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....

2010 Vol.23 Issue.5, Published 2010-10-30 Orignal Article        Orignal Article 593 A Fast Warping Visual Homing Algorithm Based on Random Search ZHENG Zhong,WANG Zeng-Fu An improved warping algorithm is proposed for visual robot homing. The original warping algorithm has to search the whole parameter space, which results in the problem of high computational cost and limited application scope. To solve this problem, a gradient descent method based on randomly selected initial points is implemented, which significantly decreases the tested parameters and improves the algorithm efficiency. Experimental results of several real scenes show that the proposed algorithm significantly improves the computational efficiency by one order of magnitude with the robustness. 2010 Vol. 23 (5): 593-600 [Abstract] ( 937 ) [HTML 1KB] [ PDF 528KB] ( 561 ) 601 Individual Difference of Emotional Spontaneous Transferring Process in Human-Robot Interaction WANG Wei,WANG Zhi-Liang,ZHENG Si-Yi,GU Xue-Jing Harmonious human-robot interaction needs emotional model. In Markov model of emotional spontaneous transferring, aiming at the emotional distinguishing problem whether parameters adjusting have any effects on individual emotional difference and how much effect it brings in, researching method of individual artificial emotional difference is proposed based on metric multidimensional scaling. Scalar product matrix is calculated by dissimilarity matrix. Subsequently, individual attribute reconstructing matrix is got using principal components factor analysis to display individual emotional difference in low dimension space. Parameters in model are adjusted instructionally by means of experimental results. In addition, some of the results are carried out with mathematical verification. 2010 Vol. 23 (5): 601-604 [Abstract] ( 276 ) [HTML 1KB] [ PDF 301KB] ( 585 ) 606 Uncertainty Measure Rule Sets of Rough Sets HU Jun,WANG Guo-Yin Since some uncertainty measures of rough sets are unreasonable under some circumstances, a basic rule set of uncertainty measure of rough set is proposed from the perspective of intuition. All the uncertainty measures except the quadratic fuzziness satisfy the basic rule set. The uncertainty measures satisfying the basic rule set still have unreasonability, and thus an extended rule set is further developed. The fuzzy entropy and revised fuzziness are the uncertainty measures satisfying the extended rule set, while the roughness, rough entropy and linear fuzziness are not. The results provide theoretical basis of the reasonability or unreasonability for the existing uncertainty measures, and it is a foundation for designing new uncertainty measures. 2010 Vol. 23 (5): 606-615 [Abstract] ( 262 ) [HTML 1KB] [ PDF 571KB] ( 659 ) 616 Selective Visual Attention Model Based on Pulsed Cosine Transform YU Ying,WANG Bin,ZHANG Li-Ming A visual attention model based on pulsed cosine transform is proposed, which mimics the generating mechanism of bottom-up visual attention. Due to its simple architecture and high computational speed, the proposed model can be used in real-time systems. The visual salience of the model is represented in binary codes, which agrees with the firing pattern of neurons in the human brain. The motion salience is generated by these binary codes as well. Moreover, the model can be extended to Hebbian-based neural networks. Experimental results show that the proposed model has better performance in human fixation prediction than other state-of-the-art models of visual attention. 2010 Vol. 23 (5): 616-623 [Abstract] ( 612 ) [HTML 1KB] [ PDF 583KB] ( 546 ) 624 Metasynthetic Engineering with Knowledge Conversion Process for Complex Problem Solving ZHANG Guang-Jun,DAI Ru-Wei Complex problem solving is a knowledge intensive task, hence it is important to understand the related knowledge conversion process in metasynthetic engineering. According to the complex features of the problem, a knowledge conversion process model is proposed. In the proposed model, knowledge is created through the conversion of tacit and explicit knowledge, and the knowledge creation process is a dynamic spiral process including cognition, externalization, integration and internalization. Based on the knowledge model, metasynthetic engineering is enhanced to enrich the problem solving process. An example of complex disaster is provided to demonstrate the application of metasynthetic engineering with knowledge management on solving complex problem. 2010 Vol. 23 (5): 624-629 [Abstract] ( 288 ) [HTML 1KB] [ PDF 476KB] ( 534 ) 630 Attribute Reduction Algorithm Based on Common Discernibility Degree TENG Shu-Hua, ZAN De-Cai,SUN Ji-Xiang,TAN Zhi-Guo From the point of knowledge classifications ability, the definition of common discernibility degree and the corresponding properties are introduced. By utilizing common discernibility degree to depict the relative importance of attribute in information system, a heuristic reduced algorithm based on information viewpoint is proposed and proved. It can be directly applied to both complete and incomplete information systems to reduce attributes without pretreatment. The approach ensures the relatively high reduction rate and simultaneously makes the worst time complexity in complete information system fall to 公式 Finally, results of numerical experiments are used to illustrate the high efficiency of the algorithm in incomplete and complete information systems. 