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.2, Published 2010-04-25 Orignal Article        Orignal Article 137 An Improved PSO Based on Diversity of Particle Symmetrical Distribution SUN Yue-Hong,WEI Jian-Xiang,XIA De-Shen Particle swarm optimization (PSO) is easy to fall into the local optimum as the diversity of population gets worse and worse during the evolution. To overcome the shortcoming, an improved PSO based on the diversity of particle symmetrical distribution (sdPSO) is developed. Over the research of the spatial distribution of particles, it can be found that the convergence probability to the global optimum solution is greatly improved with more symmetrical particle distribution surrounding the optimum solution of particles. A diversity population function is proposed and an adjustment algorithm for the diversity is introduced into the basic PSO. The spatial distribution of particles varies between asymmetry and symmetry repeatedly while the population diversity is adjusted continually, which make the improved algorithm search in a wider range. The simulation results show that the improved sdPSO algorithm achieves better convergence precision than the basic PSO by the experiment of benchmark functions. 2010 Vol. 23 (2): 137-143 [Abstract] ( 447 ) [HTML 1KB] [ PDF 469KB] ( 743 ) 144 LDS Trajectories under One-to-One Mappings in Interval Type-2 Fuzzy Sets MO Hong,WANG Fei-Yue,ZHAO Liang The interval type-2 fuzzy extension principle is presented and the conventional one-to-one mapping is abstracted as its interval type-2 fuzzy counterpart. Then, the computing with words procedure based on the interval type-2 fuzzy extension principle is introduced. Finally, the linguistic dynamic trajectories of interval type-2 fuzzy sets are analyzed. 2010 Vol. 23 (2): 144-147 [Abstract] ( 372 ) [HTML 1KB] [ PDF 225KB] ( 675 ) 148 Micro-Expression Recognition Framework Using Time Series Analysis WANG Shang-Fei,ZHANG Feng,WANG Xu-Fa Relying on existing exaggerated expression video database and micro-expression being regarded as former part of exaggerated expression image series, a micro-expression recognition framework based on time series analysis is presented. Firstly, five dimensions feature series, action direction and intensity rate of eyebrows, nose and mouth, are extracted by fusing optical flow field of binary videos and gray ones. Secondly, hidden Markov models are trained by adopting exaggerated expression feature series, the relationship being analyzed between feature series and exaggerated expressions. Finally, these models are used to predict the variety trend of micro-expressions and recognize them and boosting algorithm is employed to increase recognition accuracy. The effectiveness of the approach is evaluated on Cohn Kanade facial expression database and a preferable experiment result is obtained. 2010 Vol. 23 (2): 148-153 [Abstract] ( 343 ) [HTML 1KB] [ PDF 454KB] ( 833 ) 154 Mean Field Interval Propagation Algorithm Based on Ising Computation Tree CHEN Ya-Rui, LIAO Shi-Zhong A mean filed interval propagation algorithm is designed based on incomplete functional iterations. This algorithm can yield the expectation bound of variables. Firstly, a concept of computation tree is proposed to reveal the iteration computation process of Ising mean field. Then, a mean field interval propagation algorithm based on the Ising computation tree is put forward, which propagates message intervals through the computation tree and presents the mean intervals of random variables in root node. It is proved that the variable mean interval computed by the interval propagation algorithm with 2-layer computation tree contains the exact value, called the mean bound of random variable. Finally, theoretical and experimental results show that the interval propagation algorithm is valid and the mean bound is tight. 2010 Vol. 23 (2): 154-159 [Abstract] ( 339 ) [HTML 1KB] [ PDF 361KB] ( 541 ) 160 Geometric Interpretations and Applications of the Extrinsic Parameters Derived from the Camera Calibration Based on Spheres JIA Jing,JIANG Guang,WU Cheng-Ke Camera calibration based on spheres has been lucubrated in recent years. The geometric relationship between the projections of spheres and absolute conic (IAC) are reinterpreted in this paper. The geometric meaning between spheres and the extrinsic parameters is presented. A method for calculating the sphere centers based on the orthogonal decomposition is put forward. Compared with the existing methods, the proposed method is clear and simple. A method by using three spheres to calculate the camera extrinsic parameters is proposed as well. Experimental results show the two methods have high precision and can be widely applied in visual platform for parameters calculation of camera motion. 