模式识别与人工智能
Home      About Journal      Editorial Board      Instructions      Ethics Statement      Contact Us                   中文
Pattern Recognition and Artificial Intelligence
22 Judgement and Disposal of Academic Misconduct Article
22 Copyright Transfer Agreement
22 Proof of Confidentiality
22 Requirements for Electronic Version
More....
22 Chinese Association of Automation
22 National ResearchCenter for Intelligent Computing System
22 Institute of Intelligent Machines,Chinese Academy of Sciences
More....
 
 
2012 Vol.25 Issue.1, Published 2012-02-25
Orignal Article   
   
Orignal Article
1 Open Interior-Places Perception Algorithm of Robot Based on Prototype
ZHU Bo, DAI Xian-Zhong, LI Xin-De

As the open idea becomes popular in modern interior design, some function places are gradually transformed into open or half-open places, and the cognition of these types of places becomes a new challenge to service robots. An algorithm based on the prototype theory of cognitive psychology is proposed to improve the robots’ ability of open interior-places cognition. Firstly, the prototype model of place concept is designed, and it mainly includes a description about feature objects belonging to a place concept and a description about typical spatial relation among these objects. Secondly, a similarity measure function and a scoring criterion for spatial relation are proposed, and they are used to measure similarity between current environment and prototypes of place concepts. Finally, the perception toward place region is considered, and the influence on place-concepts perception is also discussed which is caused by overlap of place regions. The algorithm is verified through simulation, and the results suggest that the open interior-places perception can be achieved by the proposed algorithm. Moreover, it is flexible and robust to some extent.
2012 Vol. 25 (1): 1-10 [Abstract] ( 691 ) [HTML 1KB] [ PDF 662KB] ( 527 )
11 A Representation Model of Direction Relations on Hemispherical Surface
OUYANG Ji-Hong, LIU Yu-Feng

A Minimum Bounding Sector model (MBS) is proposed which represents the direction relationship on hemispherical surface. The model projects the region on the hemispherical surface to the bottom and divides the bottom region into 12 atom areas based on the projection of the region on the hemispherical surface. In this way, the model not only represents the direction relations such as left and right, but also represents the height information qualitatively by means of the distance between the points and the center on the projection plane. And the 1932 cases that exists in the MBS is also represented. As to the further research on reasoning and the composition table, the relational matrix is proposed through which the direction relations between regions on the hemispherical surface are deduced conveniently.
2012 Vol. 25 (1): 11-15 [Abstract] ( 503 ) [HTML 1KB] [ PDF 272KB] ( 442 )
16 Human Activity Classification Based on Features of 3D Micro-Doppler Signatures Shape
CHEN Yi-Wang, ZHANG Pin, FU Qiang
Different human activity classification based on 3 dimension shape of micro-Doppler signatures is studied. The 3D shape information, including time, frequency and power, is achieved by short-time Fourier transform to Doppler to obtain human pose and motion. The algorithm of point description image is used for obtaining the 3D shape characteristic. The data of 4 motions of 20 people are measured by Doppler radar. The action characteristics are learned by SVM with conformal transforming hyperkernel functions. Then the decision tree model is used for action classification. The improved of kernel functions is studied. The proposed human activity classification and the improved kernel functions are validated by experiments.
2012 Vol. 25 (1): 16-22 [Abstract] ( 638 ) [HTML 1KB] [ PDF 490KB] ( 695 )
23 Smile Recognition Based on PHOG Feature Extraction and Clustering Feature Selection
GUO Li-Hua, BAI Yang, JIN Lian-Wen
Gabor features are successfully applied to solve the problems of facial expression recognition. However, the dimension of Gabor features is usually too high to be practically applicable. A method based on Pyramid Histogram of Oriented Gradients (PHOG) feature and Clustering Linear Discriminate Analysis (CLDA) is proposed for smile expression recognition. The main merits of the proposed system are that the complexity can be decreased with low-dimension PHOG feature, and the multi-model problem can be overcome by CLDA. The experimental results show that system with PHOG feature achieves competitive or even higher recognition accuracy than with the Gabor feature, but with much lower of computation time cost. Moreover, the performance of CLDA does not be degraded significantly when decreasing the feature dimension.
