Home      About Journal      Editorial Board      Instructions      Ethics Statement      Contact Us                   中文

Judgement and Disposal of Academic Misconduct Article Copyright Transfer Agreement Proof of Confidentiality Requirements for Electronic Version More.... Chinese Association of Automation National ResearchCenter for Intelligent Computing System Institute of Intelligent Machines,Chinese Academy of Sciences More....

2013 Vol.26 Issue.2, Published 2013-02-28 Orignal Article        Orignal Article 129 Kernel Sparse Representation Classification and Multi-Scale Block Rotation-Extension Based Robust Image Recognition Method KUANG Jin-Jun,XIONG Qing-Yu,CHAI Yi The random permutations and combinations of local images in image recognition tasks are complex problems. In this paper,an algorithm based on kernel sparse representation classification and multi-scale block rotation-extension (KSRC-MSBRE) is proposed to solve these problems. Firstly,the multi-scale grids are used to segment the training image,and the rotation-extended methods are applied to create a dictionary which adapts to the random permutations and the combinations of local images in test sets. To enhance the sparsity of the dictionary and improve the efficiency of the system,a new strategy is proposed to reduce the dimensions of the dictionary. Then,a kernel random coordinate descent method is proposed to solve the convex optimization problem in the KSRC. The experimental results show the proposed method has robust performance when dealing with the random permutations and the combinations of local images,and it outperforms other classical image recognition methods. 2013 Vol. 26 (2): 129-135 [Abstract] ( 323 ) [HTML 0KB] [ PDF 623KB] ( 753 ) 136 A Margin Based Greedy Ensemble Pruning Method GUO Hua Ping,FAN Ming,ZHI Wei Mei Theoretical and experimental results indicate that for the ensemble classifiers with the same training error the one with higher margin distribution on training examples has better generalization performance. Therefore,the concept of margins of examples is introduced to ensemble pruning and it is employed to supervise the design of ensemble pruning methods. Based on the margins,a new metric called margin based metric (MBM) is designed to evaluate the importance of a classifier to an ensemble and an example set,and then a greedy ensemble pruning method called MBM based ensemble selection is proposed to reduce the ensemble size and improve its accuracy. The experimental results on 30 UCI datasets show that compared with other state of the art greedy ensemble pruning methods,the ensembles selected by the proposed method have better performance. 2013 Vol. 26 (2): 136-143 [Abstract] ( 531 ) [HTML 0KB] [ PDF 454KB] ( 614 ) 144 Memory-Based Cognitive Modeling for Visual Information Processing WANG Yan-Jiang,QI Yu-Juan Inspired by the way in which humans perceive the environment,a memory-based cognitive model for visual information processing is proposed to imitate some cognitive functions of human brain. The proposed model includes five components: information granule,memory spaces cognitive behaviors,rules for manipulating information among memory spaces,and decision-making processes. According to the three-stage memory model of human brain,three memory spaces are defined to store the current,temporal and permanent visual information respectively,i.e. ultra short-term memory space (USTMS),short-term memory space (STMS) and long-term memory space (LTMS). The past scenes can be remembered or forgotten by the proposed model,and thus the model can adapt to the variation of the scene. The proposed model is applied to two hot issues in computer vision: background modeling and object tracking. Experimental results show that the proposed model can deal with scenes with sudden background,object appearance changing and heavy object occlusions under complex background. 2013 Vol. 26 (2): 144-150 [Abstract] ( 647 ) [HTML 0KB] [ PDF 576KB] ( 1036 ) 151 Image Object Segmentation Algorithmby Combining Depth Discontinuities and Color Information PI Zhi-Ming,WANG Zeng-Fu The image segmentation using stereo image pairs is discussed. An image segmentation algorithm combining depth discontinuities and color information is proposed. Mean-shift segmentation algorithm is applied to the over-segmentation of the image,and meanwhile the dense depth map of the image pairs can be calculated by using stereo vision algorithm. Then,through combining color image over-segmentation and depth discontinuities,multiple seed regions for accurate segmentation are selected along the depth discontinuities. By using graph cut algorithm,unlabeled regions are assigned with seed regions′ labels. Next,the neighbor regions with different labels but without discontinuous depth boundary between them are merged together as well. Compared with the traditional feature clustering image segmentation algorithms,the proposed algorithm overcomes the problems of over-segmentation and under-segmentation,and semantic object segmentation results can be achieved. Experimental results show the validity of the proposed algorithm. 