2010 Vol. 23 (2): 160-164 [Abstract] ( 429 ) [HTML 1KB] [ PDF 294KB] ( 971 ) 165 An Automatic Language Identification Method Based on Supervector Subspace Analysis SONG Yan,DAI Li-Rong,WANG Ren-Hua In automatic spoken language identification on telephone conversation speech, the interference caused by different signal path, different speech content or different speakers is a major factor affecting the performance. To tackle this problem, a supervector subspace analysis based automatic language identification method is proposed. In this method the supervector is first introduced to represent the training utterances. Then a discriminative training method is applied based on SVM model. Furthermore the subspace analysis is utilized to estimate the noise subspace. Finally the noise is subtracted from the distance metric. The experiments on NIST 07 30 and 10 sec evaluation task show the advantage of this method and compared with baseline system the performance is increased clearly, the EER(Equal Error Rate) being reduced relatively about 20%. 2010 Vol. 23 (2): 165-170 [Abstract] ( 325 ) [HTML 1KB] [ PDF 425KB] ( 648 ) 171 An Improved Fractional Differential Mask WANG Wei-Xing,YU Xin,LAI Jun Applying the fractional differential theory to image processing is a new research topic. According to the functional characteristics of fractional differential Tiansi operator, an algorithm is proposed to improve Tiansi operator. The algorithm significantly enhances the edge information value. It is based on the enhancement ability of fractional differential to image details and the mechanism analysis of fractional differential. Firstly, Tiansi template is divided into eight sub-templates with different directions around the detecting pixel, and then the eight weight sum values for the eight sub-templates are obtained. Next, the eight weights are divided into different groups. Consequently, three improved algorithms with different enhancing ranges are gained. Finally, the experimental results of edge enhancement show that the improve algorithm enhances edge information effectively and reveals more detailed information than traditional operators for rock fracture images. 2010 Vol. 23 (2): 171-177 [Abstract] ( 538 ) [HTML 1KB] [ PDF 417KB] ( 1363 ) 178 Discriminant Maximum Margin Criterion Based on Locality Preserving Projections LIN Ke-Zheng,WANG Hui-Xin,BU Xue-Na,LIN Sheng A manifold learning algorithm is proposed called discriminant maximum margin criterion (DMMC). It adopts linear projective maps and optimally preserves the local structure and the global information of the data set simultaneously. DMMC tries to find the intrinsic manifold that discriminates different face classes best by maximizing the between-class scatter and minimizing the within-class scatter. The recognition rate of the proposed algorithm exceeds those of the single PCA,Fisherfaces,MMC and LPP greatly. Experimental results on YALE and JAFFE face databases indicate that the proposed algorithm is effective. 2010 Vol. 23 (2): 178-185 [Abstract] ( 365 ) [HTML 1KB] [ PDF 492KB] ( 657 ) 186 Image Segmentation Based on Geometric Active Contour Model CHEN Bo,DAI Qiu-Ping A geometric active contour model is proposed to reduce the influence of noise on image segmentation. The corresponding partial differential equations evolved by the level set curve are got through variational principle. Prior information of the regions and boundaries of the image is considered in this model and the statistical information of the image is considered as well. Moreover, a penalized term is used as an internal energy term to avoid the time-consuming re-initialization process. To verify the efficiency of the proposed model, a segmentation instance based on simple Gauss probability density function is given, and the additive operator splitting (AOS) scheme which is efficient and unconditionally stable is employed. Experimental results show that the proposed model has high accuracy, efficiency and noisy resistance. 