模式识别与人工智能
Home      About Journal      Editorial Board      Instructions      Ethics Statement      Contact Us                   中文
Pattern Recognition and Artificial Intelligence
22 Judgement and Disposal of Academic Misconduct Article
22 Copyright Transfer Agreement
22 Proof of Confidentiality
22 Requirements for Electronic Version
More....
22 Chinese Association of Automation
22 National ResearchCenter for Intelligent Computing System
22 Institute of Intelligent Machines,Chinese Academy of Sciences
More....
 
 
2014 Vol.27 Issue.6, Published 2014-06-30
Papers and Reports    Researches and Applications   
   
Papers and Reports
481 Covering Matroid and Its Graphical Representation
LI Qing-Yin, LIN Zi-Qiong, ZHU William
Matroid theory has a powerful axiomatic system, which lays a solid foundation for the combination of matroids and other theories. A matroidal structure is constructed by a covering of a universe, and the graphical representation of the matroid is studied. Using the indiscernible neighborhoods, the partition of a universe is induced by a covering of the universe. Through the partition, the matroidal structure of the covering is constructed. The set of all circuits of the matroid is represented by the covering. Finally, the matroid is proved to be a graphic matroid.
2014 Vol. 27 (6): 481-486 [Abstract] ( 438 ) [HTML 1KB] [ PDF 318KB] ( 814 )
487 Extensive-Domain-Search Robust Ant Colony Algorithm for Continuous Function Optimization
CHEN Zhi-Ming, CHEN Zhi-Xiang
A concept of extensive-domain search is proposed to make ant colony optimization for continuous function overcome the sensitivity to initial domains of independent variables. By adding new elements to the gridding method, such as self-adaptive domain adjustment, self-adaptive ant size, self-adaptive pheromone increment and self-adaptive domain division, extensive-domain-search ant colony optimization (EDS-ACO) is put forward. Thus, the optimal solution can be found by EDS-ACO through an extensive search in the whole range of real numbers. Experiments show that EDS-ACO has the robustness since it can obtain the correct results in the case of initial domains without the optimal solution. The variation of initial domains has a small influence on convergence speed and computational accuracy of EDS-ACO.
2014 Vol. 27 (6): 487-495 [Abstract] ( 416 ) [HTML 1KB] [ PDF 688KB] ( 690 )
496 Feature Extraction Method of sEMG Based on Auto Permutation Entropy
XIE Ping, WEI Xiu-Li, DU Yi-Hao, CHEN Xiao-Ling
Due to the nonstationarity and the nonlinearity of the surface electromyogram (sEMG), a method based on the permutation entropy and auto mutual information is proposed to quantitatively describe the internal dynamic features of the sEMG and realize the description of nonlinear characteristics under different motion states. Experiments are carried out to acquire the sEMG data of elbow joint at different bending angles. The auto permutation entropies of the signals are calculated and used as the inputs of support vector machines to identify different motion states. The validity of the proposed method is verified by the comparative analysis of auto permutation entropy and other indexes describing the sEMG features.
2014 Vol. 27 (6): 496-501 [Abstract] ( 623 ) [HTML 1KB] [ PDF 693KB] ( 991 )
502 Rare Category Detection Algorithm Based on Cluster Separability
YAN Xuan-Hui, GUO Gong-De
The rare category mining, which is an important research field in data mining, is widely applied. Aiming at the defects of the traditional rare category recognition methods, an rare category detection algorithm based on cluster separability(RDACS), is proposed based on the combination of density difference and inter-cluster separability criterion for rare category mining. An active-learning scenario is used to detect rare category. The similarity of feature weight is applied to the separability of rare category cluster and its surrounding samples. The experimental results on UCI public datasets and KDD99 datasets show that compared with the existing similar algorithms, the RDACS algorithm has an advantage in the number of inquiries, which can significantly improve the efficiency and reduce human errors. RDACS is complementary to the existing rare category recognition methods.
2014 Vol. 27 (6): 502-508 [Abstract] ( 378 ) [HTML 1KB] [ PDF 382KB] ( 713 )
509 A Method for Image Classification Based on Polyharmonic Random Weights Networks and Curvelet Transform
ZHAO Jian-Wei, ZHOU Zheng-Hua, CAO Fei-Long
Image classification is one of the most important and basic problems in image processing, and designing an effective feature extraction method and a fast classifier with a high recognition rate are two key points in image classification. Polyharmonic random weights networks (P-RWNs) are proposed based on the random weights networks (RWNs) and the advantage of polynomial that it can approximate the part with small variation effectively. Based on the proposed P-RWNs, a method for image classification is presented by integrating fast discrete curvelet transform (FDCT) and discriminative locality alignment (DLA). In the proposed method, FDCT is used to extract features from images, then the dimensionalities of these features are reduced by DLA before the features are input to the proposed P-RWNs classifier for recognition. Experimental results show that the proposed image classification method achieves higher recognition rate and recognition speed.
2014 Vol. 27 (6): 509-516 [Abstract] ( 403 ) [HTML 1KB] [ PDF 537KB] ( 657 )
Researches and Applications
517 Solar Cells Surface Defects Detection Based on Deep Learning
WANG Xian-Bao, LI Jie, YAO Ming-Hai, HE Wen-Xiu, QIAN Yun-Tao
Defects of solar cells are detected mainly by manual operation, and they are difficult to be detected automatically by traditional charge-coupled device(CCD) imaging system. As a training multi-layer neural network, deep learning draws great attention due to its strong ability to extract features from input sample data. A method for solar cells surface defects detection based on deep learning is proposed. Firstly, deep belief networks(DBN) are established and trained according to the sample features to obtain the initial weights of the networks. Then, the traditional BP algorithm is conducted to fine-tune the network parameters to get the mapping relationship between the training samples and the defect-free template. Finally, the defects of testing samples are detected by the contrast between the reconstruction image and the defect image. Experimental results show that DBN perfectly establishes the mapping relationship, and it can quickly detect defects with a high accuracy.
