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....

2009 Vol.22 Issue.1, Published 2009-02-25 Papers and Reports    Researches and Applications    Surveys and Reviews        Papers and Reports 1 A Hybrid Algorithm Based on PBIL Algorithm and Zooming Algorithm WANG Gao-Peng, DOU Li-Hua, CHEN Jie, ZHANG Juan, CHEN Chen Population based incremental learning (PBIL) algorithm has the advantage of simple execution process, quick and accurate solutions to problem. Aiming at the domino phenomenon of convergence from the highest position to the lowest position of binary coding in PBIL algorithm, a zooming algorithm is employed to improve the search efficiency and solution accuracy. The simulation results based on Benchmark functions of different dimensions verify that the proposed hybrid algorithm has the advantage of global convergence, high solution precision and search efficiency. 2009 Vol. 22 (1): 1-7 [Abstract] ( 248 ) [HTML 1KB] [ PDF 429KB] ( 658 ) 8 Total Margin v-Support Vector Machine and Its Geometric Problem PENG Xin-Jun, WANG Yi-Fei A total margin v-support vector machine (TM-v-SVM) is presented and it has better theoretical classification performance than v-SVM. The theoretical research shows that the TM-v-SVM is equivalent to the problem of finding the closest pair of points between two compressed convex hulls (CCHs) in the feature space. A geometric algorithm based on the theoretical properties of CCHs is proposed. Simulation results show that the TM-v-SVM and its geometric algorithm have better performance than the previous methods. 2009 Vol. 22 (1): 8-16 [Abstract] ( 312 ) [HTML 1KB] [ PDF 1059KB] ( 505 ) 17 Chinese Dependency Parsing Based on Treebank LIU Hai-Tao, ZHAO Yi-Yi Aiming at exploring the possibility of increasing the parsing accuracy by linguistic means, an experiment of Chinese dependency parsing is conducted by using MaltParser and a self-built treebank. Through the detailed analysis for the parsing results, the possible suggestion about improving the performance of the parser is provided and it is used as the guidance to modify the annotation scheme of the treebank. Experimental results show that the accuracy of unlabeled dependency attachment score increases 5.5%, and the accuracy of labeled score raises 7.5%. 2009 Vol. 22 (1): 17-21 [Abstract] ( 316 ) [HTML 1KB] [ PDF 400KB] ( 709 ) 22 Granular Computing Based Hierarchical Concept Capture Algorithm in Multi-Valued Information System QIU Tao-Rong, LIU Qing, HUANG Hou-Kuan An approach is introduced to obtain hierarchical concepts from multi-valued information system with inaccurate or uncertain values. The approximate representations of attribute values are analyzed, and the way of generating tolerance information granules is described. The framework and algorithms are proposed for acquiring hierarchical concepts. Experimental results show that the proposed algorithms are useful and effective. 2009 Vol. 22 (1): 22-27 [Abstract] ( 226 ) [HTML 1KB] [ PDF 369KB] ( 424 ) 28 Speech Enhancement Method Based on Wavelet Transform and Kalman Filters ZHANG En-Dong , HUANG Wen-Hao Using kalman filter model based on wavelet transform, an efficient speech enhancement algorithm is proposed aiming at the speech signal with additive noise. Dyadic wavelet transform, parameter estimation for Kalman filter model and the divergence of Kalman filter are analyzed in detail. The quality of the resulting enhanced speech is evaluated by means of signal to noise ratio (SNR). Simulation results indicate that the proposed method is valid under white Gaussian noise and colored noise. 2009 Vol. 22 (1): 28-31 [Abstract] ( 296 ) [HTML 1KB] [ PDF 359KB] ( 508 ) 32 Mean-Standard Deviation Descriptor and Line Matching WANG Zhi-Heng, WU Fu-Chao An idea is put forward for automatic line matching based on line descriptor. Firstly, a parallel neighborhood for a line segment is defined and it is decomposed into several parallel line segments. Next, a line description matrix (DM) is formed by selecting an image feature. Finally, the descriptor of line descriptor is obtained by computing the mean and standard deviation of column vectors of DM. Thus, the line descriptor construction is accomplished. Based on different image features (gray, gradient and gradient magnitude), three line descriptors are proposed for line matching and all of them are invariant to image shift, rotation, and linear illumination changes. Experimental results show that these descriptors have good performance in line matching. 