2013 Vol. 26 (2): 151-158 [Abstract] ( 811 ) [HTML 0KB] [ PDF 652KB] ( 1733 ) 159 A Fast Learning Algorithm Based on Minimum Enclosing Ballfor Large Domain Adaptation XU Min,WANG Shi-Tong,GU Xin,YU Lin The data fields detected from different times,places or devices are not always complete even if they come from the same data resource. To solve the problem of effectively transferring the knowledge between the two fields,the theorem is proposed that the difference between two probability distributions from two domains can be expressed by the center of each domain′s minimum enclosing ball and its up limit has nothing to do with the radius. Based on the theorem,a fast center calibration domain adaptive algorithm,center calibration-core sets support vector data description (CC-CSVDD),is proposed for large domain adaptation by modifying the original support vector domain description (SVDD) algorithm. The validity of the proposed algorithm is experimentally verified on the artificial datasets and the real KDD CUP-99 datasets. Experimental results show that the proposed algorithm has good performance. 2013 Vol. 26 (2): 159-168 [Abstract] ( 592 ) [HTML 0KB] [ PDF 1108KB] ( 627 ) 169 Kernel Coupled Metric Learning and Its Application to Gait Recognition WANG Ke Jun,YAN Tao To solve the problems of the disaster of dimensionality and the shortage on describing the nonlinear model that the linear coupled metric learning has when solving practical problems,the kernel coupled metric learning is proposed by introducing kernel method. Firstly,the nonlinear transformations are used to map the data from different sets into a high dimensional coupled space to make the elements of two sets with correlation as close as possible to each other after the projection. Then,the traditional kernel method is used in the public coupled space. The proposed method is applied to gait recognition to solve the match problem of different sets. Experiments and analysis are made on the CASIA(B) gait database,and the experimental results show that the proposed method has satisfactory recognition results. 2013 Vol. 26 (2): 169-175 [Abstract] ( 614 ) [HTML 0KB] [ PDF 365KB] ( 623 ) 176 Flickr Group Recommendation Model by Integrating Tags in Group and Users′ Contacts BAO Hong-Yun,LI Qiu-Dan,GAO Heng,ZHENG Nan In Flickr,one of the most popular photo service websites,groups are superior for photos′ propagation and they can gather photos of similar themes,which brings convenience to users′ easy browsing. Therefore,many researchers are studying how to help users to find out which groups they may be interested in. In this paper,a probabilistic matrix factorization (PMF)based model is proposed for Flickr group recommendation by employing the information of users′ contacts and tags in group. The complexity analysis indicates that the proposed model is efficient and it can be applied to large datasets. The experimental results on a Flickr dataset show the effectiveness of the proposed model. Finally,a Flickr group recommendation system is developed based on the proposed model. 2013 Vol. 26 (2): 176-181 [Abstract] ( 498 ) [HTML 0KB] [ PDF 615KB] ( 768 ) 182 Identification Method Based on Sound Wave of Teeth Collision ZHU Bing-Cheng,WU Le-Nan,WANG Wei An identification method based on sound wave of teeth collision is presented to overcome the problems in voiceprint identification,including speech segmentation,Signal noise ratio(SNR) enhancement and feature extraction. Characteristic for sound wave of teeth collision is studied and a mathematic model derived from a second-order differential equation is proposed to describe the teeth vibration process. Based on the model,an identification algorithm is presented. Experimental results show that the sound wave of teeth collision works better in identification than voiceprint signal due to its stability and lower processing complexity. Besides,the presented algorithm reduces the amount of training samples and the computational complexity,compared with the support vector machine and the nearest neighbor algorithm. 