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....

2006 Vol.19 Issue.1, Published 2006-02-25 Papers and Reports    Researches and Applications    Surveys and Reviews        Papers and Reports 1 OnLine Signature Verification Based on Shape and Writing Forces FANG Ping, MENG Ming, WU ZhongCheng, GE YunJian, YU Yong To solve the problem that the present signature collecting devices are not able to get the writing forces and signature shape simultaneously or only get the writing pressure, a selfdevised F_Tablet writing tablet is used to collect the twodimension signature shape and threedimension forces during writing. And a novel technique using stroke segmentation, multitemplate and iterative experiment puts forward to verify the signatures with signature shape and writing forces. The validity of this technique is proved by the experimental results based on the signature database constructed with F_Tablet. 2006 Vol. 19 (1): 1-6 [Abstract] ( 263 ) [HTML 1KB] [ PDF 547KB] ( 421 ) 7 Intrinsinc Dimension Estimation Based on LLE TAN Lu, WU Yi, YI DongYun A new algorithm to estimate the intrinsic dimension of data sets is proposed. The method is constructed by approximation and separation, which comes from the topological structure of data set and the distance characteristics of high dimensional space. Thereinto, the topological structure is intruduced by LLE. It discloses the relation between dimension and neighborhood then improves LLE. Experiments show that this method is reasonable and reliable than PCA. 2006 Vol. 19 (1): 7-13 [Abstract] ( 303 ) [HTML 1KB] [ PDF 1605KB] ( 514 ) 14 A Novel Heuristic Search Algorithm for Edge Extraction in Noise Image DONG YinWen, GUO Lei, YAO Jun A novel heuristic search algorithm based on subedge selfreinforce for edge extraction in noise image is proposed in this paper. Firstly, The noise image is filtered by a small scale Gaussian Filter. Then a new Large Template Edge Detector is designed in order to get more accurate leading information, and the corresponding search trajectories are selfreinforced according to this information. Finally, the real edge of noise image is extracted according to the accumulated degree of selfreinforces. The new Large Template Edge Detector has good performance in orientation precision, noise resistance and false edge. Experimental results on image with noise demonstrate better performance of the proposed method, which keeps more image details in extracting real edges of objects, compared with the classical methods, especially Canny Operator. 2006 Vol. 19 (1): 14-19 [Abstract] ( 227 ) [HTML 1KB] [ PDF 889KB] ( 509 ) 20 The Important Schema and Its Properties in Genetic Algorithms LI YunQiang, YU ZhaoPing BuildingBlock Hypothesis is an important theoretical foundation of genetic algorithms. Although there are many practical evidences to support it, it has not been proven. This paper proposes the definition of important schema, discusses the properties of important schema, proves a special example of BuildingBlock Hypothesis, presents the condition of whether there is only one best solution and gets the characteristic of the best schemata of some special function species. 2006 Vol. 19 (1): 20-23 [Abstract] ( 296 ) [HTML 1KB] [ PDF 265KB] ( 635 ) 24 The Property of Rough Invariant Subgroup YU JiaLi, SHU Lan In this paper, the application of rough sets theory on group is discussed. Based on the notions of the congruence relation in a rough group, rough coset and rough invariant subgroup, some properties of the rough invariant subgroup are given and proved. Thus the rough group theory is completed. 2006 Vol. 19 (1): 24-26 [Abstract] ( 232 ) [HTML 1KB] [ PDF 224KB] ( 369 ) 27 An Optimal Algorithm for the Farthest Pair Problem of a Point Set QU JiLin, KOU JiSong, LI MinQiang An optimal algorithm for the farthest pair problem is presented in this paper. Given a set of n points in the plane, this paper proposes a new method to find the antipodal pairs of the convex hull of the point set. Using this method, the farthest pair problem can be solved in O(nlogn) time. 2006 Vol. 19 (1): 27-30 [Abstract] ( 279 ) [HTML 1KB] [ PDF 290KB] ( 973 ) 31 A Method of Bayesian Network Construction Combining Knowledge and Data YANG ShanLin, HU XiaoXuan, MAO XueMin Learning the structure of a Bayesian network from data may be time expensive due to huge search space. Because a Bayesian network contains causal semantics, experts can use their knowledge to confirm cause and effect among variables. In this paper, experts’ opinions are collected and combined using DempsterShafer evidence theory. The network structures without semantics are eliminated, then learning network from data is continued. This method fuses expert knowledge which is used to reduce search space with data to construct a Bayesian network. It avoids the subjective bias of single expert. The experimental results show that this method can improve learning efficiency. 