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....
 
 
2015 Vol.28 Issue.2, Published 2015-02-28
Papers and Reports    Researches and Applications   
   
Papers and Reports
97 Three Kinds of Fuzzy Modus Tollens and Algorithms Based on Different Fuzzy Negations
PAN Zheng-Hua, ZHAO Jie-Xin, WANG Shan-Shan

Fuzzy modus tollens (FMT) is one of the most basic inference forms in fuzzy reasoning. As a premise of FMT, fuzzy negation plays a major role in inference. In this paper, based on the fuzzy propositional logic with contradictory negation, opposite negation and medium negation (FLCOM), contradictory negation, opposite negation and medium negation are proved to be different fuzzy negations, and thus three kinds of fuzzy modus tollens, FMT1, FMT2 and FMT3, are proposed,which are different from FMT respectively based on contradictory negation, opposite negation and medium negation. Furthermore, based on the R-implication operator IR, a fuzzy NR-implication operator INR is defined which is associated with IR. The algorithms of FMT1, FMT2 and FMT3 are proposed according to the algorithm of FMT, and it is proved that algorithms of FMT1, FMT2 and FMT3 are reducible algorithms when I ≤INR.
2015 Vol. 28 (2): 97-104 [Abstract] ( 613 ) [HTML 1KB] [ PDF 362KB] ( 598 )
105 Object Tracking under Constraint of Adaptive Structure-Preserving
CHEN Chen-Shu, ZHANG Jun, XIE Zhao, GAO Jun
With structure preserving property in tracking, a representation method for object structural appearance is proposed. In the proposed method, regional nodes are built to describe the local property of the object. And several soft/hard constraints are defined on regional nodes so that local and global properties of the object and the spatial structure of regional nodes are unified and described by the object structural representation. During the tracking procedure, state of objects can be roughly estimated by matching scale-invariant feature transform (SIFT) flow of local regions between successive frames. Then, through soft/hard constraints on regional nodes, the tracking result can be adaptively adjusted, which is called adaptive structure-preserving (ASP). Experimental results show that ASP performs better than other methods in tracking objects with deformation, shadows and illumination changes. Furthermore, ASP shows robustness and good generalization ability when the resolution of video sequences is low and the object is similar to the background.
2015 Vol. 28 (2): 105-115 [Abstract] ( 496 ) [HTML 1KB] [ PDF 2978KB] ( 528 )
116 Consistent Amendment Algorithm for Inconsistent Ordered Decision System
WENG Shi-Zhou, L Yue-Jin
Firstly, the concept of inconsistent degree is proposed to measure the inconsistent degree of the ordered decision system. By using decision information entropy of decision attribute and inconsistent degree as heuristic information, the consistent amendment algorithm is proposed. The decision information entropy is used as the first heuristic information to ensure the reasonable distribution of amendment results as consistent as possible with the distribution of the original data. The inconsistent degree is used as the second heuristic information to make the algorithm converge quickly and the inconsistent degree reduce to zero at a high speed. The feasibility and the effectiveness of the algorithm is demonstrated by examples and the simulation analysis. Finally, by the comparative analyses of the decision rules derived from the original data and the coordination data, the results show that the modified decision rule is more reasonable.
2015 Vol. 28 (2): 116-124 [Abstract] ( 449 ) [HTML 1KB] [ PDF 415KB] ( 394 )
125 Multirecord Webpage Extraction Based on DOM Tree Hierarchical Feature
CHEN Qiao-Ling, LIAO Xiang-Wen, WEI Jing-Jing, CHEN Guo-Long
The existing multirecord webpage extraction methods usually make overall longitudinal analyses of the document object model (DOM) tree. The computional structural similarity is always low, and therefore record regions can not be identified correctly. Different from the previous work, a method named data record extraction based on DOM tree hierarchical feature (DEBHF) is proposed to make transverse analyses of the DOM tree by distinguishing different roles of nodes at different levels. Thus, the problem of searching similar sub-trees is converted into the problem of searching similar sub-blocks in data blocks. Finally, the two-way search for non-overlapped and repeated sub-blocks is adopted to segment the record regions. Experimental results show that the proposed approach can deal with webpages which can not be obtained by the existing methods and the extraction results of different data sources demonstrate its effectiveness.