2009 Vol. 22 (1): 32-39 [Abstract] ( 295 ) [HTML 1KB] [ PDF 1524KB] ( 392 ) 40 Design of Data Integration System Supporting Intelligent Match-Retrieval XIE Xing-Sheng, ZHANG Yi-Ming, YU Yin, ZHUANG Zhen-Quan A method for integrating distributed heterogeneous data is proposed. The data query processing is implemented by issuing and registering data services actively in an unit for all network nodes, called data service cell (DS-Cell). Then the data services are retrieved by an intelligent matcher. By effectively interfusing the techniques of semantic web, web ontology language (OWL), and description logic, the proposed method can work with data formal semantics and ontology based reasoning. A prototype of data integration system is designed based on the proposed method, and the experimental tests on the core component of the system are conducted. The experimental results show that the proposed system works effectively and reliably with favorable flexibility, scalability and performance. 2009 Vol. 22 (1): 40-46 [Abstract] ( 243 ) [HTML 1KB] [ PDF 807KB] ( 391 ) 47 New Class Recognition Method Based on Distance Metric Learning XIE Mao-Qiang, HUANG Ya-Lou, YIN Ai-Ru, JIANG Hao, LI Dong In online classification tasks, new class of patterns sometimes emerges, which makes the distribution change significantly and current classification models invalid. A method based on distance metric learning is proposed to recognize new class from existing classes without the apriori knowledge about emerging class. And it can make the class similarity represented by using distance between two objects, which is the key of promoting the performance of recognition. Therefore, the proposed method can be applied to the adaptive classification. The experimental results show that the proposed method can recognize new class well, and on the basis of this method the online classifier is adapted and it can predict the instance better than the original one. 2009 Vol. 22 (1): 47-52 [Abstract] ( 223 ) [HTML 1KB] [ PDF 458KB] ( 560 ) 53 Rough Set Models of Incomplete Information System Based on Random Fuzzy Set QIU Wei-Gen The traditional Pawlak rough set theory has some limitations in treatment of the incomplete information systems, therefore, it is of great importance to study its related theories and methods. A fuzzy-values completion method of incomplete information systems is proposed according to the expert professional knowledge. Two kinds of composite fuzzy relations based on the random fuzzy set are constructed in objects universe, which is a starting point of construction of the composition rough set model. The Krysckiewcz rough set model and the Stefanowki rough set model are generalized to the fuzzy case, and some related important concepts are extended accordingly as well. The results provide a way for the rough set theory to utilize the incomplete information systems. 2009 Vol. 22 (1): 53-59 [Abstract] ( 241 ) [HTML 1KB] [ PDF 330KB] ( 466 ) 60 Face and Expression Recognition Method by Combing 2DFLD with LPP ZHU Ming-Han, LUO Da-Yong A face and expression recognition method by combing 2DFLD with LPP is proposed. Firstly, each facial image in training set is mapped by using 2DFLD according to its identity. Then, expression manifolds of training set are obtained by LPP. Finally, each identity and expression probability of test image is computed according to the given probability metric, and face and expression recognition of the test image is thus accomplished. The experimental results on CMU-AMP and JAFFE face database show the effectiveness of the proposed method. 2009 Vol. 22 (1): 60-64 [Abstract] ( 277 ) [HTML 1KB] [ PDF 325KB] ( 535 ) 65 Knowledge Representation Using Partition Granularity FENG Qin-Rong, MIAO Duo-Qian,CHENG Yi, XU Fei-Fei A quantitative representation for the classification ability of knowledge is described, and a representation for knowledge is given. Firstly, the algebraic representation of main concepts in rough sets is presented. Next, the concept of partition granularity is defined, and its properties are analyzed. Finally, the equivalence between the algebraic representation of knowledge and partition granularity representation is proved. 