2014 Vol. 27 (6): 517-523 [Abstract] ( 1093 ) [HTML 1KB] [ PDF 646KB] ( 2644 )
524 Interestingness Rule Mining Algorithm Based on Information Entropy
JIN Zhou, WANG Ru-Jing
With the development of data collection and storage techniques, excessive and unorderly rules are generated by traditional association rule mining, which can not meet interest of users. To solve this problem, an interestingness measure of association rules based on information entropy is proposed to mine interestingness association rules. Correlation analysis for categorical variables is adopted to eliminate false and erroneous rules from the primitive set, and a framework for evaluating the interestingness degree of rules based on information entropy is proposed. Since the method does not depend on the prior knowledge of users, it can represent the information hidden in the data accurately. Simulation results on both real and synthetic datasets show that the proposed algorithm performs better than the traditional algorithms, and it discovers interestingness rules from large database efficiently.
2014 Vol. 27 (6): 524-532 [Abstract] ( 380 ) [HTML 1KB] [ PDF 595KB] ( 940 )
533 Particle Swarm Optimization Algorithm with Double-Flight Modes
LI Jing-Yang, WANG Yong, LI Chun-Lei
An optimization algorithm is proposed based on the simulation of flight modes of the real birds, namely particle swarm optimization algorithm with double-flight modes(DMPSO). Particles can use maneuver flight-mode or non-maneuver flight-mode to fly during searching. Each particle chooses its flight-mode according to the feedback of the swarm information and its own state in the search. To test the performance of DMPSO, experiments are carried out on some typical complex high-dimensional optimization problems. The experimental results show that the DMPSO avoids the premature convergence problems and it is effective when solving complex high dimensional optimization problems.
2014 Vol. 27 (6): 533-539 [Abstract] ( 424 ) [HTML 1KB] [ PDF 508KB] ( 631 )
540 Bayesian Network Structure Learning Based on Hybrid Differential Evolution and Bee Colony Algorithm
GUO Tong, LIN Feng
Bayesian network structure learning is the core of Bayesian network theory and the current algorithms of learning Bayesian network structures are always inefficient. A method of learning Bayesian network structure based on hybrid differential evolution and bee colony algorithm is proposed. The maximum weight spanning tree is used to generate the candidate networks, and then the differential evolution algorithm is used to optimize the initial populations. In the process of using the differential evolution algorithm, the bee colony algorithm is introduced into variation stage and optimizing cross stage, and better candidates are selected by applying cloud-based adaptive theory to the choose stage. Simulation results on classic Bayesian network show that the proposed algorithm has a strong searching ability in Bayesian network structure learning.
2014 Vol. 27 (6): 540-545 [Abstract] ( 518 ) [HTML 1KB] [ PDF 450KB] ( 889 )
546 Fast Harmonic and Sparse Image Decomposition Model and Its Application
ZHENG Cheng-Yong
An image decomposition model, harmonic and sparse image decomposition (HSID), is firstly put forward to decompose an image into a harmonic component and a sparse component. Then, based on augmented Lagrangian alternating direction method (ALADM), an algorithm, namely HSID_ALADM, is presented to solve HSID. The main computational load of each iteration in HSID_ALADM is computing fast Fourier transform (FFT), which makes HSID_ALADM fast. HSID_ALADM can be used to decompose an infrared image with small targets into a harmonic component and a sparse component. The harmonic component is considered as the modeling of the background, and the sparse component as the small target component. By searching for the maximum local energy regions in the sparse component, the infrared targets in the infrared image can be easily and accurately located. Experimental results of small infrared target detection for real infrared images and image completion and inpainting show good performance of HSID_ALAD.
2014 Vol. 27 (6): 546-553 [Abstract] ( 383 ) [HTML 1KB] [ PDF 1359KB] ( 725 )
554 An English Fricative Detection Method Based on Energy Spectrum Entropy
LI Li-Yong, ZHANG Lian-Hai
According to the spectrum characteristics of fricatives, a fricative detection method based on the energy spectrum entropy is proposed. Firstly, phone boundaries are detected based on spectrum of different phonemes. Then, each spectrum entropy of speech segments is computed and the segments whose entropy exceeds the threshold are selected as candidates. Finally, post processing is conducted to remove the insertion errors according to parameters of segment length and the sudden changing of energy at segment starts and ends. The experimental results show that the accuracy of the proposed method is up to 96.9% in clean circumstance when the tolerance is 20 ms.
2014 Vol. 27 (6): 554-560 [Abstract] ( 375 ) [HTML 1KB] [ PDF 1084KB] ( 689 )
561 An Adaptive Thresholding Image Denoising Method Based on Morphological Component Analysis and Contourlet Transform
JI Jian, XU Shuang-Xing, LI Xiao
Aiming at the noise image with rich texture and edge feature, an adaptive thresholding image denoising method based on morphological component analysis (MCA) and contourlet transform is proposed. Firstly, MCA method is introduced to separate the image into the low frequency part and the high frequency part. Then, an adaptive thresholding processing method is designed. Finally, according to the characteristics of noise distribution, the threshold estimation and contourlet transform are used in the low frequency part and the high frequency part to effectively remove the noise from the noisy image. The experimental results on noise images illustrate that the proposed method reserves better textures and edges of the image, and its denoising performance is better than that of the mean filter, themedian filter, the wavelet multilevel threshold denoising and the contourlet multilevel threshold denoising.
2014 Vol. 27 (6): 561-568 [Abstract] ( 348 ) [HTML 1KB] [ PDF 714KB] ( 1281 )
569 ε-Pareto Dominance Strategy Based on Angle Preference in MOEA
ZHENG Jin-Hua, LAI Nian, GUO Guan-Qi

