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....
 
 
2014 Vol.27 Issue.9, Published 2014-09-30
Papers and Reports    Researches and Applications   
   
Papers and Reports
769 Human Pose Estimation Based on Fusion of HOG and Color Feature
HAN Gui-Jin, ZHU Hong
Appearance model plays an important role in the human pose estimation. To improve the estimation accuracy, how to set up the appearance model by using the histogram of oriented gradient (HOG) and color features is studied. The sub-classifier for each cell unit of the body part is built by using the support vector data description (SVDD) algorithm, and then the HOG-based appearance model is constructed by the linear combination of sub-classifiers with different weights. The specific location probability is learned by using those states, which has higher similarity with the HOG-based appearance model, and the corresponding color histogram is calculated, which is the color-based appearance model. According to the illumination and the color contrast between the clothing and background in the static image to be proceeded, the weights of two appearance models are determined, and then two appearance models are combined linearly to construct appearance model based on the fusion of the HOG and color features. The proposed appearance model is used for human pose estimation, and the experimental results show it is more effectively and gets higher pose estimation accuracy.
2014 Vol. 27 (9): 769-777 [Abstract] ( 462 ) [HTML 1KB] [ PDF 921KB] ( 19275 )
778 Calculation Analysis and Attribute Reduction for Double-Quantitative Rough Set Model Based on Logical OR
ZHANG Xian-Yong, MIAO Duo-Qian
Double quantification has a fundamental function for completely describing approximate space in rough set, and the rough set model based on logical OR of precision and grade is a basic extended model with double quantification. Aiming at this model, calculation analysis is mainly conducted, and attribute reduction in approximate space is further explored. Based on both basic structures and calculation formulas of model regions, macroscopic algorithm and structural algorithm are constructed. The analysis and comparison results show that the structural algorithm has more advantages in calculation complexity. Based on approximate space, basic properties on four-region preservation are discussed, and attribute reduction with the region preservation criterion is proposed and investigated. In particular, a type of extended quantitative reduction is obtained for the classical qualitative reduction. Some generalized thoughts are provided for optimal calculations and reduction applications of double-quantitative rough set models.
2014 Vol. 27 (9): 778-786 [Abstract] ( 415 ) [HTML 1KB] [ PDF 411KB] ( 640 )
787 Super-Resolution Image Reconstruction Based on Local Structural Similarity and Collaborative Representation
CAI Miao-Miao, TAN Yuan-Peng, CAO Fei-Long
An approach for super-resolution image reconstruction is presented based on local structural similarity and collaborative representation. The collaborative representation l2-norm regularization and local similarity constraint are employed to seek a linear combination for a patch of low-resolution input image with respect to the low-resolution dictionary. Then, the high-resolution image patch is generated by virtue of the coefficients of this combination and the high-resolution dictionary. In addition, the l2-norm based objective function implies an analytical solution and it does not involve local minima. Hence, it performs at a lower complexity compared to l1-sparsity constraint model. The experimental results demonstrate that the proposed method is feasible and effective for small super-resolution image reconstruction and outperforms the bicubic interpolation method and sparse representation super-resolution model on both visual effect and numerical results. Further research shows that the proposed method also performs well for large magnification factors and noisy data.
2014 Vol. 27 (9): 787-793 [Abstract] ( 381 ) [HTML 1KB] [ PDF 789KB] ( 762 )
794 Ensemble Incomplete Wavelet Packet Subspaces for Face Recognition Based on Fuzzy Integral
ZHAI Jun-Hai, WANG Xi-Zhao, ZHANG Su-Fang
An ensemble incomplete wavelet packet subspaces method based on fuzzy integral for face recognition is proposed, and it is compared with 5 related approaches. Firstly, face images are decomposed into different sub-images with incomplete wavelet packet transform. For sub-images with low frequency information in two directions, features are extracted directly. And for high frequency sub-images with low frequency information in one direction, features are extracted after these images are averaged. Next, fuzzy classifiers are trained by the obtained wavelet subspace images. Finally, the trained classifiers are integrated by fuzzy integral. The proposed method makes full use of the information provided by sub-images with different frequency and improves the accuracy of face recognition. The experimental results on ORL, YALE, JAFFE and FERET show that the proposed method has higher accuracy than 5 related approaches.