2012 Vol. 25 (1): 23-28 [Abstract] ( 636 ) [HTML 1KB] [ PDF 406KB] ( 682 )
29 Behavior Cognition Computational ModelBased on Cerebellum and Basal Ganglia Mechanism
CHEN Jing, RUAN Xiao-Gang, DAI Li-Zhen
Aiming at agent’s behavioral cognition problem, a behavior cognition computational model based on the coordination of cerebellum and basal ganglia is proposed. Operant conditioning learning algorithm is the central algorithm including evaluation mechanism, action selection mechanism, tropism mechanism, and the coordination mechanism between cerebellum and basal ganglia. The learning signals come from not only the Inferior Olive but also the Substantia Nigra in the beginning. The convergence of the algorithm can be guaranteed in the sense of entropy. With the proposed method, a motor nerve cognitive system for the self-balancing two-wheeled robot has been built using the RBF neural network as the actor and evaluation function approximator. The simulation results show that the learning speed is increased as well as the failure times are reduced by the proposed method than by the Actor-Critic method with the only Basal Ganglia mechanism. Through decreasing temperature in the late stage, the learning speed is increased and the vibration disappeares eventually, and the learning effect is improved.
2012 Vol. 25 (1): 29-36 [Abstract] ( 511 ) [HTML 1KB] [ PDF 517KB] ( 928 )
37 Danger Theory Based Dynamic Constrained Immune Optimization
ZHANG Zhu-Hong
Based on the danger theory, an immune algorithm for dynamic constrained single-objective function optimization is proposed. The key of the algorithm is to construct two functional modules: environmental detection and co-evolution. Relying upon the Antigen Presenting Cells (APCs) infected by distressed or apoptotic cells, the change of the environment is detected and the environmental level is confirmed. The co-evolving scheme based on self-reactive, effective and environmental memory cells is explored. The proposed approach can online detect the change of the environment with the merits of simplicity, flexibility and dynamic runtime. The experimental results show that the proposed approach performs better than the compared algorithms, it has potential use for dynamic constrained optimization problems while achieves the reasonable balance between effect and efficiency.
2012 Vol. 25 (1): 37-44 [Abstract] ( 547 ) [HTML 1KB] [ PDF 525KB] ( 727 )
45 3D Face Recognition Based on Local Descriptor
XU Jun, DA Fei-Peng
A 3D face recognition algorithm based on a local descriptor is presented. The local feature of every sampling point is defined as the area projected by the points in the neighborhood whose size is determined by the angle of normal vector and 3 axes. The contour lines being equidistant with the nose tip are firstly extracted after the original face is preprocessed and normalized to the same posture. Then, the useful sampling points in the contour lines are selected by resampling and local features of these useful sampling points are extracted. Finally, the relationships of corresponding points between faces are established and the local features with weighted fusion are used for matching. Experimental results certify that the proposed method obtains better recognition rate and is robust to occlusion and spikes.
2012 Vol. 25 (1): 45-53 [Abstract] ( 620 ) [HTML 1KB] [ PDF 626KB] ( 18024 )
54 Properties of Fuzzy Morphological Bidirectional Associative Memories
ZENG Shui-Ling, XU Wei-Hong, YANG Jing-Yu
A learning algorithm is proposed for a class of fuzzy morphological bidirectional associative memories (FMBAM). It is proved theoretically that, for any given set of pattern pairs, if existing pairs of connection weight matrices which make the set become a set of the equilibrium states of FMBAM, the proposed learning algorithm can give the maximum of all such pairs of weight matrices. And the learning algorithm ensure that the FMBAM with this maximal pair of connection weight matrices can be convergent to an equilibrium state in one iterative process for any input. Any equilibrium state of FMBAM is Lyapunov stable. FMBAM can converge to equilbrium state for its any input vector. The robustness of FMBAM is good when the learning algorithm is used to train FMBAM and training pattern pairs have perturbations.
2012 Vol. 25 (1): 54-62 [Abstract] ( 680 ) [HTML 1KB] [ PDF 473KB] ( 455 )
63 Music Emotion Fuzzy Computing Model Based on Evolving Kernel Clustering
L Lan-Lan, ZHOU Chang-Le
Music emotion computing is a complex problem of emotion representation, which has multi-level and multi-dimensional structure. Its characteristics of fuzziness, subtleness and diversity result in the inefficiency of traditional methods. In order to improve recognition accuracy, firstly, the non-linear mapping of Gaussian radial basis function is used to identify, extract and magnify more details. Then, six key emotional features are extracted, by analyzing Chinese Guqin music in depth, and the fuzzy classification model for music emotion is constructed based on kernel clustering evolutionary algorithm. Moreover, aiming at the shortcoming of setting uniform cluster radius threshold in algorithm, the corresponding optimization strategy is proposed based on ACO. Finally, the optimized model is compared with Beyes classification model, and the experimental results show that the proposed method is effective.