2010 Vol. 23 (2): 186-190 [Abstract] ( 468 ) [HTML 1KB] [ PDF 363KB] ( 698 ) 191 Image Semantic Analysis and Understanding: A Review GAO Jun,XIE Zhao,ZHANG Jun,WU Ke-Wei Semantic analysis is the importance and difficulty of high-level interpretation in image understanding, in which there are two key issues of text-image semantic gap and text description polysemy. Concentrating on semantization of images ontology, three sophisticated methodologies are roundly reviewed as generative, discriminative and descriptive grammar on the basis of concluding images semantic features and context expression. The objective benchmark and evaluation for semantic vocabulary are induced as well. Finally, the summarized directions for further researches on semantics in image understanding are discussed intensively. 2010 Vol. 23 (2): 191-202 [Abstract] ( 1045 ) [HTML 1KB] [ PDF 999KB] ( 2452 ) 203 Writer Identification of Offline Chinese Handwriting Documents Based on Feature Fusion YAN Yu-Chen,CHEN Qing-Hu,YUAN Feng,DENG Wei A text-independent method for writer identification is proposed based on the fusion of text-dependent features. A two-factor model is developed, and thus the feature of handwriting is decomposed into the character factor and the writing factor. According to the deviation analysis of the two factors and data mining, the writing factor is separated, the text-independent feature based on text-dependent method is extracted, and the classifier is developed. The proposed method is effective for the samples with few characters, especially for the samples without the same characters between the unknown script and the corresponding one in the database. The proposed method provides a new way for writer identification with few characters in the writing sample. 2010 Vol. 23 (2): 203-209 [Abstract] ( 531 ) [HTML 1KB] [ PDF 443KB] ( 688 ) 210 SVM Parameter Selection Algorithm Based on Maximum Kernel Similarity Diversity TANG Yao-Hua,GUO Wei-Min,GAO Jing-Huai Aiming at support vector machine (SVM) parameter selection problem, a novel Gaussian kernel parameter rapid selection algorithm is proposed on the basis of kernel similarity diversity maximum (MSD) by analyzing the equivalent network model and the classification principle of SVM. In addition, MSD is combined with parameter search algorithm based on cross validation, and thus it is a composite parameter selection algorithm (MSD-GS) to the realize rapid and optimal selection of kernel parameter and regularization parameter. Simulation experiment results on data sets from UCI show that MSD-GS has the merits of simpleness, celerity and accurate parameter selection with no need of adding prior knowledge. The parameter selection result is better than the traversing exponential grid search algorithm. The selected couple of SVM parameters can make SVM get high generalization performance. 2010 Vol. 23 (2): 210-215 [Abstract] ( 383 ) [HTML 1KB] [ PDF 470KB] ( 886 ) 216 An Edge-Based Color Image Retrieval by Using Multiple Features WANG Xiang-Yang,CHEN Jing-Wei,YU Yong-Jian Feature extraction and representation is one of the most important techniques in the content-based image retrieval (CBIR). An image edge is the boundary between an object and the background and indicates the boundary between overlapping objects. In this paper, an edge-based color image retrieval by using multiple features is proposed. Firstly, the color edge is extracted by using canny detection operator. Secondly, the weighted color histogram, the angle histogram and the gradient orientation histogram about the extracted color edge image are computed as image features. Finally, the similarity between color images is computed by combined feature index based on three kinds of histograms. Experimental results show that the proposed image retrieval is accurate and efficient in the user-interested images retrieval. 2010 Vol. 23 (2): 216-221 [Abstract] ( 332 ) [HTML 1KB] [ PDF 416KB] ( 857 ) 222 A Variational Model for Integrating Registration and Segmentation via Level Set Evolution BAI Xiao-Jing,ZHANG Jie-Yu,SUN Quan-Sen,XIA De-Shen,SUN Huai-Jiang A variational model for integrating registration and segmentation is proposed. A non-parametric registration method based on the abstract matching flow model is adopted as the registration term to go along with the non-parametric segmentation term, handling the problem of inconsistence on definition format and solving plan between parametric registration based on B spline and non-parametric active contour model. An edge-based active contour model is applied to segment the region of interest, and the improved model by adding region statistic information deals with the problem of sensitivity to the initialization. The integrated model is directly defined by the level set function and has the merits of intuitionstic definition and simple numerical solution. The validity of the model is verified via the experiments on single modal and multimodal brain images. 