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.8, Published 2014-08-30
Papers and Reports    Researches and Applications    Surveys and Reviews   
   
Papers and Reports
673 Tag Recommendation Method Considering Users Tagging Status
YU Hong, DENG Ming-Yao, HU Feng
To improve the quality of tag recommendation, a tag recommendation method considering users current tagging status is proposed. Firstly, the statistical analysis shows the total number of tags used by a user is changed with time in a social tagging system. Then, three tagging statuses are defined, i.e. the growing status, the mature status and the dormant status, and a user current tagging status is one of the above statuses. Finally, according to the characteristics of the current tagging status, different strategies are developed to compute the tag probability distribution to recommend tags to users. Results of comparative experiments show that the proposed method has better accuracy of tag recommendation.
2014 Vol. 27 (8): 673-682 [Abstract] ( 390 ) [HTML 1KB] [ PDF 471KB] ( 986 )
683 U-Chord Curvature: A Computational Method of Discrete Curvature
GUO Juan-Juan, ZHONG Bao-Jiang
Discrete curvature computation of digital curves is widely applied to various tasks of image analysis and computer vision. A computational method of discrete curvature, U-chord curvature, is proposed. For each point in a digital curve, its support region is determined by two points with a given chord distance to the point, and then the U-chord curvature of the point is estimated. A theoretical analysis shows that there is a close relationship between the U-chord curvature and the real curvature of the curve. Compared with the existing computational methods of discrete curvature, the U-chord curvature is more stable under rotation transformations and noise condition. Therefore, it is suitable for image and vision tasks which require a high stability of curvature estimation, such as curve matching. Simulation experiments show the efficiency of the proposed method.
2014 Vol. 27 (8): 683-691 [Abstract] ( 652 ) [HTML 1KB] [ PDF 820KB] ( 1648 )
692 Optimized Cutting Plane Method for Linear SVM via Inexact Step-Length Search
CHU De-Jun, TAO An, GAO Qian-Kun, JIANG Ji-Yuan, TAO Qing
Cutting plane method efficiently solves the primal problem of linear support vector machines by adding cutting planes incrementally, and thus it can be accelerated through the exact line search. In this paper, an optimized cutting plane method with inexact line search is presented, and it determines the interval containing the optimal step size with fewer iterations. The acceptable step size is obtained by the closed-form solution of quadratic interpolation with two points. The theoretical analysis shows that the proposed method has the same optimal convergence bound as the exact line search method with a higher speed and low cost. The experiments on large-scale datasets demonstrate that the proposed method outperforms the exact line search method. In some cases, it achieves even more than 50% speedup.
2014 Vol. 27 (8): 692-700 [Abstract] ( 444 ) [HTML 1KB] [ PDF 721KB] ( 1090 )
701 Decision-Theoretic Rough Set Attribute Reduction and Classification Based on Fuzzification
GUO Min, JIA Xiu-Yi, SHANG Lin
The decision-theoretic rough set (DTRS) is a kind of probabilistic rough set model with certain tolerance based on the Bayesian risk minimization principle. However, the current research on DTRS model is restricted to processing information tables with discrete data. In this paper, the decision-theoretic rough set theory is combined with fuzzy sets, and the fuzzy membership functions are employed to replace the posterior probability calculating method when calculating the expected risk losses in the DTRS model. Thus, the new decision rules are derived to effectively deal with the information system with continuous data. Experiments show that the proposed method is feasible and it has a better classification performance by adjusting the membership functions.
2014 Vol. 27 (8): 701-707 [Abstract] ( 369 ) [HTML 1KB] [ PDF 344KB] ( 2130 )
708 An Occluded Facial Expression Recognition Method Based on Sparse Representation
ZHU Ming-Han, LI Shu-Tao, YE Hua
Occlusion dictionary does not have redundancy and facial expression classification is easily disturbed by identity features, which sparse representation based classification(SRC) is used to recognize occluded facial expression. A method for occluded facial expression recognition is proposed to solve this problem. Firstly, an occlusion dictionary with redundancy is constructed by multilevel blocking of the image. Next, sparse representation coefficients of the test image are gained by spare decomposition. Finally, the expression category of test image is judged in its individual subspace. The proposed method makes decomposition coefficients of the test image sparser and avoids identity feature interference to expression classification. The experimental results on Cohn-Kanade and JAFFE face databases show that the proposed method is robust to occluded facial expression recognition.
2014 Vol. 27 (8): 708-712 [Abstract] ( 516 ) [HTML 1KB] [ PDF 680KB] ( 998 )
713 KFLD-SIFT with RVM Fuzzy Integral Fusion Recognition of Human Action Based on Tensor
XIAO Di, NAN Lei-Guang
Due to the large sample and multiple characteristics of video sequence in the field of human action recognition, a method of kernel Fisher nonlinear discriminant (KFLD) - scale invariant feature transform (SIFT) and relevance vector machine (RVM) fuzzy integral fusion recognition based on tensor is proposed. Firstly, video sequence is pre-processed into binary video sequence, and then it is described as third-order tensor. Furthermore, as for large sample characteristics, a local feature extraction method of KFLD-SIFT is proposed to reduce the dimension around the key points under different initial scales. Meanwhile, RVM fuzzy integral fusion algorithm for behavior classification is presented. Finally, the proposed method and other relevant methods are compared through four kinds of evolution indexes and average recognition rates. The video sequence of KTH human action database and triple-cross verification method are used to test the recognition methods. Experimental results show that the proposed method achieves good recognition effect, and its average recognition rate rises by at least 2.3% compared to other mainstream methods for human action recognition.