2010 Vol. 23 (5): 630-638 [Abstract] ( 564 ) [HTML 1KB] [ PDF 522KB] ( 596 ) 639 Convergence Analysis of Genetic Algorithm Based on Space Mating ZHENG Jin-Hua,L Hui,WU Jun,ZHOU Cong,LI Ke,LI Mi-Qing The genetic algorithm based on space mating (GASM) with space mating operator overcomes the premature convergence effectively, but it lacks theoretical analysis. In this paper, the convergent properties of the genetic algorithm based on space mating are analyzed by homogeneous finite Markov chain. It is proved that the GASM with the elitist mechanism can converge to the global optimum, and the GASM can converge to it with probability one on the condition of no mutation operator. By comparing the experimental results of four test problems, in which three of them are multi-peak complex issues, it is shown that the convergence of GASM is better than that of the genetic algorithm with the elitist mechanism, namely elitist genetic algorithm (EGA) in solving the multi-peak complex problems. The algorithm is compared with the algorithm of fast marriage in honey bees optimization as well. 2010 Vol. 23 (5): 639-645 [Abstract] ( 627 ) [HTML 1KB] [ PDF 444KB] ( 709 ) 646 Automatic Object Tracking Method Based on SAD and UKF-Mean Shift LIU Xian-Ru,CAI Zi-Xing,TANG Jin Aiming at the problems in accurate motion detection and tracking location under complex scene, an automatic object tracking method combined sum of absolute difference (SAD) and Mean shift with unscented Kalman filter (UKF) is proposed. Firstly block matching method based on SAD is used to estimate the displacement between current frame and successive frame. Then the disparity cues are utilized to detect the moving object automatically and build the object model and state-space model for following tracking task. Finally Mean shift with UKF is employed to filter and estimate the state of the object and locate the object in subsequence image frame. The experimental results show that the proposed moving object detection method effectively detects moving objects in scene and acquires the motion information of objects. Compared with the related methods, the proposed tracking strategy based on UKF-Mean shift has better tracking results and time property. 2010 Vol. 23 (5): 646-652 [Abstract] ( 390 ) [HTML 1KB] [ PDF 508KB] ( 732 ) 653 Maximal Gravitation Optimization Algorithm for Function Optimization JIN Lin-Peng,LI Jun-Li,WEI Ping,CHEN Gang A global function optimization algorithm based on Newtons law of universal gravitation is proposed, namely maximal gravitation optimization algorithm (MGOA). The search agents are updated through the processes of gravitational clustering and gravitational elimination, which are two main strategies in MGOA. Four lemmas are provided to describe the mathematical foundation, and the convergence of MGOA is strictly proved. Furthermore, the proposed algorithm is improved. The experimental result shows MGOA has good performance in solving continuous function optimization problems, compared with some well-known heuristic search methods such as Particle Swarm Optimization, Differential Evolution, and Guo Tao algorithm. 2010 Vol. 23 (5): 653-662 [Abstract] ( 630 ) [HTML 1KB] [ PDF 659KB] ( 825 ) 663 Layered Fuzzy Facial Expression Generation Based on Sociality, Emotion and Physiology XUE Yu-Li,MAO Xia,Catalin-Daniel CALEANU,CHANG Qing In the existing models of facial expression generation, few models include comprehensive types of emotions, sources of facial expressions besides emotion and variable facial expressions. Thus a novel model of layered fuzzy facial expression generation is proposed. In this model, social, emotional and physiological factors are considered in different layers producing different effect on facial expression generation, and fuzzy theory is utilized to generate smart and abundant facial expressions. Then, a layered intelligent fuzzy facial expression generation system based on this model is founded and evaluated. The results show the effectiveness of the layered fuzzy facial expression generation is fine in human computer interaction. 2010 Vol. 23 (5): 663-670 [Abstract] ( 293 ) [HTML 1KB] [ PDF 530KB] ( 454 ) 671 Color Image Segmentation Approach by Combining EFD and NCut XU Li-Yan,ZHANG Jie-Yu,SUN Quan-Sen,XIA De-Shen To overcome the over segmentation phenomenon of edgeflow-driven anisotropic diffusion (EFD) and the high computational complexity of normalized cut (NCut) , a color image segmentation algorithm based on EFD and NCut is presented. EFD is applied to the image to get a preliminary result. Then, the segmented regions are taken as nodes to construct a weighted undirected graph G, and the NCut is applied to perform globally optimized clustering. Segmentation results are achieved after proper post-process. The graph structure is based on segmented regions instead of image pixels, and thus the proposed algorithm requires lower computational complexity. In addition, EFD focuses on local detail while NCut captures global property, so this algorithm combines both advantages. Experimental results show that this algorithm can get appropriate segmentation results. 