2010 Vol. 23 (2): 222-227 [Abstract] ( 310 ) [HTML 1KB] [ PDF 433KB] ( 543 ) 228 Cost-Sensitive ROI Detection Method for Medical Images Based on Cascade Architecture LI Ning,GUO Qiao-Jin,XIE Jun-Yuan,CHEN Shi-Fu Regions of Interest (ROI) in medical images contain important information and are of great significance to the analysis and diagnosis. A cost-sensitive ROI detection method for medical images based on Cascade architecture is proposed in this paper, which combines the characters of medical images and applies machine learning and image process. This method achieves high sensitivity and efficiency by effectively integrating cost-sensitive classifier method and Cascade architecture. Experimental results on mammograms show that the method is more efficient and less in calculated amount than pixel-based methods, meanwhile avoids the difficulty of detecting masses by using traditional segmentation and filtering techniques with region-based approach. 2010 Vol. 23 (2): 228-234 [Abstract] ( 301 ) [HTML 1KB] [ PDF 548KB] ( 729 ) 235 Feature Selection Algorithm Based on Kernel Distance Measure CAI Zhe-Yuan, YU Jian-Guo, LI Xian-Peng,JIN Zhen-Dong The kernel distance measure is proposed as a new type of class separability. The distance of samples between two classes is measured in the kernel space and used to evaluate the separability of subsets. Using the sequential forward selection algorithm as the search strategy, tests are carried out on both synthetic and real datasets. Experimental results demonstrate that the proposed method outperforms the traditional non-kernel class separability. Moreover, the proposed method is superior or close to the kernel scatter matrix measures proposed by Wang and its running time is an order of magnitude faster. When applied to the pancreatic EUS image classification, the proposed method receives a good result. 2010 Vol. 23 (2): 235-240 [Abstract] ( 445 ) [HTML 1KB] [ PDF 381KB] ( 808 ) 241 Robust Least Squares Support Vector Machine Regression Based on Hypothesis-Testing and Outlier-Elimination WEN Wen,HAO Zhi-Feng,YANG Xiao-Wei,ZHAN Yin-Wei A robust least squares support vector machine (RLS-SVM) algorithm for regression is proposed based on the recursive outlier-elimination. In each loop, the sample with the largest error is investigated and diagnosed by statistical hypothesis testing. If the sample is diagnosed as an outlier, it is eliminated and the LS-SVM is re-trained by using the rest samples to provide more accurate information for the successive outlier diagnosis and elimination. The decremental-learning method is introduced into the re-training stage to reduce the computations. Thus, the additional computational complexity of RLS-SVM is less than O(N3). Experimental results on simulated and real-world datasets demonstrate the validity of the proposed algorithm and reveal the potential of the algorithm in building an outlier detector. 2010 Vol. 23 (2): 241-249 [Abstract] ( 329 ) [HTML 1KB] [ PDF 623KB] ( 1100 ) 250 Object Tracking MethodBased on Incrementally Updating Linear Discriminant Subspace QIAN Cheng,XU Shu-Chang,ZHANG Yin, ZHANG San-Yuan A method for object tracking is presented to track objects steadily based on incremental linear discriminant analysis. The locations and poses of the object are represented by a set of affine parameters. Resorting to a state transition model, a group of image samples are predicted as candidates of the image patches of the object in the next frame. Their likelihoods of being the object image patch are measured after they are projected into a linear discriminant subspace. Then a sample with maximum likelihood is regarded as the object image patch. Finally, sufficient spanning sets of total scatter matrix and between-class scatter matrix are rotated to update projection matrix for maintaining the discrimination power of the subspace. Experimental results show that the method is robust to variation in appearances of objects and surrounding background, and it is available in affine invariant tracking. 