needed to investigate human walking mechanism, design bio-inspired robot and develop rehabilitation device. In this paper, we combine multibody dynamics and biomechanics to develop novel method for human walking modeling. Firstly, a human bone model with accurate anthropometric data is constructed based on reverse engineering to solve the problem that the human model is not accurate enough, and then the model motion data is obtained by human gait experiment to simulate the walking process in ADAMS. Finally, the walking simulation results used to verify the human walking model. We hope to provide a more efficient and accurate human modeling and walking simulation method for researchers in related fields.

Although non-convex fuzzy set (FS) has high potential in modeling data and controlling, it is seldom used and discussed because of the lack of linguistic explanation. A method named “parametric qualitative fuzzy set (PQ FS) plus choice strategy” is put forward in this paper for the construction of non-convex FS, in which PQ FS is defined as a collection of convex FSs with specific structure and choice strategy is an approach to choose one convex FS in PQ FS for each situation. Some theorems are shown to demonstrate that the way “PQ FS plus choice strategy” can construct a non-convex FS by some proper choice strategy. Furthermore, the algorithm of constructing given non-convex FSs is also proposed by homotopy. Based on the construction way, the linguistic explanation of non-convex FS is obtained. The result shows that Non-convex FSs do not violate the routine of representing linguistic variables by convex FSs, for linguistic viable is represented by convex FS at each situation and non-convex FS is just the trajectory of those chosen convex FSs under all the situations.

The growing share of electricity production from natural gas combined with the scarcity of firm contracts for fuel procurement creates vulnerability when gas supplies are tight. To examine the effects of uncertainty in the natural gas spot price on the dispatch cost, we formulate a model of economic dispatch that accounts for limits on fixed price gas availability from contracts. Probability distributions of the hourly electric load and a daily natural gas spot price are generated using a temporal and weather conditional epi-spline based load prediction method and the kernel density estimation method, respectively. Monte Carlo simulations are conducted with randomly generated electric load variates both with and without uncertainty in the gas spot prices, to generate the corresponding dispatch cost probability distributions and quantify the effects of gas price uncertainty using the mass transportation distance. The proposed method is tested on a 24-bus power system with a 20-node gas system.

Constructing fuzzy set models from data is a key step in computing with words (CWW). The enhanced interval approach (EIA) is one of the most important approach for constructing interval type-2 fuzzy set (IT2 FS) from data intervals. This paper proposes an interval approach (NIA), which introduces a normal distribution associated with a free parameter (FP) and Gaussian footprint of uncertainty (FOU) as the supplement of uniform distribution and trapezoidal FOU. The NIA includes a data part, fuzzy set (FS) part and FOU part. The data part maps data intervals to probability distributions. The FS part encodes the distributions in the data part to fuzzy membership functions. Gaussian membership function (MF) is discussed and the parameter transformation is obtained. In the FOU part, the FOU of IT2 FS is built by collecting the obtained T1 FSs. The way to construct a Gaussian FOU is, for the first time, proposed in this paper. Owing to FP, the FOUs could be consist with data statistics. The effectiveness of NIA is validated by modeling a group of words, showing that NIA has greater flexibility and accuracy when approximating the distributions of data intervals.

The deep learning based trackers can always achieve high tracking precision and strong adaptability in different scenarios. However, due to the fact that the number of the parameter is large and the fine-tuning is challenging, the time complexity is high. In order to improve the efficiency, we proposed a tracker based on fast deep learning through constructing a new network with less redundancy. Based on the theory of deep learning, we proposed a deep neural network to describe essential features of images. Furthermore, fast deep learning can be achieved by restricting the size of network. With the help of GPU, the time complexity of the network training is released to a large extent. Under the framework of particle filter, the proposed method combined the deep learning extractor with an SVM scoring professor to distinguish the target from the background. The condensed network structure reduced the complexity of the model. Compared with some other deep learning based tracker, the proposed method can achieve higher efficiency. The frame rate keeps at 22 frames per second on average. Experiments on an open tracking benchmark demonstrate that both the robustness and the timeliness of the proposed tracker are promising when the appearance of the target changes containing translation, rotation and scale or the interference containing illumination, occlusion and cluttered background. Unfortunately, it is not robust enough when the target moves fast or the motion blur and some similar objects exist.

Previous studies show that language differences cause corresponding influences on Second Language Acquisition. Based on the Corpus of Chinese, English, and Tibetan spoken by Tibetan Speakers (CETTS), this study conducted English stress production and perception experiments and analyze the English stress acquisition by native speakers of Tibetan, to find out whether stress position and syllable structures will affect their stress acquisition. By analyzing 2,176 productions and 4,352 perception response by 17 subjects, as well as adopting ANOVA analysis and Mann-Whitney U test, this study finds out that: (1) The Tibetan group has a lower accuracy rate in English stress production and perception experiments; besides, unlike Americans, native Tibetan speakers do not employ the range of F0 to realize English stress. (2) Stress position affects stress production by the Tibetan group, but exerts no effect to stress production by the American group, and stress perception by both groups. (3) The tonal distribution rule of Tibetan does not affect the acquisition of English stress. (4) Schwa makes no statistical difference in accuracy rate, but can shorten the subjects' response time. Such findings will provide empirical data for more effective teaching of English to minority groups.

