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....

2016 Vol.29 Issue.4, Published 2016-04-30 Papers and Reports    Researches and Applications        Papers and Reports 289 A Semi-supervised Method for Phrase-Level Sentiment Analysis Odbal, WANG Zengfu The existing methods for sentiment analysis can not dig more complex linguistic phenomena in emotional expression and they encounter the challenge of sparse features. An innovative semi-supervised phrase-level sentiment analysis method based on semantic space model is proposed. Firstly, the problem of word representation in semantic space is discussed and word-level semantic distribution computing methods based on dependency grammar semantic space model are proposed, and the computational procedure is completed by using unsupervised method. Secondly, the problems of phrase recognition and representation are discussed and nonlinear combinations of word-level semantic distribution are used to represent the multi-word structures. Finally, a neutral network algorithm is used to design the phrase-level sentiment analysis system based on word level semantic distribution and phrasal structure representation. Experimental results on real Chinese corpora show the expected recognition accuracy of the model. 2016 Vol. 29 (4): 289-297 [Abstract] ( 591 ) [HTML 1KB] [ PDF 487KB] ( 590 ) 298 Disturbing-Valued Fuzzy Finite-State Automata and Their Languages PENG Jiayin The concepts of disturbing-valued fuzzy finite-state automata and their languages are introduced. The extension of state transition function for disturbing-valued fuzzy finite-state automata is discussed. The non-deterministic disturbing-valued fuzzy finite-state automata and three kinds of deterministic disturbing-valued fuzzy finite-state automata are equivalent to each other. The closeness of the languages families of disturbing-valued fuzzy finite-state automata under regular operations are studied. 2016 Vol. 29 (4): 298-312 [Abstract] ( 481 ) [HTML 1KB] [ PDF 489KB] ( 422 ) 313 Human Action Recognition Algorithm Based on Selective Ensemble Rotation Forest TANG Chao, WANG Wenjian, LI Wei , LI Guobin, CAO Feng, ZHANG Miaohui The representation of high dimensional human actions and the construction of accurate and stable human classification model are key issues in human action recognition. An efficient action recognition algorithm based on mixed features is proposed. Key joints of human body polar coordinates features based on appearance structure and motion features based on optical flow are fused into the proposed algorithm to capture motion information in video sequences and improve the recognition instantaneity. Meanwhile, the selective ensemble rotation forest model (SERF) based on frame is developed and the selection ensemble strategy is used to select the base classifier of rotation forest and increase differences among the classifiers. Experimental results show the better classification accuracy and robustness of the proposed model. 2016 Vol. 29 (4): 313-321 [Abstract] ( 550 ) [HTML 1KB] [ PDF 7390KB] ( 627 ) 322 Undecimated Wavelet Bayesian Image Denoising Method with Its Threshold Determined by Curve Fitting WANG Xianghai, LIU Xiaoqian, ZHANG Aidi, FU Bo The undecimated discrete wavelet transform(UDWT) possesses local features of time and frequency and shift-invariant property of reducing the pseudo-Gibbs phenomenon. In this paper, after the UDWT coefficients are analyzed, the conclusion that the UDWT coefficients have strong non-Gaussian statistical property is obtained. Grounded on the property, a generalized guassian distribution model is established. To improve the precision of standard deviation estimation of the noise image, a method of curve fitting is proposed based on the standard deviation of image, and thus the denoising threshold is determined. Based on the shift-invariant property of UDWT, the proposed method effectively reduces the pseudo-Gibbs phenomenon of the traditional wavelet denoising method. Meanwhile, the denoising effect is enhanced by improving the accuracy of denoising threshold. A large number of simulation experiments verifies the effectiveness of the proposed method. 2016 Vol. 29 (4): 322-331 [Abstract] ( 630 ) [HTML 1KB] [ PDF 2412KB] ( 893 ) 332 Payoff Distribution Strategy of Overlapping Coalitions for Concurrent Multiple Tasks GUI Haixia , JIANG Jianguo, ZHANG Guofu Payoff distribution of overlapping coalitions is a difficult topic in multi-agent systems. A payoff distribution strategy of overlapping coalitions for concurrent multiple tasks is proposed in this paper. Based on the idea of more abilities for more works, multiple concurrent tasks are dispatched in parallel by proportional allocation. Meanwhile, the payoff of overlapping coalitions is distributed according to the results of task dispatch. Then, a sufficient and necessary condition that one agent satisfies the principle of non-reducing utility when joining multiple coalitions is deduced. Finally, the effectiveness of the proposed method is proved by an example, and a comparative analysis between the proposed strategy and the serial utility allocation is carried out. The result shows that when a new agent applies for joining coalitions, the proposed strategy can satisfy the condition of non-reducing utility more easily and it has better timeliness. 