2012 Vol. 25 (1): 63-70 [Abstract] ( 603 ) [HTML 1KB] [ PDF 477KB] ( 578 )
71 Review on Pedestrian Gait Feature Expression and Recognition
BEN Xian-Ye, XU Sen, WANG Ke-Jun
Various methods for gait recognition are summarized by unique metods from anthropometry, spatial temporal, kinematics, kinetics, and video stream data forms. And video stream data are analyzed in detail. The expressions, meanings and characteristics including static, dynamic, and time-varying information among different energy image species methods are compared. Moreover, the fusion of gait features, gait and other biometric and extended gait recognition are reviewed. The assessment methods for gait recognition research are also presented. In addition, the future research directions of gait recognition are addressed.
2012 Vol. 25 (1): 71-81 [Abstract] ( 866 ) [HTML 1KB] [ PDF 788KB] ( 1732 )
82 An Area Ratio between Rings Based Translation, Rotation and Scale Invariant Descriptor
WANG Xiao-Nian, QIU Li-Ke, CHENG Yu, JIANG Ping, ZHU Jin
A descriptor with rotation, translation, scaling and mirroring invariant is presented. The descriptor based on the spatial-partition of binary image adopts distributions of objects in each rings of concentric circles as features. Firstly, four invariants of the descriptor are proved, and then a gesture recognition experiment is demonstrated using this descriptor. The results of matching experiments show that the descriptor performs well under certain deformation of objects, the accuracy rate is higher than that of other algorithm, and the robustness of the feature is validated. In addition, as the feature is based on the concept of spatial-partition idea, it can be easily extended to other situations, such as Hu moments, energy of curve, even chain code and Fourier descriptors to form new features.
2012 Vol. 25 (1): 82-88 [Abstract] ( 469 ) [HTML 1KB] [ PDF 905KB] ( 884 )
96 Research on Computing Minimum Entropy Based Attribute Reduction via Stochastic Optimization Algorithms
MA Sheng-Lan, YE Dong-Yi
Existing heuristic attribute reduction algorithms generally fail to get a minimum entropy-based attribute reduction of a decision table. Some stochastic optimization algorithms are discussed to solve the problem of entropy-based attribute reduction. Firstly, a proper fitness function is defined to transform the minimum attribute reduction problem into a fitness optimization problem without additional constraints and the equivalence of transformation is proved. Then, the solving efficiency and the solution quality of some stochastic optimization algorithms are studied such as Genetic Algorithm, Particle Swarm Optimization, Tabu search and Ant Colony Optimization. Some UCI datasets are applied to test those performances. The experimental results show that the fully informed PSO based attribute reduction algorithm with refine scheme has a higher probability to find a minimum entropy-based attribute reduction and good performance.
2012 Vol. 25 (1): 96-104 [Abstract] ( 700 ) [HTML 1KB] [ PDF 523KB] ( 493 )
105 Factor Analysis for Language Identification Based on Phoneme Recognition
ZHONG Hai-Bing, SONG Yan, DAI Li-Rong
In the phoneme recognition based language identification system, the key issue is whether the tokens or the token sequence can reflect the language related information or not. However, it is observed that for certain utterance, the noise in the output token sequence from the phone recognizer is introduced due to the channel, speaker and background clutters. To address this problem, each utterance is represented in n-gram vector. And in this vector space, the factor analysis is applied to model the noise subspace, which will be reduced in final modeling process. The experiment results on NIST LRE 2007 show that the proposed method can outperform the existing phone recognition based language identification system. In 30s evaluation task, the equal error rate (EER) of recognition reduces relatively about 14.4% against the baseline phone recognition followed by language modeling (PRLM) system, while about 12.9% against the baseline phone recognition followed by support vector machine (PRSVM) system.
2012 Vol. 25 (1): 105-110 [Abstract] ( 706 ) [HTML 1KB] [ PDF 403KB] ( 561 )
111 Incremental Image Classification Method Based on Semi-Supervised Learning
LIANG Peng, LI Shao-Fa, QIN Jiang-Wei, LUO Jian-Gao
In order to use large numbers of unlabeled images effectively, an image classification method is proposed based on semi-supervised learning. The proposed method bridges a large amount of unlabeled images and limited numbers of labeled images by exploiting the common topics. The classification accuracy is improved by using the must-link constraint and cannot-link constraint of labeled images. The experimental results on Caltech-101 and 7-classes image dataset demonstrate that the classification accuracy improves about 10% by the proposed method. Furthermore, due to the present semi-supervised image classification methods lacking of incremental learning ability, an incremental implementation of our method is proposed. Comparing with non-incremental learning model in literature, the incremental learning method improves the computation efficiency of nearly 90%.