2009 Vol. 22 (1): 65-69 [Abstract] ( 289 ) [HTML 1KB] [ PDF 0KB] ( 139 ) Surveys and Reviews 70 Some Advances in Human Motion Analysis LI Hong-Song, LI Da Visual analysis includes moving object detection, moving object classification, human tracking and activity recognition and description. It has broad application prospects in many fields, such as smart vision surveillance, visual reality, intelligent human-computer interface, video compression and computer-aided clinical diagnosis. A comprehensive survey on vision-based human motion analysis is presented from the above four aspects, and the challenges and future directions are discussed. 2009 Vol. 22 (1): 70-78 [Abstract] ( 313 ) [HTML 1KB] [ PDF 473KB] ( 787 ) Researches and Applications 79 Image Recognition with Kernel Matching Pursuit Classifier Ensemble Based on Immune Clone GOU Shui-Ping, JIAO Li-Cheng, ZHANG Xiang-Rong An algorithm for kernel matching pursuit classifier (KMPC) ensemble based on immune clone for image recognition is presented to select a subset of the optimal individual from classifier ensemble system and eventually to improve the performance of classifiers. Based on the strong abilities of global optimal search and local search of immune clone algorithm, the proposed method can get ensemble system with better general performance by selecting son kernel matching pursuit from training classifiers. The recognition experiments are made on Brodatz texture image sets and SAR image. The results show that the performance of the proposed algorithm is better than that of the traditional ensemble system and the genetic algorithm based selective ensemble KMPC. 2009 Vol. 22 (1): 79-85 [Abstract] ( 253 ) [HTML 1KB] [ PDF 1421KB] ( 342 ) 86 Multilingual Acoustic Modeling Method Based on Phoneme Clustering MENG Meng, LIANG Jia-En, XU Bo A clustering method is proposed to generate multilingual global phoneme based on the decrease of model self-likelihood. Two linguistic limitations are used in the clustering procedure, and the phonemes in same language or belonging to different international phonetic alphabet (IPA) classes are not merged. In telephone speech keyword spotting system, the performance of several Chinese-English bilingual model are compared which are generated by different phoneme clustering methods. The experimental results show that the merged phoneme set of an appropriate size can generate acoustic models with good quality, far above the results without merging. Moreover, the linguistic limitations added to clustering procedure can improve the performance. 2009 Vol. 22 (1): 86-90 [Abstract] ( 351 ) [HTML 1KB] [ PDF 386KB] ( 649 ) 91 An Adaptive Niche Genetic Algorithm for Multimodal Function Optimization LU Qing, LIANG Chang-Yong, YANG Shan-Lin, ZHANG Jun-Ling An adaptive niche genetic algorithm is proposed to solve the problems of the inaccurate niche identification and the conflict between quick convergence and population diversity maintaining in niche genetic algorithms. In the proposed algorithm, an improved niche identification method is designed to identify the niches of the population, and a concept of niche entropy is introduced to measure the diversity of the population. The evolutionary parameters of the algorithm can be adjusted adaptively on the basis of the niche entropy of population. And the strategies of selection and crossover are also improved in the algorithm. The strategies divide the operation of crossover into the out-niche crossover and the in-niche crossover to enhance global searching ability and local convergence rate of the algorithm. Experimental results show that the proposed algorithm can solve the multimodal function optimization problems with quick convergence rate, low computational complexity and avoidance of the genetic drift. 2009 Vol. 22 (1): 91-100 [Abstract] ( 279 ) [HTML 1KB] [ PDF 673KB] ( 796 ) 101 An Optimization Algorithm for MultiObjective Flow Shop Scheduling in Uncertain Condition ZHOU Qiang, CUI Xun-Xue The model of optimizing multi-objective flow shop scheduling with stochastic processing time and machine breakdown is analyzed. The multi-objective flow shop scheduling problem is modeled with the stochastic processing time and machine breakdown. A mathematical scheme is designed for the solutions with the longest flow time or the longest delay time. A hybrid multi-objective genetic algorithm is proposed to solve the optimization problems iteratively in uncertain condition. The simulation results show that the proposed algorithm has good performance for the flow shop scheduling problems in uncertain condition. 