2014 Vol. 27 (9): 794-801 [Abstract] ( 405 ) [HTML 1KB] [ PDF 806KB] ( 587 )
802 Worst Separation Spatially Smooth Discriminant Analysis with Constrained Average Compactness
NIU Lu-Lu, CHEN Song-Can, YU Lu
Spatially Smooth Linear Discriminant Analysis(SLDA) and IMage Euclidean Distance Discriminant Analysis(IMEDA)combined with spatial structure information of the imagesare two main discriminant methods to reduce dimension, and the classification performance of SLDA and IMEDA is better than that of LDA. Different from SLDA and IMEDA, the solutions in the proposed algorithms called WSLDA and WIMEDAare obtained by parameterizing projection directions, maintaining an upper bound for average within-class scatter and maximizing the minimal between-class scatter.Also their solution can simply be attributed to solve a well-known eigenvalue optimization problem called minimization for the maximal eigenvalue of a symmetric matrix. It overcomes the shortcoming that many algorithms need to use full eigenvalue decomposition. In addition, experiments on standard face dataset Yale、AR and FERET validate the effectiveness of WSLDA and WIMEDA.
2014 Vol. 27 (9): 802-807 [Abstract] ( 329 ) [HTML 1KB] [ PDF 454KB] ( 486 )
808 High-Order Adaptive Super-Twisting Sliding Mode Control forUncertain Underactuated Systems
YANG Xing-Ming, GAO Yin-Ping
To achieve good robustness against disturbances for a class of uncertain underactuated systems, a second-order adaptive sliding mode control method is proposed based on quadratic Lyapunov function to reduce the inherent chattering of conventional sliding mode control (SMC). Firstly, a second-order super-twisting algorithm is used by the discontinuous part of controller, which acts on the second-order derivative of sliding mode variables. Secondly, as for the effects of unknown disturbances on sliding mode surface, an adaptive law is designed to adjust the parameters. This method eliminates the restriction of the first derivative of disturbances boundary being known in the traditional second-order sliding mode control, which not only keeps convergence of sliding mode surface but also reduces chattering. Finally, a two-wheeled self-balancing cart is used to test the proposed approach. The simulation results show that compared with conventional SMC and ordinary second-order SMC, the proposed method outperforms the above methods on effectiveness and reducing chattering.
2014 Vol. 27 (9): 808-814 [Abstract] ( 571 ) [HTML 1KB] [ PDF 503KB] ( 587 )
Researches and Applications
815 An Enhanced Hybrid Genetic Simulated Annealing Algorithm for VLSI Standard Cell Placement
CHEN Xiong-Feng, WU Jing-Lan, ZHU Wen-Xing
A hybrid genetic simulated annealing algorithm is presented for solving the problem of VLSI standard cell placement with up to millions of cells. Firstly, to make genetic algorithm be capable of handling very large scale of standard cell placement, the strategies of small size population, dynamic updating population, and crossover localization are adopted, and the global search and local search of genetic algorithm are coordinated. Then, by introducing hill climbing (HC) and simulated annealing (SA) into the framework of genetic algorithm and the internal procedure of its operators, an effective crossover operator named Net Cycle Crossover and local search algorithms for the placement problem are designed to further improve the evolutionary efficiency of the algorithm and the quality of its placement results. In the algorithm procedure, HC method and SA method focus on array placement and non-array placement respectively. The experimental results on Peko suite3, Peko suite4 and ISPD04 benchmark circuits show that the proposed algorithm can handle array and non-array placements with 10,000~1,600,000 cells and 10,000~210,000 cells respectively, and can effectively improve the quality of placement results in a reasonable running time.
2014 Vol. 27 (9): 815-825 [Abstract] ( 439 ) [HTML 1KB] [ PDF 572KB] ( 609 )
826 Emotion Reasoning Algorithm Based on Emotional Context of Speech
MAO Qi-Rong, BAI Li-Juan, WANG Li, ZHAN Yong-Zhao
Since the change of emotion state is continuous in daily conversations, an emotion reasoning algorithm based on emotional context for speech emotion recognition (SER) is put forward. In this algorithm, contextual speech emotion features and widely-used acoustic speech emotion features are used to recognize emotion state in continuous speech utterances respectively. Then, the emotional interaction matrix and the confidence coefficient are used to fuse the recognition results of these two kinds of features.Finally, the emotion reasoning rule based on the emotional context is proposed to adjust the fusion results according to the emotional context of the emotional utterance to be analyzed. The fusion results after adjusting are used as the emotion state of the emotional utterance to be analyzed.The experimental results on the recorded emotional speech corpus with respect to 6 basic emotion states show that the proposed algorithm can improve the emotion recognition accuracies of the continuous speech, and compared with the method by widely-used acoustic speech emotional features, the average recognition accuracy of the proposed algorithm rises by 12.17%.