2014 Vol. 27 (8): 713-719 [Abstract] ( 376 ) [HTML 1KB] [ PDF 466KB] ( 893 )
Surveys and Reviews
720 Survey of Recommendation Based on Collaborative Filtering
LENG Ya-Jun, LU Qing, LIANG Chang-Yong
Collaborative filtering is a widely used technique in recommender systems. Extensive studies are carried out on collaborative filtering. However, systematic summary of this field is scarce. In this paper, research of collaborative filtering is reviewed. The meaning and key issues of collaborative filtering, including sparsity, multiple-content and scalability, are described firstly, and then the solutions to the above key issues are introduced in detail. Finally, the future work of collaborative filtering is pointed out. The knowledge framework of collaborative filtering is introduced. It makes the research clues of collaborative filtering clear, provides a reference to other scholars, and improves the performance of personalized information services.
2014 Vol. 27 (8): 720-734 [Abstract] ( 556 ) [HTML 1KB] [ PDF 698KB] ( 3357 )
Researches and Applications
735 Image Set Matching Based on Support Vector Domain Description
ZENG Qing-Song
An image set matching method based on support vector domain description is proposed. Firstly, each image set from the original input space is mapped into the high dimensional feature space by support vector machine learning, and then they are modeled using support vector domain description. In feature space, the model is described by a smallest enclosing ball, which encloses the most of the mapped data. Next, by introducing an efficient similarity metric based on support vector domain, the distance between two image sets is converted to the distance between pairwise support vector domains. Finally, the proposed method is evaluated on face recognition and object classification tasks based on datasets. Experimental results show that the proposed method outperforms other state-of-the-art set based matching methods. The recognition rates of the proposed method reaches 96.37%, 100% and 95.32% on ETH80 object database, HondaUCSD and YouTube video databases, respectively.
2014 Vol. 27 (8): 735-740 [Abstract] ( 437 ) [HTML 1KB] [ PDF 421KB] ( 935 )
741 Chinese Sign Language Recognition Method Based on Depth Image Information and SURF-BoW
YANG Quan, PENG Jin-Ye
To realize the accurate recognition of sign language in the video, an algorithm based on depth image CamShift(DI_CamShift) and speeded up robust features-bag of words (SURF-BoW) is proposed. Kinect is used as the sign language video capture device to obtain both of the color video and depth image information of sign language gestures. Firstly, spindle direction angle and mass center position of the depth images are calculated and the search window is adjusted to track gesture. Next, an OTSU algorithm based on depth integral image is used for gesture segmentation, and the SURF features are extracted. Finally, SURF-BoW is built as the feature of sign language and SVM is utilized for recognition. The best recognition rate of single manual alphabet reaches 99.37%, and the average recognition rate is up to 96.24%.
2014 Vol. 27 (8): 741-749 [Abstract] ( 486 ) [HTML 1KB] [ PDF 956KB] ( 1012 )
750 Natural Images Classification and Retrieval Based on Improved SDA
XU Shou-Jing, HAN Li-Xin, ZENG Xiao-Qin
Stacked denoising auto-encoder(SDA) is introduced into the image recognition. Convolutional auto-encoder (CAE) is used to improve SDA in the area of natural image retrieval. The unsupervised greedy layer-wise training algorithm is used to initialize the weight of the network. The parameters of the network are optimized by the back propagation algorithm. The improved SDA is trained for extracting features from natural images and the softmax classifier is used for classification. Finally, the extracted feature combined with scale invariant feature transform (SIFT) is used for realizing images retrieval. The experimental results show that the improved stacked denoising auto-encoder(ISDA) method can greatly reduce the time of network training, enhance the fault-tolerant ability of network, raise the classification precision of the classifier and eventually improve the image retrieval performance.
2014 Vol. 27 (8): 750-757 [Abstract] ( 354 ) [HTML 1KB] [ PDF 742KB] ( 1130 )
758 Structured Information Extraction Based on Pattern Matching
SHAO Kun, YANG Chun-Lei, QIAN Li-Bin, FANG Shuai

The information extraction results extracted from the semi-structured texts are coarse-grained, which results in ineffective semantic analysis. A structured information extraction method based on pattern matching is proposed. The proposed method is targeted at the web-presented semi-structured texts, and the suitable lexicon is loaded through domain recognition of the coarse-grained extraction results. Roles are mapped to the corresponding words in the word sequence according to the part of speech of the role in the patterns. Thus, the structured information can be extracted and it provides support for the accurate semantic analysis. Experiments show more accurate extraction results can be achieved by the proposed method.
2014 Vol. 27 (8): 758-768 [Abstract] ( 473 ) [HTML 1KB] [ PDF 711KB] ( 1196 )
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