2010 Vol. 23 (2): 250-255 [Abstract] ( 301 ) [HTML 1KB] [ PDF 475KB] ( 596 ) 256 Feature Extraction of Customer Purchase Behavior Based on Genetic Algorithm ZHANG Zhi-Hong,KOU Ji-Song,CHEN Fu-Zan,LI Min-Qiang A feature extraction method for customer purchase behavior based on genetic algorithm (GA) is proposed. Firstly, Tanimoto similarity is used to measure purchase behavior similarity between customers, and a clustering method based on genetic algorithm is designed to cluster customers who have similar purchase behavior in the same subpopulation. Then, an customer feature extraction method based on multi-population genetic algorithm is presented to find out knowledge from all kinds of subpopulation. To promote coevolution within the population and the quality of rule set, q-nearest neighbor replacement policy and local search are adopted. The proposed algorithm is validated by using real-world retail data and is compared with Apriori algorithm. Experimental results show that the proposed algorithm can efficiently yield condensed rule sets without generating frequent itemsets and is more flexible in rule form as well. Finally, the experimental results are analyzed in detail. 2010 Vol. 23 (2): 256-266 [Abstract] ( 294 ) [HTML 1KB] [ PDF 760KB] ( 834 ) 267 Human Motion Analysis Based on Improved Dynamic Texture Model CHEN Chang-Hong,ZHAO Heng,HU Hai-Hong,LIANG Ji-Min Human motion analysis is one of the most active subjects in computer vision. Two improved dynamic texture models are proposed for human motion sequence description, binary dynamic texture model and tensor subspace dynamic texture model. A binary image is supposed to submit to Bernoulli distribution, and the logistic principle component analysis is used to learn the parameters of the binary dynamic texture model. In tensor subspace dynamic texture model, a binary image is treated as a tensor with dimensions of column and row reduced by tensor subspace analysis, and then it is transformed to a low-dimensional gray image. The dynamic texture model is applied to describe the gray image sequence. Experimental results on human activity and gait databases show the validity of the two proposed improved dynamic texture models. 2010 Vol. 23 (2): 267-272 [Abstract] ( 340 ) [HTML 1KB] [ PDF 427KB] ( 698 ) 273 Optimization Algorithm for Edge-Based Semantic Similarity Calculation WANG Zhi-Xiao,ZHANG Da-Lu Concept semantic similarity calculation is a problem in applications. To simplify the single ontology concept semantic similarity calculation, an optimization algorithm for edge-based similarity calculation is put forward. It utilizes hierarchical relationship between concepts to simplify calculation process. Based on the semantic similarity between a pair of ontology concepts, the optimization algorithm gives all semantic similarity between each pair of ontology concepts. Results of simulation experiments show that the calculation complexity is reduced considerably, and the similarity calculation speed is improved by a factor of two. 2010 Vol. 23 (2): 273-277 [Abstract] ( 309 ) [HTML 1KB] [ PDF 285KB] ( 761 ) 278 Facial Expression Recognition Based on SLLE with Expression Weighted Distances YING Zi-Lu,LI Jing-Wen,ZHANG You-Wei The traditional locally linear embedding (LLE) algorithm doesn’t take into consideration the class label information of training samples while the supervised LLE (SLLE) algorithm treats the different between classes equally which is inappropriate for facial expression recognition. Taking the differences between expressions into account, a new supervised locally linear embedding (SLLE) algorithm called expression related SLLE (ERSLLE) is designed which uses different weights for sample distance calculation in determining neighborhood samples. The proposed algorithm is applied to the facial expression recognition on the Japanese female facial expression database (JAFFE). The results show that the proposed algorithm is effective and superior to the traditional LLE and SLLE. Better performance is obtained for facial expression recognition in a certain range of the number k of the nearest neighborhood, compared with LLE and SLLE. 2010 Vol. 23 (2): 278-283 [Abstract] ( 308 ) [HTML 1KB] [ PDF 416KB] ( 742 ) 284 A Genetic Algorithm Based on Random Uniform Design Point Set for Solving MVCP REN Zhe,ZHOU Ben-Da,CHEN Ming-Hua Based on the mechanism analysis of ideal density model and by utilizing the principle and method of random uniform design (RUD), the crossover operation in genetic algorithm (GA) is redesigned. Then, on the basis of characteristic analysis of the minimum vertices covering problem (MVCP) in graph and combining scan-repair and local improvement techniques, a GA based on RUD point set is presented to solve the MVCP. Compared with simple GA and Good Point GA for solving this problem, the simulation results show that the presented GA has superiority in speed, accuracy and overcoming premature. 2010 Vol. 23 (2): 284-290 [Abstract] ( 337 ) [HTML 1KB] [ PDF 350KB] ( 545 )

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