The radio access network (RAN) cache is recently proposed to improve quality of experience perceived by mobile users on Internet content access and reduce intense pressure on the backhaul network. However, because RAN caches (e.g., associated with base stations) are located below mobility management entities in mobile core network, the continuity of content service for the handover users with unfinished RANcache content transmission is destroyed. Thus motivated, we design the seamless RAN-cache handover framework based on mobility prediction algorithm (MPA), where the smart terminal with unfinished content service during mobility predicts the break and the target handover access node (such as the BS or the WiFi access point), and pre-triggers the source RAN cache to return the content service to CDN servers outside the core network (CN) or to notify the RAN cache associated with the target handover access node to prepare for serving the handover user. The framework incurs no change to cellular infrastructure including RAN and CN. Different from prior arts of assumption that users' mobility is pre-known, we design the parallel autoregressive (AR) based received signal strength (RSS) prediction approach to achieve mobility prediction, and mathematically define its false probability during different tolerant intervals. The simulation scenario deployment engine and the user movement engine are designed, considering practical issues. Through simulations, the the false probability of seamless RAN-cache handover pre-trigger through MPA is less than 1.36%, and it can be guaranteed less than 1% as the maximal RAN-cache handover processing time is increased by 8%.

Distributed caching in the radio access network (RAN) has been a common approach for improving QoE per- ceived by mobile users and reducing backhaul traffic load. In this paper, we focus on the problem of cooperative and online searching and routing for content objects cached according to their local popularity in small- cell networks (SCNs), considering practical issues of implementation complexity and content prefer- ence diversity, which is different from prior arts mainly focusing on global content caching optimization with pre-knowledge of content requests and/or mobility pattern of users. We propose an in-RAN caches orchestration strategy (named as iCacheOS), where cooperative content searching and sharing is achieved through content joint wireless and backhaul routing algorithm. Specifically, 1) the in-RAN caches orchestration framework is designed under the software defined networking inspired system architecture consisting of the central orchestrator and in-RAN caches with the local routing function; 2) the content joint wireless and backhaul routing problem is constructed with the objective to minimizing the SCN content transmission delay, which is proved to be a quadratic assignment problem (QAP); 3) to assure high computation efficiency, a joint dimension reduction, Lagrangian relaxation and accelerated branch and bound algorithm is pro- posed. Numerical results reveal that iCacheOS can significantly reduce the content transmission delay and outgoing backhaul traffic load, and is more robust to content preference diversities, in comparison with other strategies.

Inspired by recent advances in deep learning, this paper reviews the deep learning methodologies and its applications in object tracking. To overcome the complexity and low-efficiency of existing full-connected deep learning based tracker, we use a novel convolutional deep belief network (CDBN) with convolution, weights sharing and pooling to have much fewer parameters, in addition to gain translation invariance which would benefit the tracker performance. Empirical evaluation demonstrates our CDBN based tracker outperforms several state-of-the-art methods on an open tracker benchmark.

Ambient backscatter is a new communication technology that utilizes ambient radio frequency signals to enable battery-free devices to communicate with each other. In this paper, we study the problem of signal detection and bit error rate (BER) performance for this novel communication system that utilizes ambient backscatter and multiple receiving antennas. Specifically, we first build up the theoretical system model and then propose a maximum likelihood detection method. We also obtain the optimal closed-form detection threshold and the corresponding BER expression. It is found that increasing the number of the receiving antennas can bring about larger BER gains at low SNR, while little gains in high SNR. Finally, simulation results are provided to corroborate our theoretical analysis.

Distributed and cooperative caches (DCCs) enabled small-cell network (SCN) has been a common approach for improving QoE perceived by mobile users (UEs) and reducing backhaul traffic load. In this paper, we propose a joint wireless and backhaul load balancing (JWBLB) framework for DCCs enabled SCNs with the objective to minimizing the system content transmission delay. The JWBLB is achieved through content-request UEs and small-cell base stations re-association, and back-haul links and cooperative caches selection, with the knowledge of the cached content distribution, jointly considering traffic load of wireless and backhaul links. Specifically, 1) a software defined networking (SDN) inspired load balancing control architecture, consisting of the coordinator and DCCs, is designed. 2) The optimal JWBLB problem is formulated with the objective to minimizing the system content transmission delay, which is proved to be a quadratic assignment problem (QAP). 3) A joint Lagrangian relaxation and decomposition, and accelerated branch and bound algorithm is proposed to efficiently find the optimal binary solutions of the considered QAP. Simulation results show that JWBLB achieves great reduction in the average transmission delay and outgoing backhaul traffic load. For instance, 65% average transmission delay can be reduced as only 50% UEs with potential re-association possibility, and 48.4% outgoing traffic load caused by content fetches from Internet is reduced as the Zipf distribution parameter equals 0.6.