2006 Vol. 19 (1): 31-34 [Abstract] ( 293 ) [HTML 1KB] [ PDF 353KB] ( 536 ) 35 Qualitative Mapping Model from Judgment to Recognition and Fuzzy Artificial Neuron FENG JiaLi It is shown that a judgment of property (proposition) based on the conversion of a quantity of a attribute a(o) x into a property p(x,o) of a(o), whose judging result or the truth value of property varies according to its qualitative criterion [α,β], can be extracted as a dynamic Qualitative Mapping τp(x,[α,β]) whose parameter is criterion [α,β]. Several QM definitions are respectively given whose qualitative criteria are a interval, a vector of intervals and a grid of ndimension parallelepiped. The geometric mean that a weight w acting on the qualitative criterion [α,β], the relation between a qualitative mapping and an artificial neuron, the mechanism and the condition that a qualitative mapping for judging truth value of property converted into a pattern recognition mapping are respectively discussed. Furthermore, the degree function that a quantity of an attribute converted into a quality or a property and a mechanism fuzzized the boundary of qualitative criterion are presented too. And it’s shown that if qualitative mapping is substituted by a fuzzized conversion degree function respectively the judgment and recognition would be fuzzy. And a series of examples to test the validity of qualitative mapping and conversion degree function are given. 2006 Vol. 19 (1): 35-46 [Abstract] ( 342 ) [HTML 1KB] [ PDF 1533KB] ( 458 ) 47 An Improved MultiPattern String Matching Algorithm DAI LiuLing, HUANG HeYan , CHEN ZhaoXiong A new algorithm for matching multiple strings at the same time is suggested. The new algorithm is based on the ideas of QS and SunWu algorithm, named as QMS (Quick Multipattern Searching) algorithm in this paper. QMS uses hashing and PREFIX table to decrease the number of comparisons. During the computation of the shift distance, the character closely after the current window is considered. Because the shift distance is computed with more accurate technique, larger average shift distance is acquired. More characters can be skipped when the text is scanned, so the algorithm becomes very efficient. Tests on an actual corpus show that QMS algorithm is much more efficient than SunWu algorithm under common circumstances. 2006 Vol. 19 (1): 47-51 [Abstract] ( 228 ) [HTML 1KB] [ PDF 439KB] ( 428 ) 52 Dimension Reduction and Similarity Search for Time Series Based on Regression Coefficient HUANG Chao, ZHU YangYong Dimension reduction is always necessary when similarity search is conducted in time series. The privious methods are of high time complexity and unintuitive (such as DFT and DWT), or can’t be used for accurate similarity search (such as PAA). This paper brings forward a new dimension reduction method which is called Piecewise Regression Approximation (PRA) for time series based on regression coefficient. The PRA method is of linear time complexity and not sensitive to independent noises. It is proved the similarity search based on PRA method satisfies the lowerbounding lemma, so it is practical and effective. The experiments conducted on reallife datasets validate our conclusions. 2006 Vol. 19 (1): 52-57 [Abstract] ( 209 ) [HTML 1KB] [ PDF 457KB] ( 412 ) Surveys and Reviews 58 A Survey of Face Recognition LI WuJun, WANG ChongJun, ZHANG Wei, CHEN ShiFu Due to various applications in the areas of pattern recognition, image processing, computer vision, cognitive science etc., face recognition has drawn much attention in recent years. This paper presents an uptodate survey on the history and stateoftheart face recognition research, systematically classifying face recognition methods into several categories. Furthermore, this paper expatiates on the evolution of the recent algorithms which are used to deal with the illumination variation problem and the pose variation problem. In addition, several major issues for further exploration are also pointed out at the end of this paper. 