2009 Vol. 22 (1): 101-107 [Abstract] ( 289 ) [HTML 1KB] [ PDF 590KB] ( 554 ) 108 Mongolian Word Segmentation Based on Statistical Language Model HOU Hong-Xu, LIU Qun, Nasanurtu, Murengaowa, LI Jin-Tao Based on the analysis of Mongolian segmentation technique and the rules used as the foundation of word segmentation, a hybrid word segmentation method is proposed. It uses Mongolian statistical language model to eliminate the ambiguity in Mongolian word segmentation. A POS language model and a Skip-N language model are used, and an experiment system is thus created. The experimental results are better than those of the system based on rules. 2009 Vol. 22 (1): 108-112 [Abstract] ( 291 ) [HTML 1KB] [ PDF 320KB] ( 459 ) 113 Random Projection Based Clustering Method of Parallel Data Streams CHEN Hua-Hui, SHI Bo-Le A synopsis is maintained dynamically for each data stream. The construction of the synopsis is based on random projections and it utilizes the amnesic feature of data stream. Using the synopsis, the approximate distances between streams and the cluster center can be computed fast. And an efficient online version of the classical K-means clustering algorithm is developed. The experimental results show the method can be performed effectively with a good clustering quality. 2009 Vol. 22 (1): 113-122 [Abstract] ( 278 ) [HTML 1KB] [ PDF 492KB] ( 519 ) 123 Facial Expression Recognition Based on Advanced Local Binary Pattern Histogram Projection FU Xiao-Feng, WEI Wei A method of advanced local binary pattern histogram projection (ALBPHP) is proposed. It projects advanced local binary pattern (ALBP) histogram, which has complete label information and unified label position, onto locality preserving projection (LPP) space to get the lower dimensional ALBPHP feature. Compared with ALBP feature, ALBPHP feature has lower dimensionality and stronger discriminative power in representing face image. Comparative experiments on JAFFE database and Cohn-Kanade database demonstrate that the recognition rate of ALBPHP is higher than that of ALBP when same classifiers are used. 2009 Vol. 22 (1): 123-128 [Abstract] ( 283 ) [HTML 1KB] [ PDF 410KB] ( 444 ) 129 Quality Classification Method for Fingerprint Image Based on Support Vector Machine ZHANG Yu, YIN Yi-Long, LUO Gong-Qing In an automatic fingerprint identification system, estimating the quality of fingerprint image has significant value for segmentation, enhancement and matching processes. Besides, the quality classification of fingerprint image is of paramount significance in the applicability research of fingerprint recognition algorithm. In this paper, a method for quality classification of fingerprint image is proposed based on the support vector machine (SVM). The gradient, Gabor feature, and directional contrast are used as the quality index, and SVM is applied to achieve quality classification of fingerprint image. Meanwhile, synthetic minority over sampling technique (SMOTE) method is employed to reduce the influence of class imbalance problem. Both the theoretical analysis and the experimental results indicate the validity of the proposed method. 2009 Vol. 22 (1): 129-135 [Abstract] ( 249 ) [HTML 1KB] [ PDF 1192KB] ( 599 ) 136 Speaker Recognition Based on Pitch-Dependent Affective Speech Clustering LI Dong-Dong, WU Zhao-Hui, YANG Ying-Chun Speech with various emotions aggravates the performance of speaker recognition system. A pitch-dependent affective speech clustering method for speaker modeling is proposed. This method aims to exploiting the affective material effectively in the speaker systems. Thresholds for pitches are determined for the male and the female separately. The cepstral features in the same pitch range are clustered. Different pitch-dependent models are built with the corresponding cluster features by map adaptation for each speaker. The maximum likelihood rule is applied to the matched models and the identification of the person. The proposed method is evaluated on the mandarin affective speech corpus. Experimental results show that the proposed approach is more powerful and efficient than the cepstral feature based method and the structure training method for speaker recognition. 