Due to the vast amount of bug reports in many large-scale open source projects, detecting duplicate reports has received much attention in recent years. Textual information and different kinds of categorical information have been leveraged in many detection methods. Although past studies have shown the influence of the temporal information in the submission interval of duplicate bug reports, only few methods consider the temporal information in its plain interval form. In this paper, we propose a temporal model to improve the detection performance by considering the submission time and the version information. We discuss the effectiveness by implementing it based on a prestigious BM25Fext-based method. Empirical experiments have been conducted on a dataset collected from the Eclipse open source project. Compared with the baseline BM25Fext-based method, the proposed scheme can obtain performance improvement consistently.

The gap between the stable system resource supply and the traffic load (i.e, the demand) with temporal fluctuations and spatial disparities results in deficiency in the system resource (e.g, power and spectrum resources) utility in cellular networks. To bridge the variant gap, cell breathing schemes are adopted in prior arts, which mainly focus on the temporal variance. Therefore, there is every reason to further explore the improvement room in terms of efficiency in the system resource utilization of cell breathing, jointly considering temporal fluctuations and spatial disparities of the traffic load, especially under Ultra-Dense HetNets (UDHNs), the promising delivery of 5G networks. We propose the High-Resolution Cell Breathing (HiRCB) strategy for improving energy efficiency for UDHNs: 1) inspired by pixels in digital imaging, the UDHN coverage area is logically divided into Traffic Lattices (TLs), capturing spatial disparities of the traffic load; 2) the TL association problem with the objective of maximizing the accumulative energy efficiency utility is formulated; 3) to solve the formulated problem, the Parallel and Integer TL Association Algorithm (PITLAA) is proposed, featuring parallel computations and algorithmic conversions from multiple to single TL association. Numerical results show that HiRCB improves at least 50% energy efficiency.

STSA(symbolic time series analysis) is a effective tool used in many fields to measure randomness and deterministic of nonlinear system. In order to investigate the complex characteristics of radar intrapulse modulation signals in modern electronic warfare, the entropy statistics of STSA is defined in this paper. The symbolic entropy algorithm of STSA based on information theory, chaos time series analysis and symbolic dynamics is developed for quantitatively extracting intrinsic information of radar emitter signals. Simulated signals derived from seven typical radar modulation signals and their overlapping signals are examined and compared when SNR varies in the range from 0dB to 30dB. The simulation results demonstrate that the introduced STSA method has better characteristics of separating the same signals with different modulation parameter and the outlier signals overlapped by different modulation signals. Moreover, the algorithm can be easily implemented and computationally efficient.

Focus on constructing of general type-2 fuzzy set, the essence of type-n fuzzy sets are discovered. We first proved the consistency between interval estimation and nested sets representation theorem of type-1 fuzzy sets, which guarantees the rationality of our methods in this paper. Then, based on the fact that a type-1 fuzzy set can be constructed by a collection of intervals, an algorithm for constructing general type-2 fuzzy sets based on the collections of interval type-2 fuzzy sets was proposed. Furthermore, a new concept, interval type-2 fuzzy sets with uncertain bounds, was defined. And the ways for constructing general type-2 fuzzy sets by give interval type-2 fuzzy sets with uncertain bounds and constructing interval type-2 fuzzy sets with uncertain bounds by given general type-2 fuzzy sets were proposed. Finally, from the viewpoint of uncertain bounds, the essence of type-n fuzzy sets was discussed.

In this work, eight antennas are selected with same resonance amplitude but different phase sampled from 0 to 2π with equal step, thus one can use these antennas to from arbitrary elements based on the theory of diffractive optics.

Sensitivity analysis is used to estimate the effective variable (set) in a system. It plays an important role in model simplification, quality assurance of models and codes, identification of crucial regions in the parameter space and so on. Many interesting results have been obtained in the sensitivity analysis of traditional feed forward neural network. But the related work hasn't been done in fuzzy neural network (FNN). In this paper, we choose fuzzy inference neural network (FINN), which has the advantages of representing the uncertain information and higher approximation capability, to study the sensitivity analysis of FNN. We firstly propose spFINN which is the simplification of a FINN introduced by Takatoshi Nishina and present a corresponding training algorithm. The simplification can make the sensitivity analysis process easier without reducing the approximation capability. Then a procedure called FINNSI is proposed to evaluate the sensitivity of input variable of spFINN. As the kernel idea of FINNSI, the way of separation and integration of the sensitivity of neurons is introduced because the sensitivity of neurons is transferred from the last layer (output layer) to the first layer (input layer). The procedure in FINNSI can be easily generalized to the sensitivity analysis of other fuzzy systems or networks. Finally, we discuss the application of FINNSI in band selection. The sensitive bands can be found to help the subsequent steps of spectral data processing such as object recognition and mineral classification.