2014 Vol. 27 (9): 826-834 [Abstract] ( 480 ) [HTML 1KB] [ PDF 879KB] ( 627 )
835 An Fingerprint Matching Algorithm Based on Minutia Global Confidence
FU Xiang, FENG Ju-Fu
The local structural similarity is used in traditional minutia-based fingerprint matching methods to describe the potential associations of each minutia pair.The concept of minutia global confidence is proposed to define the geometric consistency and global matching possibility between one minutia pair and all the other candidate pairs. It can be seen as a supplement to local structural similarity. The global confidence of each minutia pair is acquired by calculating the principal eigenvector of the pairwise compatibility matrix and using spectral relaxation techniques. The correlation matrix can be constructed by using large local structural similarity and large global confidence. Minutia pairs with large local structural similarity and large global confidence are judged to be matched. The proposed approach utilizes the information of local topology and global compatibitity well and has better robustness. The experiments on FVC 2002 and 2004 databases demonstrate its effectiveness and efficiency.
2014 Vol. 27 (9): 835-840 [Abstract] ( 425 ) [HTML 1KB] [ PDF 501KB] ( 1091 )
841 Kernelized Fuzzy C-Means Clustering Algorithm Based on Hybrid Ant Colony Optimization for Continuous Domains
GUO Xiao-Fang, LI Feng, SONG Xiao-Ning, WANG Wei-Dong
To further improve the clustering performance of kernelized fuzzy C-means clustering algorithm, a kernelized fuzzy C-means clustering algorithm based on hybrid ant colony optimization of continuous domain (KFCM-HACO) is proposed. Kernel function parameters value of KFCM algorithm is optimized by HACO, which overcomes the shortcomings of traditional algorithm, minimizes the objective function of kernelized fuzzy clustering algorithm, and speeds up the convergence rate of the algorithm. The simulation and comparison results on UCI dataset show that the KFCM-HACO algorithm outperforms the traditional clustering algorithm and improves the accuracy of clustering.
2014 Vol. 27 (9): 841-846 [Abstract] ( 366 ) [HTML 1KB] [ PDF 424KB] ( 583 )
847 Research on Selective Clustering Ensemble Algorithm Based onNormalized Mutual Information and Fractal Dimension
WU Xiao-Xuan, NI Zhi-Wei, NI Li-Ping, ZHANG Chen
Traditional clustering ensemble algorithm can not eliminate the influence of inferior quality clustering members and is also characterized with lower clustering accuracy. To solve the problems, a selective clustering ensemble algorithm based on fractal dimension is proposed. Firstly, the proposed algorithm is used to realize incremental clustering and can find arbitrary shape clustering. Then, according to the selection strategy of weight values based on normalized mutual information, the proposed algorithm selects high quality clustering members to realize integration by using weighted co-association matrix and get the final clustering results. The experimental results show that compared to the traditional clustering ensemble algorithm, the proposed algorithm improves the clustering quality and has good extensibility.
2014 Vol. 27 (9): 847-855 [Abstract] ( 396 ) [HTML 1KB] [ PDF 751KB] ( 1032 )
856 Spectral Clustering Based on Local Density Estimation and Neighbor Propagation
GE Hong-Wei, LI Zhi-Wei, YANG Jin-Long

Neighbor propagation based spectral clustering can be used to cluster the dataset with inhomogeneous density. However, sometimes it propagates different clustering samples into the same subset with high similarity, which can not obtain the real similarity matrix and accurate clustering results. To solve this problem, a local density estimation and neighbor propagation based spectral clustering algorithm (LDENP-SC) is proposed. In this algorithm, the local density of the samples is firstly estimated and the dimensions of the datasets are increased. Then, the similarity matrix is updated by using neighbor propagation and the new dataset is clustered by spectral clustering. Also, a simple local density estimation method is proposed by with the local density of the samples can be estimated accurately and fast. Moreover, based on propagation algorithm, a method for updating the similarity of the samples in different subsets is adopted to get more actual similarity matrix. The experimental results show that LDENP-SC algorithm can obtain similarity matrix close to the ideal and accurate clustering results, has good generalization ability and is robust to a certain range ofparameter σ.
2014 Vol. 27 (9): 856-864 [Abstract] ( 392 ) [HTML 1KB] [ PDF 1993KB] ( 678 )
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