2006 Vol. 19 (1): 58-66 [Abstract] ( 361 ) [HTML 1KB] [ PDF 447KB] ( 2074 ) Researches and Applications 67 A Method for Determining Joint Position of Human Body Based on Silhouette Image FAN Bin, WANG ZengFu A method for articulated pose identification from the silhouette of a human body is presented in this paper. First the method employs an energy function to extract a dummy skeleton of a human body from the silhouette. Then it estimates the position of the joints by the knowledge of anthropotomy and the three rules given by the standard human skeleton model, which can help us judge whether a point on the skeleton is a joint. The experimental results show that the method is effective. It doesn’t require any additional restrictions on the motion and color of a human body. In the meantime the method is robust against the noise in the silhouette, which makes it have good performance in the complicated background. 2006 Vol. 19 (1): 67-72 [Abstract] ( 234 ) [HTML 1KB] [ PDF 528KB] ( 493 ) 73 Immune Ant Colony Algorithm for MultiTask Scheduling Problem ZHONG YiWen, YANG JianGang1 This paper presents an immune ant colony algorithm for multitask scheduling in parallel and distributed systems. Ant colony algorithm is used to evolve a priority list firstly, then the priority list is mapped to a schedule by a greedy strategy. In order to avoid premature stagnation, immune principle is used to preserve the diversity of the population of ants. The simulation results compared with those of genetic algorithm and list scheduling which are typical in literature, show that it produces encouraging results in both solution complexity and execution time. 2006 Vol. 19 (1): 73-78 [Abstract] ( 218 ) [HTML 1KB] [ PDF 358KB] ( 630 ) 79 A Clustering Algorithm Based on the Trace of Sample Covariance Matrix HUANG XiaoBin, WANG JianWei, ZHANG Yan Aiming at the shortage of traditional clustering algorithm when dealing data with some nonsphericalshape distribution, a novel clustering algorithm based on the trace of sample covariance matrix is presented in this paper. This algorithm is made up of the three main parts-uniform for data, constitution of initial patterns and fusion of initial patterns. The simulation results show that compared with the traditional FCA, the proposed algorithm has good clustering performance for data with some nonsphericalshape distribution without the number of clustering. 2006 Vol. 19 (1): 79-83 [Abstract] ( 428 ) [HTML 1KB] [ PDF 515KB] ( 663 ) 84 Research on Moving Objects Detection Using HOS YUAN Jie, DU SiDan, GAO DunTang In this paper, static background image abstraction algorithms based on video techniques is proposed. It works on the intrinsic properties of the high order statistic value of the two dimension images. The static background image can be distilled through a number of continuous or uncontinuous video frames no matter whether it contains the moving objects or not. Since the static background image has been got, the difference operation between the later video frames and the abstracted static background image can perfectly locate the moving target. The comparison between the method mentioned in this paper and traditional difference algorithm shows the former has the advantages in noise resistance and selfadaptation. Experiments are provided to show the validation of our algorithm. 2006 Vol. 19 (1): 84-88 [Abstract] ( 266 ) [HTML 1KB] [ PDF 770KB] ( 383 ) 89 Mountain Clustering Based on Improved PSO Algorithm SHEN HongYuan, PENG XiaoQi, WANG JunNian, HU ZhiKun The PSO (particle swarm optimization) algorithm is reformed so that it can be used in multimodel function optimization. The improved PSO is combined with a mountain clustering method. A mountain clustering based on improved PSO (MCBIPSO) algorithm is presented. The principle and steps are supplied in this paper. The simulation results show that the mechanism of the MCBIPSO algorithm is definite. When the MCBIPSO algorithm is used in clustering based on density, the calculation is easier and more efficient in deciding the clustering centers of data samples. The MCBIPSO algorithm can realize accuracy clustering based on the samples density. 2006 Vol. 19 (1): 89-93 [Abstract] ( 219 ) [HTML 1KB] [ PDF 868KB] ( 461 ) 94 A Dynamic RBF Neural Network for Pattern Recognition HAN Min, CUI PiSuo The problem of training RBF (Radial Basis Function) neural network for pattern recognition is considered. In this paper, taking account of the specific feature of classification problem, a new training algorithm based on the regional mapping and novelty condition of RAN (Resource Allocating Network) is proposed. The results show the effectiveness of the proposed approach in RBF network training for pattern recognition, mainly in shortening the learning time, simplifying the structure of network and improving the classification accuracy. 