In face recognition multi-scales Gabor features contain a large amount of redundant information and the value of the features are different from each other. The paper proposes multi-scales Gabor uncertainty weighted fusion algorithm in face recognition, multi-scales Gabor features were used to show facial features firstly, and then weights and uncertainties were calculated to get the last facial feature by multi-scales uncertainty weighted fusion algorithm. At the same time 2DPCA method was exploited to construct the linear subspace and Euclidean distance based classifier was adopted for classification. With the ORL database, the experimental results indicated that compared with the use of 2DPCA, traditional wavelet and Gabor wavelet feature extraction method, the proposed method in this paper had obtained better recognition rate.

In the fierce market competition and gradually standardized policy environment, it is imperative to improve corporate governance structure and establish modern enterprise system for family enterprises. Independent director system is a good solution to control the risk of the family listed company. Independent directors are able to give constructive advice based on their expertise and independence judgment to the company development, so as to have a significant and positive impact on enterprise value. This paper focuses on the study on the relationship between independent director and family business, elaborating further on how to create enterprise value by family enterprise's independent directors.

Except for over-fitting, excessive generalization should lead to high error rate of the learnt rule set, which is seldom discussed by literatures. When excessive generalization is occurred, the rule set will give multiple classification for a particular instance. The errors caused by generalization actually result in the increased inner conflict of the generalized rule set. In this paper, the inner conflict of rule set is defined based on the expanded knowledge of rules and a novel algorithm named RES(reduced error specialization) is proposed for the error rate reduction of rule sets. The best merit of RES is that it can eliminate the inner conflict of a rule set completely while the unknown knowledge of the rule set is unchanged. This fact will guarantee the error rate of the rule set on every test data will be determinedly reduced.

The position and attitude identification of rover is very important for the lunar exploration, which is the precondition of path planning and trajectory tracking control of rover. It is a effective method by visual servo tracking for rover. The motion tracking model based on Kalman filter is proposed by visual feedback signal. The equations are derived for attitude identification based on four rhomboid point configurations. The conclusion by simulation and experiments prove the feasibility of trajectory tracking and attitude identification for rover based on computer vision. The research results in the paper provide the theoretical basis for the position and attitude identification of rover based on stereo vision and has some reference values in practical application.

The two-tier femtocell network is comprised of a central macrocell underlaid with shorter range femtocell hotspots. Due to the universal frequency reuse, this kind of new system architecture brings about urgent problems of the interference management and the resource allocation. Motivated by these problems, the following contributions are made in this paper: 1) a novel joint power and subchannel allocation problem for Orthogonal Frequency Division Multiple Access (OFDMA) downlink based femtocells is formulated on the premise of minimizing Femto BSs' radiating interference; 2) a pseudo-handover based scheduling information exchange method is proposed to avoid the collision interference; 3) an iterative scheme of subchannel allocation and power control is proposed to solve the formulated problem, which is an NP-complete problem. Through simulations and comparisons with three other schemes, the proposed scheme shows better performance in reducing interference and the Femto BS's transmit power, and improving the spectrum efficiency.

The expression of lexical semantics is the crucial factor for natural semantic processing. This paper proposes a new theoretical model for constructing a lexical semantic knowledge base. According to this theory, semantic genes are the carriers of lexical meanings. They can be inherited from hypernyms to hyponyms, and during the inheritance they may be mutated. By heredity, recombination and variation of these semantic genes, new words are produces constantly. According to this rule the whole lexical semantics of a language can be described systematically.

One key step in text mining is the categorization of texts, i.e., to put texts of the same or similar contents into one group so as to distinguish texts of different contents. However, traditional word-frequency-based statistical approaches, such as VSM model, failed to reflect the complicated meaning in texts. This paper ushers in domain ontology and constructs new conceptual vector space model in the pre-processing stage of text clustering, substituting the initial matrix (lexicon-text matrix) in the latent semantic analysis with concept-text matrix. In the clustering analysis stage, this model adopts semantic similarity, partially overcoming the difficulty in accurately and effectively evaluating the degree of similarity of text due to simply taking into account the frequency of words and/or phrases in the text. Experimental results indicate that this method is helpful in improving the result of text clustering.

Considering the time delay and uncertainly in the network system, the GPC based on fuzzy compensation is applied in the NCS to reduce model mismatch and noise disturbance. Simulation results show the validity of the proposed control method.