2010 Vol. 23 (5): 671-677 [Abstract] ( 571 ) [HTML 1KB] [ PDF 478KB] ( 593 ) 678 A Batch Constructing Algorithm of Frequent Weighted Concept Lattice WANG Xin-Xin,ZHANG Ji-Fu,ZHANG Su-Lan Concept lattice is an effective tool for knowledge representation and data analysis. Weighted concept lattice is a concept lattice structure which depicts the importance of intention. A batch constructing algorithm of frequent weighted concept lattice is proposed by using the concept of virtual node. Firstly, it is proved that frequent weighted concept lattice is a complete lattice by defining the concept of virtual node, thereby the defect of having no supremum/infimum for some frequent weighted nodes in previous frequent weighted concept lattics proposed by Zhang is avoided. Secondly, frequent node, virtual node and their edges are generated from the bottom to the top. Thus, the time and the storage complexity of constructing the lattice is reduced and the efficiency of batch constructing the frequent weighted concept lattice is improved. Finally, the experimental results validate the correctness and the validity of the proposed algorithm by taking the star spectrum data as the formal contexts. 2010 Vol. 23 (5): 678-685 [Abstract] ( 574 ) [HTML 1KB] [ PDF 562KB] ( 548 ) 686 An Improved 3D Face Reconstruction Method ZHOU Jia-Li,ZHANG Shu-You,WU Min An improved 3D face reconstruction method as well as a binocular stereo vision system based on single camera is proposed. Under the assumption that face is symmetrical, the point cloud is optimized automatically by correction and holes filling. Then, a simplified Candide-3 model is used as initial subdivision controlling mesh, locally refined and levelly fitted. Meanwhile, geodesic mapping technique is applied to normalize different expressions and face databases are built respectively. Experimental results show that the proposed stereo vision system improves the reconstruction accuracy and avoids robust decreasing caused by non synchronous shooting of two cameras. Furthermore, subdivision surfaces used as storage saves space and provides theoretical support for comparison. Considering its low cost, the proposed system is feasible to spread in many fields. 2010 Vol. 23 (5): 686-694 [Abstract] ( 692 ) [HTML 1KB] [ PDF 680KB] ( 586 ) 695 Pattern Recognition of Hand Motions Based on WPT and LVQ LUO Zhi-Zeng,XIONG Jing,LIU Zhi-Hong To recognize hand motions based on the surface electromyography (SEMG), a neural network classifier is put forward by using wavelet packet transform (WPT) and learning vector quantization (LVQ) algorithms. The decomposition coefficients of each node for SEMG are gained by optimal wavelet package decomposition based on entropy criterion. The coefficient energy corresponding to sub-band of each node is calculated. Then the feature vectors via normalization are inputted into LVQ neural networks to realize recognition of hand motions. The experimental results show that four motion patterns including wrist extension, wrist flexion, hand extension and hand grasp can be identified by the classifier using two-channel SEMG with the recognition accuracy up to 96%. Consequently, the classifier is applicable to myoelectric prosthetic hand control of 2 degrees of freedom (DOF) because of its superior recognition capability. 2010 Vol. 23 (5): 695-700 [Abstract] ( 615 ) [HTML 1KB] [ PDF 399KB] ( 712 ) 701 KNN Model Based Incremental Learning Algorithm GUO Gong-De,HUANG Jie,CHEN Li-Fei KNN Model is an improved version of the k-nearest neighbor method. However, KNN Model is a non-incremental learning method, which restricts it from some real applications. A KNN Model based incremental learning method is proposed by introducing level concept for created clusters. It constructs few clusters for new coming data with different levels assignment to adjust and optimize previous generated KNN Model. Experimental results show the effectiveness of the proposed method. 2010 Vol. 23 (5): 701-707 [Abstract] ( 926 ) [HTML 1KB] [ PDF 490KB] ( 1127 ) 708 Multisurface Support Vector Machines via Weight Vector Projection YE Qiao-Lin ,YE Ning,CUI Jing,CHEN Yan-Nan,WU Bo A multisurface support vector machine classifier is proposed called multisurface support vector machines via weight vector projection. It generates two weight vectors by solving two simple eigenvalue problems without consideration of the matrix singularity in it. Unlike the standard classifiers, the solution of the specific hyperplane is not required. According to the decision rule of the proposed approach, a unseen point is assigned to the closest projected mean. The proposed approach obtains comparable computational efficiency compared with proximal support vector machine via generalized eigenvalues (GEPSVM). Moreover, it solves some complex XOR problems as well. The experimental results on artificial and UCI datasets show that the classification performance of the proposed approach outperforms that of GEPSVM. 