2016 Vol. 29 (4): 332-340 [Abstract] ( 384 ) [HTML 1KB] [ PDF 421KB] ( 383 ) Researches and Applications 341 Video Anomaly Detection Algorithm Based on Weighted Sample Selection and Active Learning LIN Ling, LIAO De, GAO Yang, YANG Wanqi Due to the surge in public security issues, anomaly detection in video surveillance is a hot topic in computer vision. Taking characteristics of the dataset for video anomaly detection into account, a video anomaly detection algorithm based on weighted sample selection and active learning is proposed. According to the distribution characteristics of the dataset, appropriate weights for instances are selected to remove the effect of imbalanced data on the classifier. Active learning is used to select the uncertain instances. To adapt to the complex environment, the model is updated iteratively. Experimental results on UCSD dataset show that the proposed algorithm outperforms traditional algorithms. 2016 Vol. 29 (4): 341-349 [Abstract] ( 622 ) [HTML 1KB] [ PDF 1360KB] ( 827 ) 350 Out-of-Vocabulary Word Recognition Based on Lattice Combination of Complement Sub-lexical Units FAN Zhengguang, QU Dan, CHEN Bin Different sub-lexical units used in hybrid model often provide complementary information for each other during out-of-vocabulary (OOV) words recognition. In this paper, a lattice combination method of complement sub-lexical units for out-of-vocabulary words recognition is proposed. Firstly, two hybrid model systems with performance difference are built respectively by using syllables and graphones. Next, the recognition lattices are obtained from the built systems and the sub-lexical units are preprocessed for the purpose of combination. Finally, the combination strategies based on lattices union and lattices intersection are respectively explored to combine the lattices to acquire the better result of OOV Words recognition . The experimental results show the proposed method is superior to individual system and the recognizer output voting error reduction (ROVER) system in OOV words recognition. 2016 Vol. 29 (4): 350-358 [Abstract] ( 433 ) [HTML 1KB] [ PDF 685KB] ( 747 ) 359 Collaborative Recommendation Framework Based on Ratings and Textual Reviews TAN Yunzhi, ZHANG Min, LIU Yiqun, MA Shaoping The feedback of users usually contains a numeric rating and a textual review. In this paper, textual review information is used to learn the distributions of item features on different topics and the user preference to different features of items. Then, the topic-based user preference similarity is incorporated into the traditional collaborative filtering recommendation systems. A recommendation framework based on ratings and textual reviews is proposed. With the proposed framework, review information can be easily introduced into the existing recommendation algorithms. By employing textual reviews, the problem of data sparsity in the traditional recommendation algorithms is relieved. Experiments are conducted on 22 real-world datasets from Amazon and the experimental results demonstrate the advantages and the effectiveness of the proposed framework. 2016 Vol. 29 (4): 359-366 [Abstract] ( 627 ) [HTML 1KB] [ PDF 471KB] ( 908 ) 367 Decision Tree Ensemble Based Partial Label Learning Algorithm YU Fei, ZHANG Minling To overcome the problem of the missing supervision information in partial label learning, a special splitting measure for the generation of decision tree is designed according to the property of partial label examples and the growth algorithm of decision tree is modified. In the proposed algorithm, bootstrap sampling is employed to construct multiple decision trees, and then the final prediction result is obtained by voting on the classification results of each decision tree. Experiments on artificial datasets and real-world datasets validate the good performance of the proposed algorithm. 2016 Vol. 29 (4): 367-375 [Abstract] ( 678 ) [HTML 1KB] [ PDF 582KB] ( 699 ) 376 Compression History Detection for WAV Audio Based on Phase Spectrum Differences ZHOU Jinglei, WANG Rangding, JIN Chao, YAN Diqun, CHEN Ya′nan As an important branch of audio forensics, audio compression history detection is of great significance for audio tampering and forgery detection. In this paper, a WAV audio compression history detection algorithm is proposed based on the statistics of the phase spectrum differences. The mean, the variance and the kurtosis of phase spectrum differences are utilized as the detection features. The proposed algorithm can determine whether the suspicious WAV audio is compressed and decompressed by any of the four popular encoders. Furthermore, the bit rate used in the compression can also be estimated. Experimental results show that the performance of the proposed algorithm is better than that of state-of-the-art detection algorithms. 2016 Vol. 29 (4): 376-384 [Abstract] ( 511 ) [HTML 1KB] [ PDF 702KB] ( 618 )

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