By using reference points and angle values, decision maker's preferences are introduced into ε-multi-objective evolutionary algorithm(ε-MOEA). The objective space is divided into preference area and non-preference area by the preferences. Moreover, an angle preference based ε-Pareto dominance strategy is presented. It establishes a strict partial order relation to distinguish the preference solutions and non-preference solutions among non-dominated solutions. To demonstrate the effectiveness of the proposed strategy, it is integrated into ε-MOEA,and thus ε-Pareto dominance strategy based on angle preference in MOEA(AP-ε-MOEA) is put forward . The comparative experiments of AP-ε-MOEA, g-dominance and r-dominance show that AP-ε-MOEA can converge to Pareto optimal front with a higher speed and meanwhile meet the decision maker′s preferences.
2014 Vol. 27 (6): 569-576 [Abstract] ( 525 ) [HTML 1KB] [ PDF 711KB] ( 691 )
模式识别与人工智能
 

Supervised by
China Association for Science and Technology
Sponsored by
Chinese Association of Automation
NationalResearchCenter for Intelligent Computing System
Institute of Intelligent Machines, Chinese Academy of Sciences
Published by
Science Press
 
Copyright © 2010 Editorial Office of Pattern Recognition and Artificial Intelligence
Address: No.350 Shushanhu Road, Hefei, Anhui Province, P.R. China Tel: 0551-65591176 Fax:0551-65591176 Email: bjb@iim.ac.cn
Supported by Beijing Magtech 　Email:support@magtech.com.cn