2015 Vol. 28 (2): 125-131 [Abstract] ( 440 ) [HTML 1KB] [ PDF 457KB] ( 761 )
132 Fractal Mutation Factor Correcting Differential Evolution Algorithm
QIU Xiao-Hong, JIANG Yang, LI Bo
To get better solution of the differential evolution (DE) algorithm, the mutation strategy of DE is proposed and divided into two parts to reflect the changes of the target population trends and their random variation. Fractal mutation factor differential evolution (FMDE) algorithm is put forward and it consists of an additional mutation factor simulated by a different Hurst index fractal Brownian motion. FMDE is tested on 25 benchmark functions presented at 2005 IEEE congress on evolutionary computation. The optimization results of at least 10 benchmark functions are better than the results obtained by other differential evolution algorithms, and the rest of the test results are approximate. Experimental results show that FMDE significantly improves the accuracy and adaptability of the optimization.
2015 Vol. 28 (2): 132-138 [Abstract] ( 402 ) [HTML 1KB] [ PDF 640KB] ( 556 )
139 Recoverable Watermarking Algorithm for Text Authentication and Synonym Replacement Based on Hopfield Neural Network
WANG Jing, TANG Xiang-Hong, LIN Xin-Jian
Aiming at problems of tamper detection and recovery, a recoverable watermarking algorithm for text authentication and synonym replacement based on synonym replacement technology and associative memory function of Hopfield neural network is proposed. The text is divided into replaceable synonyms and non-replaceable words. The feature information of replaceable synonyms is extracted according to the position in their thesaurus and the feature information of non-replaceable words is extracted according to the structure and stroke of Chinese characters in the text. Then, the watermark is embedded by synonym replacement. The information of watermark and the feature information of non-replaceable words are input into the Hopfield neural network and they are trained to realize tamper detection and recovery function of replaceable synonyms. The simulation results show that the proposed algorithm has good robustness, tamper detection performance and recoverability, and by this algorithm, tamper detection and the location of replaceable synonyms and non-replaceable words are implemented to realize text authentication, recover the original replaceable synonyms and achieve recovery.
2015 Vol. 28 (2): 139-147 [Abstract] ( 425 ) [HTML 1KB] [ PDF 1314KB] ( 590 )
Researches and Applications
148 A System Identification Method for Unmanned Helicopter Yaw Channel Based on Switch Model
SHEN Hui, FANG Yong-Chun, SUN Xiu-Yun, LIANG Xiao
Aiming at the saturation nonlinearity in the yaw channel of a small-scale unmanned helicopter, a switch model based identification method for the unmanned helicopter yaw channel is proposed. Firstly, the characteristics of the yaw channel are analyzed and a switch model including a saturation unit for the controller is established. Experimental data for the model identification are collected by utilizing some sweep frequency control signals. Based on the data, the parameters of the switch model are identified through the genetic algorithm method. The residual error signal is calculated, analyzed and utilized to modify the model obtained, and thus the accuracy of the model is further increased. The effectiveness of the proposed method is validated by some flight data. By utilizing the obtained model, a flight controller is designed and then employed to control the helicopter. The experimental result further demonstrates the validity of the indentified yaw model.