2013 Vol. 26 (2): 182-188 [Abstract] ( 598 ) [HTML 0KB] [ PDF 547KB] ( 749 ) 189 A Semantic Similarity Weighted Query Term Proximity Framework for Information Retrieval QIAO Ya-Nan,LIU Yue-Hu,QI Yong Traditional proximity retrieval models treat query terms equally and they do not distinguish the proximities between query terms. Thus,the parallel concept effect is caused,and the performance of many query term proximity based information retrieval models is affected. A semantic similarity weighted query term proximity framework is proposed.The statistics of query term proximity are weighted in this framework by the semantic similarities between query terms,and then the in-depth information needs can be concluded and mined.Experimental results show that compared with traditional proximity retrieval models,the proposed framework greatly improves the performance of traditional proximity retrieval models and avoids the parallel concept effect efficiently for short queries. 2013 Vol. 26 (2): 189-194 [Abstract] ( 588 ) [HTML 0KB] [ PDF 376KB] ( 1014 ) 195 Chaotic Ant Based Collaborative Optimization Algorithm in Distributed System WEI Zhen,GE Fang-Zhen ,LU Yang,WANG Qiang,LI Li-Xiang To solve the optimization problem in complex distributed system (CDS),a collaborative optimization method based on chaotic ant swarm in CDS is presented. The basic dynamic characteristics of complex distributed systems are analyzed under the guide of system theory,and a model of collaborative optimization of CDS is proposed. Thus,a collaborative optimization in CDS is established based on the idea of chaotic ant swarm (CAS),called CAS based collaborative optimization (CAS-CO). The locality-based task allocation in complex networked multi-agent system is resolved by CAS-CO,and the comparison results of the proposed algorithm and the existing ones show that the CAS-CO algorithm is feasible and effective,and the proposed model is correct and the autonomy of an agent is of importance for the design and modeling of CDS. 2013 Vol. 26 (2): 195-204 [Abstract] ( 546 ) [HTML 0KB] [ PDF 624KB] ( 719 ) 205 Images Segmentation Based on Genetic_Kernel Fuzzy C-Means Clustering Algorithm JIN Lu,FU Meng-Yin Aiming at the characteristics of infrared images and the sensitivity of fuzzy clustering algorithm to the noise and the initial clustering center,a genetic kernel fuzzy C-Means clustering algorithm(G_KFCM) is presented. The gray values of the infrared images are clustered globally. Then the optimal clustering center and the membership matrix are calculated by the G_KFCM. The image segmentation is performed according to the clustering result and the maximum membership principle. The experimental results show G_KFCM is effective to the infrared images respectively including Gaussian noise,simple or complex background. 2013 Vol. 26 (2): 205-210 [Abstract] ( 506 ) [HTML 0KB] [ PDF 720KB] ( 570 ) 211 Principal Component Analysis Based on L1-Norm Maximization with Lp-Norm Constraints LIANG Zhi-Zheng,LI Yong,XIA Shi-Xiong,ZHOU Yong Aiming at the sensitivity of principal component analysis in dealing with the contaminated data and the property that its projection vectors are not sparse,a robust principal component analysis optimization model is proposed. The objective function of the proposed model adopts L1 norm and projective vectors are constrained by Lp norm. An iterative algorithm is used to solve the proposed model and the theoretical analysis shows that the algorithm can obtain the locally optimal solution. In addition,the kernel version is made by embedding kernel functions into the model. The experiments on UCI datasets and face datasets are performed to demonstrate the feasibility and effectiveness of the proposed method. 2013 Vol. 26 (2): 211-218 [Abstract] ( 706 ) [HTML 0KB] [ PDF 486KB] ( 905 ) 219 3D Facial Depth Map Recognition in Different Poses with Surface Contour Feature YE Chang-Ming,JIANG Jian-Guo,ZHAN Shu,ANDO Shigeru Three-dimensional face recognition has drawn more and more attention,for it overcomes the shortcomings of two-dimensional face recognition technology that two-dimensional face recognition is susceptible to the influence of light,expression changes and pose variations. A face recognition method,Fourier descriptor and contour (FDAC),is proposed in this paper. It is based on the depth maps by the three-dimensional facial imaging system in different poses. Firstly,depth maps are corrected under the guidance of thedifferential geometry theory,and the human face features are described by the contours. Secondly,Fourier descriptor is employed to extract the facial features.Finally,these extracted features are used in the face recognition process. Experimental results show that FDAC has good recognition accuracy and it performs better in time cost compared with Eigenface method. 2013 Vol. 26 (2): 219-224 [Abstract] ( 585 ) [HTML 0KB] [ PDF 465KB] ( 979 )

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