2012 Vol. 25 (1): 111-117 [Abstract] ( 805 ) [HTML 1KB] [ PDF 416KB] ( 703 )
118 Learning to Rank Based on Query Clustering
HUA Gui-Chun, ZHANG Min, LIU Yi-Qun, MA Shao-Ping, RU Li-Yun
Learning to rank,the interdisciplinary field of information retrieval and machine learning, draws increasing attention and lots of models are designed to optimize the ranking functions. However, few methods take the differences among the queries into account. In this paper,the queries are modeled as multivariate Gaussian distributions and Kullback-Leibler divergence is adopted as distance measure. The spectral clustering is applied to cluster the queries into several clusters and a ranking function is learned for each cluster.The experimental results show that the ranking functions with clustering are trained with less data,but are comparable to or even outperform the ones without clustering.
2012 Vol. 25 (1): 118-123 [Abstract] ( 770 ) [HTML 1KB] [ PDF 354KB] ( 630 )
124 A Multi-Label Learning Algorithm Based on Sparse Representation
SONG Xiang-Fa, JIAO Li-Cheng
To solve the problem of multi-label data classification, a multi-label learning algorithm based on sparse representation is proposed. The testing samples are treated as a sparse linear combination of training samples, and the sparsest coefficients are obtained by using l1-minimization. Then, the discriminating information of sparse coefficients is utilized to calculate membership function of the testing sample. Finally, the labels are ranked according to the membership function and the classification is completed. Extensive experiments are conducted on gene functional analysis, natural scene classification and web page categorization, and experimental results demonstrate the effectiveness of the proposed method. The results also show that the proposed method based on sparse representation achieves better results than other algorithms.
2012 Vol. 25 (1): 124-129 [Abstract] ( 722 ) [HTML 1KB] [ PDF 399KB] ( 824 )
130 Optimization of Decoding Thresholds Parameters in Continuous Speech Recognition
YIN Ming-Ming, LI Bi-Cheng, QU Dan, NIU Tong
As the current pruning thresholds can not take decoding speed and accuracy into account at the same time in continuous speech recognition, a joint optimization algorithm of multi-dimension pruning thresholds parameters is proposed. The pruning thresholds, including the main beam pruning, the word end pruning, the number of active modes and the tokens, are mainly studied in the proposed algorithm. The multi-objectives theory is adopted to optimize these parameters jointly. And then the strategy of segment-based dynamic thresholds pruning is introduced to deal with the results. The experimental results show that the performance of decoder is improved, the search space of decoding gets effective control, and the request of speed and accuracy can be satisfied.
2012 Vol. 25 (1): 130-135 [Abstract] ( 533 ) [HTML 1KB] [ PDF 436KB] ( 581 )
136 Simultaneous Dehazing and Denoising of Single Hazing Image
FANG Shuai, WANG Feng, ZHAN Ji Qing, CAO Yang, YUAN Hong Wu, RAO Rui Zhong

Various noise exists in images in practice, which brings great influence on dehazing results. Aiming at this, single image dehazing algorithm is proposed which can realize simultaneous dehazing and denoising based on joint bilateral filter. Firstly, the initial rough transmission map is estimated based on dark prior. Then, a joint bilateral filter is applied to refine the rough transmission map under the guidance of original image, which decreases the halo artifacts in the dehazing image effectively. Next, another bilateral filter is applied to obtain the dehzaing image, which can realize image denoising at the same time. Finally, a color factor is introduced into the bilateral filtering process to deal with the color distort problem. Various contrastive experimental results verify that the proposed algorithm realizes single image dehazing and denoising simultaneously with low computational costs. Besides, the color factor brings abundant chromatic details in the dehazing results.