2015 Vol. 28 (2): 148-154 [Abstract] ( 720 ) [HTML 1KB] [ PDF 1054KB] ( 635 )
155 Dynamic Match Lattice Spotting Integrated with Posterior Probability Confidence Measure
ZHENG Yong-Jun, ZHANG Lian-Hai, CHEN Bin
In the keyword spotting system based on dynamic match lattice spotting (DMLS), the minimum edit distance is used as the confidence measure. When the detection rate is increased, the false alarm rate is raised as well. Aiming at this problem, an approach integrating the posterior probability confidence measure with DMLS is proposed. Firstly, the posterior probability based on lattice is introduced with the index stage of DMLS. Secondly, data driven phone substitution, insertion and deletion costs are incorporated for more flexible phone sequence matching. Finally, the minimum edit distance and the posterior probability confidence measure are blended together to detect all occurrences of the keywords. The experimental results show that there is a certain complementarity between the minimum edit distance and posterior probability confidence measure and the equal error rate is relatively reduced.
2015 Vol. 28 (2): 155-161 [Abstract] ( 407 ) [HTML 1KB] [ PDF 459KB] ( 657 )
162 Network Structure-Enhanced Extremal Optimization Based Semi-supervised Algorithm for Community Detection
DU Mei, HU Xue-Gang, LI Lei, HE Wei
Community structure detection is extensively studied. However, the performance of the existing community detection methods becomes lower as the noise in the related networks increases. To solve this problem, the prior knowledge in the form of pairwise constraints and existing community detection methods are combined to guide the process of community detection, and an extremal optimization based semi-supervised algorithm is proposed for community detection. The experimental results on networks show that compared with the existing methods, the proposed method improves the accuracy of community detection and shows good performance with the noise in the network.
2015 Vol. 28 (2): 162-172 [Abstract] ( 490 ) [HTML 1KB] [ PDF 663KB] ( 589 )
173 Bottleneck Feature Extraction Method Based on Hierarchical Deep Sparse Belief Network
WANG Yi, YANG Jun-An, LIU Hui, LIU Lin
To overcome the drawbacks of original speech features that long temporal speeches and the supervised information can not be effectively utilized and the training time cost is high, a bottleneck feature extraction method based on hierarchical deep sparse belief network is presented. The overlapping group lasso is used as the sparse regularization constraint of the objective function of deep belief network to obtain a deep sparse belief network with a higher speed. To make full use of the hierarchical structure, two sparse deep belief networks are connected in series to enhance the discriminant ability of the bottleneck features. The experimental results on phoneme recognition task show that the proposed feature is effective.
2015 Vol. 28 (2): 173-180 [Abstract] ( 539 ) [HTML 1KB] [ PDF 516KB] ( 1177 )
181 Non-negative and Sparse Graph Construction Algorithm Based on Split Bregman Method
SHEN Ze-Fan, XU Lin-Li
In graph-based machine learning algorithms, the construction of the graph representing the data structure is the key issue. In this paper, a non-negative and sparse graph construction algorithm based on split Bregman method is presented. A weight matrix is learned by solving an equality formulation of the sparse representation through split Bregman method. In the weight matrix, each data sample can be represented by a non-negative linear combination of other samples. The constructed graph of the proposed algorithm can capture the linear relationship between data samples. Experimental results under semi-supervised learning framework demonstrate that the proposed algorithm can capture the latent structure information of data well.
2015 Vol. 28 (2): 181-186 [Abstract] ( 480 ) [HTML 1KB] [ PDF 435KB] ( 531 )
187 Product Attribute Extraction Based on Feature Selection and Pointwise Mutual Information Pruning
GAO Lei, DAI Xin-Yu, HUANG Shu-Jian, CHEN Jia-Jun

Product attribute extraction is a key point in sentiment analysis. In this paper, a product attribute extraction method based on feature selection and pointwise mutual information pruning strategies is proposed. Firstly, the extraction task is transferred to a feature selection task in a classifier. The classification model with l1-norm regularization, such as Lasso, can encourage a sparse model with fewer important selected features. Secondly, some extracted features are selected through a frequency threshold. The features as the product attributes are finally generated with point mutual information pruning. The experiments on the product reviews in Chinese demonstrate the effectiveness of the proposed method.
2015 Vol. 28 (2): 187-192 [Abstract] ( 459 ) [HTML 1KB] [ PDF 420KB] ( 1039 )
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