2009 Vol. 22 (1): 136-141 [Abstract] ( 298 ) [HTML 1KB] [ PDF 478KB] ( 565 ) 142 A Genetic Algorithm Based Autonomous Localization Strategy for Mobile Robots HE Feng, QIN Xiao-Li, FANG Yong-Chun Aiming at the three main problems in localization of mobile robots, position tracking, global localization and kidnapped problem, an autonomous localization strategy based on genetic algorithm is proposed. A fitness function is designed based on the similarity of position. The real-coded method is used in the crossover and the mutation steps to improve the real-time ability of the algorithm. For the kidnapped problem, a scattering mechanism is introduced into the regular genetic algorithm. Thus, the population impoverishment problem is largely alleviated. Subsequently, the population state is updated with the kinematic model to achieve continuous localization of mobile robots. The experimental results of indoor environment demonstrate the validity of the proposed localization strategy. 2009 Vol. 22 (1): 142-147 [Abstract] ( 259 ) [HTML 1KB] [ PDF 686KB] ( 492 ) 148 Cellular Automate and QPSO Based Neural Network Structure Design by Indirect Encoding BAO Fang, PAN Yong-Hui, SUN Jun, XU Wen-Bo An algorithm for neural network structure design is proposed. The algorithm introduces indirect encoding schema to represent the structure of neural network and the cell in the 2-dimension cellular automate system to represent the existence of connection in neural network. By separately evolving the coordinate and value of the cell, the growing and pruning of the network structure are achieved. The coordinate of the cell is created and evolved by binary quantum particle swarm optimization (BQPSO). The value of the cell is evolved by using properly-designed neighboring evolving rule of cellular system, and the current network is trained by float-point QPSO. Thus, the final stable structure is found. The experimental results show that the proposed algorithm has stable complexity and convergent capability with different scales of neural network structure design. 2009 Vol. 22 (1): 148-155 [Abstract] ( 260 ) [HTML 1KB] [ PDF 446KB] ( 390 ) 156 Image Similarity Measure Based on Intuitionistic Fuzzy Set XU Shao-Ping, ZHANG Hua, JIANG Shun-Liang , YE Fa-Mao, XIONG Yu-Hong An intuitionistic fuzzy model for images based on the HSV color histogram is proposed. The image can be considered as an intuitionistic fuzzy set (IFS) by this model. Similarity measures are originally introduced to express the comparison between two fuzzy sets, and they can be used to reflect the resemblance of images. Experimental results show that the proposed approach can efficiently process queries of an image database in HSV color space and its accuracy rate is 5%~10% higher than those of fuzzy similarity measures and conventional histogram distances. 2009 Vol. 22 (1): 156-161 [Abstract] ( 306 ) [HTML 1KB] [ PDF 2833KB] ( 547 ) 162 Maximum Entropy Image Thresholding Based on Two-Dimensional Histogram Oblique Segmentation WU Yi-Quan, PAN Zhe, WU Wen-Yi The obvious wrong segmentation is pointed out in the existing two-dimensional histogram vertical segmentation method. A two-dimensional histogram oblique segmentation method is proposed. Then the formula and its fast recursive algorithm of the maximum Shannon entropy thresholding are deduced based on the two-dimensional histogram oblique segmentation. Finally, the threshold images and the processing time are given in the experimental results and analysis. The results are compared with those of the original maximum Shannon entropy algorithm and its fast algorithms based on the two-dimensional histogram vertical segmentation. The experimental results show that the proposed method makes the inner part uniform and the edge accurate in the threshold image, and it has a better anti-noise property. The processing time of the fast recursive algorithm of the proposed method is about 2% of that of the original two-dimensional maximum Shannon entropy algorithm, and it is less than one third of that of two fast recursive algorithms of the maximum Shannon entropy thresholding based on the two-dimensional histogram vertical segmentation. 2009 Vol. 22 (1): 162-168 [Abstract] ( 275 ) [HTML 1KB] [ PDF 1072KB] ( 580 )

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