2006 Vol. 19 (1): 94-99 [Abstract] ( 221 ) [HTML 1KB] [ PDF 415KB] ( 615 ) 100 Visual Saliency Based Natural Landmarks Detection under Unknown Environments WANG Lu, CAI ZiXing Natural landmark detection is a basis of mobile robots navigation to represent and recognize unknown environments. A saliency based adaptive natural landmarks detecting system is presented in this paper. Firstly the detail preserving sampling scheme is designed to create multiscale image spaces where opponencies of color and texture are computed. And the Gabor filter, which can adjust parameters adaptively, is designed to analyze texture of all kinds of environments. At last the saliency map that points out where can be treated as natural landmark is created. Experiments show that this algorithm has better precision on detecting salient points and better repeatability including scale, rotation and viewpoint invariance. 2006 Vol. 19 (1): 100-105 [Abstract] ( 216 ) [HTML 1KB] [ PDF 1621KB] ( 483 ) 106 Research on Principal Component Analysis in Choosing Target Category Feature and Its Application to Target Recognition LI JunMei, HU YiHua In this paper, the opinion is presented that category feature includes two kinds: classification feature and recognition feature. The idea is described that category feature can be received by traditional PCA (principal component analysis) and improved PCA. The example calculating category feature is given in this paper. The analysis shows that the recognition precision will be improved greatly as the unknown target is compared twice by two kinds of category feature. 2006 Vol. 19 (1): 106-110 [Abstract] ( 246 ) [HTML 1KB] [ PDF 326KB] ( 321 ) 111 Method of Replacing the User with Machine in Interactive Genetic Algorithm HAO GuoSheng, GONG DunWei, SHI YouQun, SUN XiaoYan It is an important to replace a user with the machine in interactive algorithm, because compared with tireless machine, a user is apt to be tired. Firstly, three basic viewpoints are put forward. The first viewpoint is that the machine plays the role of environment to select the individual. The second is that the chance for machine’s sampling and replacing user should be in the phase during which a user’s preference doesn’s fluctuate again. The third is that the result of optimization is determined by the sampling data and the strategies that the machine applies. Next, the individual fitness estimation method based on genesenseunit fitness is given. Its efficiency is validated by comparative experiment. 2006 Vol. 19 (1): 111-115 [Abstract] ( 207 ) [HTML 1KB] [ PDF 364KB] ( 360 ) 116 Feature Extraction Based on Symmetrical ICA and Its Application to Face Recognition ZHENG YuJie, YANG JingYu, WU XiaoJun, YU DongJun Independent Component Analysis (ICA) has been extensively used in the field of signal processing and image processing. In this paper, a new algorithm called symmetrical ICA (SICA) based on facial symmetry is proposed. This algorithm is based on the theory of function decomposition in algebra and mirror symmetry in geometry. In this algorithm, mirror transform is firstly introduced. Then, even/odd symmetrical samples are produced based on the theory of the even/odd decomposition principle, and the even/odd independent components are extracted from the corresponding samples respectively. Both theoretical analysis and experimental results demonstrate that this algorithm not only enlarges the number of training samples, but also remarkably raises the recognition rate. Experiment results also show this algorithm is not sensitive to the illumination variation of human faces. 2006 Vol. 19 (1): 116-121 [Abstract] ( 295 ) [HTML 1KB] [ PDF 962KB] ( 761 ) 122 Complete Coverage Path Planning of Mobile Robots: Biologically Inspired Neural Network and Heuristic Template Approach QIU XueNa, LIU ShiRong, YU JinShou, Simon X. Yang In this paper, a novel complete coverage path planning method based on biologically inspired neural network for mobile robot motion planning is developed, which integrates heuristic searching algorithm, templatebased model and obstacle approaching algorithm. The biological neural network that is described by the shunting cooperativecompetitive feedback network is used to model the environment of the workspace of mobile robot. The templatebased model, heuristic searching algorithm and obstacle approaching algorithm are employed to plan the motion path of a mobile robot with obstacle avoidance. The obstacle approaching algorithm is used to cover the vicinity areas of the irregular obstacles so that the coverage area of the path planning is further improved. The simulation studies show that the performance of the path generated by the proposed method, such as the rate of the repeated coverage, is improved obviously, and the proposed algorithm is computationally simple and effective. 2006 Vol. 19 (1): 122-128 [Abstract] ( 532 ) [HTML 1KB] [ PDF 915KB] ( 825 )

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