2012 Vol. 25 (1): 136-142 [Abstract] ( 368 ) [HTML 1KB] [ PDF 757KB] ( 771 )
143 Supervised Locality Preserving Canonical Correlation Analysis Algorithm
HOU Shu-Dong, SUN Quan-Sen, XIA De-Shen
To use locality preserving canonical correlation analysis (LPCCA) in pattern classification and acquire fine results, a supervised locality preserving canonical correlation analysis (SLPCCA) is proposed based on LPCCA incorporated the class label information. Through maximizing the weighted correlation between corresponding samples and their near neighbors belonging to the same classes, SLPCCA effectively utilizes the class label information and preserves the local manifold structure of the data. In addition, the proposed algorithm effectively fuses the discrimination information of DCCA without the restriction of total class numbers. Besides, a kernel SLPCCA (KSLPCCA) is also proposed based on kernel methods to extract nonlinear features of the data. The experimental results on ORL, Yale, AR and FERET face databases show that the proposed algorithms are better than related canonical correlation analysis methods.
2012 Vol. 25 (1): 143-149 [Abstract] ( 431 ) [HTML 1KB] [ PDF 473KB] ( 507 )
150 Counting Pedestrains in Video SequencesBased on Non-Maxima Suppression Clustering
L Ji-Min, Zeng Zhao-Xian, Zhang Mao-Jun
Based on the background image of a fixed scene, a four-step approach to count predestrains in video sequences is presented, and the estimation result of long-range crowds is improved compared with D.Conte’s solution in 2010 EURASIP Journal. Our primary contribution lies in non-maxima suppression clustering. The proposed density-based clustering approach applies different clustering standards to crowds at different distances from camera, hence it avoids overlarge clusters and ensuing problems. Experiments on PETS 2010 database show estimation result of long-range crowds is improved significantly, as an implicit result of smaller clusters from Non-maxima Suppression Clustering.
2012 Vol. 25 (1): 150-156 [Abstract] ( 289 ) [HTML 1KB] [ PDF 448KB] ( 565 )
157 An Algorithm for Mining Complement-Alternative Relationship Based on Frequent Itemsets
CHAI Yu-Mei, WANG Chun-Li,WANG Li-Ming
Based on the TOP-k-Closed Miner algorithm, an algorithm for mining frequent itemsets based on index (Index-FIM) is proposed. Bit-vector is used to represent dataset, the breadth expansion pruning strategy and region-index skimming strategy are simultaneously introduced. The experimental results show that Index-FIM has high execution efficiency for mining frequent itemsets in sparse datasets. In order to acquire useful information for direct prediction, another algorithm for complement-alternative relationship mining based on frequent itemsets (CARM) is proposed. The relevant coefficient between the items included in frequent itemsets is computed to get their complement-alternative relationship, and the complementary-alternative relationship is represented by complement-alternative relationship graph (CAG), which is convenient for the decision maker to make reasonable and accurate judgment. Experimental results show that CAG is more efficient and precise than frequent itemsets in expressing information.
2012 Vol. 25 (1): 157-165 [Abstract] ( 335 ) [HTML 1KB] [ PDF 534KB] ( 983 )
166 A Robust Feature Parameter Extraction Algorithm for Language Identification
HUANG Shan-Qi, ZHANG Ling-Hai, QU Dan
In current language identification system, the commonly used feature parameters have not made the best use of auditory characteristics and have weak robustness in complex environments. An auditory-based robust feature extraction algorithm is proposed. Each sub-band energy of the extracted auditory features is calculated by using a Gammachirp filter bank instead of the commonly used triangle filter bank. The compensation filter using data-driven analysis for each sub-band output is obtained by a constrained optimization process which jointly minimizes the environmental distortion as well as the distortion caused by the filter itself. Experimental results show that the feature outperforms the Mel-frequency cepstral coefficient widely used in noisy environments.
2012 Vol. 25 (1): 166-171 [Abstract] ( 287 ) [HTML 1KB] [ PDF 380KB] ( 513 )
172 Cognitive Model Based on Single Unit Event
FENG Kang, YAO Nan-Sheng
To find the cognitive principle through a cognitive model, a cognitive model based on single unit event is designed. The cognitive object of the model is cognition. The cognitive structure of the model consists of sensing part, accepting part, storing part, problem-solving part, thinking part and reasoning part. In the same testing cognitive process, the simulation of cognitive process is tested, when the capacity of the memory bank is given different values, the changes of the cognition stored in memory bank are recorded, and the cognitive data are calculated. The results show that the cognitive model is able to simulate cognitive process. As the cognitive process goes on, the cognition dimensions, the cognition quantities, the cognition multiplicities and the location of cognition in the memory bank often alter. All the cognitive data increase when the capacity of the memory bank enlarges and the data are modified when the problems are solved.
2012 Vol. 25 (1): 172-180 [Abstract] ( 297 ) [HTML 1KB] [ PDF 581KB] ( 814 )
模式识别与人工智能
 

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