2010 Vol. 23 (5): 708-714 [Abstract] ( 592 ) [HTML 1KB] [ PDF 374KB] ( 474 ) 715 An Improved Distance Coherence Vector Algorithm for CBIR ZENG Jie-Xian,ZHAO Yong-Gang,FU Xiang An improved distance coherence vector for content-based image retrieval (CBIR) is proposed to improve the algorithm proposed by Sajjanhar et al. The improved algorithm regards centroidal distances vector of average coordinates from the biggest connected coherence pixels as a new feature vector. The new added feature vector is invariable to translation, scaling and rotation. Similarity of images is measured by different similar functions according to different feature vectors. The improved algorithm has better retrieval effect due to the more introduced spatial information. The experimental results indicate that the improved algorithm has high recall and precision. 2010 Vol. 23 (5): 715-719 [Abstract] ( 300 ) [HTML 1KB] [ PDF 335KB] ( 466 ) 720 Concept Lattice Attribute Reduction Based on Intersectional Reducible Equivalence Class LIN Pei-Rong,ZHANG Qi-Sen,LI Jin-Jin The concepts of intersectional reducible equivalence class and intersectional reducible element are introduced. The concept lattice attribute reduction and reduction algorithm based on intersectional reducible elements are studied, and attribute characters of different kinds are obtained. The linked list is used to show the logical structure of formal context, and based on the number of extension objects, the index is built to rapidly judge the validity of the intersection operation on attribute reduction. All unnecessary attributes are found out according to the different roles of attributes to intersection operation. Finally, the concept lattice attribute reduction is achieved. 2010 Vol. 23 (5): 720-726 [Abstract] ( 574 ) [HTML 1KB] [ PDF 386KB] ( 463 ) 727 Adaptive Threshold Video Splicing Algorithm Based on Distance Feature CHEN Xia-Yan,WU Xian-Da An adaptive threshold video splicing algorithm based on distance feature is proposed to solve the problem of video splicing in virtual walkthrough system. The algorithm is used to splice video images captured by virtual walkthrough system.By extracting distance feature, the adaptive threshold sequential similarity detection algorithms (SSDA) it used to search the matching feature in the frame images to be matched. Then, the beginning column of the overlapping part it estimated. Experimental results show the proposed algorithm realizes video splicing better, reduces workload and accelerates the splicing speed. 2010 Vol. 23 (5): 727-730 [Abstract] ( 528 ) [HTML 1KB] [ PDF 309KB] ( 476 ) 731 Improved Method of Maximizing AUC and Its Application to Obstacle Detection HAN Guang,ZHAO Chun-Xia In the obstacle detection, obstacle/non-obstacle samples have characteristics of a large range of overlapping in the feature space and uneven distribution. The traditional training method for the classifier is not competent for dealing with such data. Thus, an improved method of maximizing area under the ROC (AUC) is proposed to train classifier. An alternative function is used as the objective function of optimizing AUC. Meanwhile, the particle swarm optimization is introduced to optimize the AUC objective function, and the particle swarm optimization algorithm is improved by using the Butterworth curves and particles with the low fitness value being mutated. The experimental results show that the proposed method effectively solves the local optimization caused by the gradient descent method. Moreover, the detection performance of the proposed method is improved compared with other existing algorithms, and the algorithm is reliable and efficient. 2010 Vol. 23 (5): 731-737 [Abstract] ( 577 ) [HTML 1KB] [ PDF 533KB] ( 531 ) 738 Robust Approach for 2D Shape-Based Image Retrieval YANG Xu,YANG Xin,TIAN Xue The geometric centre is taken as the origin in the traditional centroid-radii based shape description method, which makes it sensitive to noise, and slight changes in the boundary cause errors in matching. A robust method for 2-D shape description and matching is presented to solve this problem. It uses a polar transformation of the contour points to get the shape descriptor, and the maximum of the generalized Hough transform (GHT) mapping array is taken as the reference point. The experimental results on 3 benchmark test sets show that the proposed approach is invariant to translation, rotation and scaling. Furthermore, it achieves high performance in the retrieval of partially occluded and defect images. 2010 Vol. 23 (5): 738-745 [Abstract] ( 482 ) [HTML 1KB] [ PDF 485KB] ( 552 )

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