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....

2018 Vol.31 Issue.7, Published 2018-07-25 Papers and Reports    Researches and Applications        Papers and Reports 581 Attribute Reductions of Interval-Set Concept Lattices for Decision Formal Contexts MA Jianmin, HU Lingling A concept of interval set is introduced into decision formal contexts to study attribute reduction of interval-set concept lattices. Firstly, the conditional and decision interval-set concept lattices are constructed, and a refinement between them is also introduced. Then, the interval-set consistent set and interval-set reduction are defined. Furthermore, the corresponding approaches for judging the interval-set consistent sets are shown. By introducing the discernibility interval-set attribute matrix, methods for obtaining the interval-set attribute reductions are given. 2018 Vol. 31 (7): 581-590 [Abstract] ( 522 ) [HTML 1KB] [ PDF 733KB] ( 602 ) 591 High Resolution SAR Image Segmentation Method Based on Hierarchical Gamma Mixture Model SHI Xue, LI Yu, ZHAO Quanhua To accurately model the complicated statistical characteristics of pixel intensities in a homogeneous region and obtain accurate segmentation results, a high resolution synthetic aperture radar(SAR) image segmentation algorithm based on hierarchical Gamma mixture model(HGaMM) is proposed. HGaMM is constructed by several Gamma mixture models to model the asymmetrical, heavy-tailed and multimodal distribution of pixel intensities. To reduce the influence of image noise on segmentation, Markov random field is employed to model the label field for introducing the spatial neighboring relationship between pixels into HGaMM. Based on Bayesian theory, the segmentation model is built by posterior distribution of model parameters. Markov Chain Monte Carlo algorithm is designed to simulate the segmentation model. Segmentation experiment is conducted on simulated and real SAR images. The results show that the proposed algorithm obtains more accurate segmentation results than other algorithms. 2018 Vol. 31 (7): 591-601 [Abstract] ( 437 ) [HTML 1KB] [ PDF 2482KB] ( 388 ) 602 Low-Resolution Face Recognition Based on Recursive Label Propagation Algorithm XUE Shan, ZHU Hong, WANG Jing, SHI Jing In low-resolution face recognition, the feature representation ability is not robust and the discrimination result of open-set face recognition is inaccurate. Therefore, an algorithm for low-resolution face recognition based on recursive label propagation algorithm is proposed. Firstly, VGG network is utilized to extract face representations. Secondly, the mapping relationship between high-resolution and low-resolution images can be acquired based on the similarity of feature vectors. Finally, the iteration label propagation algorithm is conducted on the labeled and unlabeled facial samples. During the iterations, the adaptive confidence threshold approaching to 100% recognition accuracy is estimated according to the confidence histogram of each category. The identified unlabeled samples are updated to the labeled sample set based on the threshold, thereby the recognition recall rate is improved. Experimental results on public face datasets show that the proposed algorithm achieves a high recall rate with 100% precision. 2018 Vol. 31 (7): 602-611 [Abstract] ( 595 ) [HTML 1KB] [ PDF 1258KB] ( 383 ) 612 Object Tracking with Multiple Memory Learning and Adaptive Correlation Filter Based on Subspace and Histogram FENG Fei, WU Xiaojun, XU Tianyang To enhance the tracking robustness of the correlation filter(CF) in occlusion and background clutter, an algorithm for object tracking with multiple memory learning and adaptive correlation filter based on subspace and histogram is proposed. CF cannot cope with target appearance difference of adjacent frames in the different periods using a single template. A strategy of random updates is proposed to learn multiple target templates and adapt to the variation of target. Several candidate targets are obtained by random updates and the representation coefficients of the previous frame is learned by subspace structure to synthetically judge the accuracy of current candidates. Since CF and subspace representations are sensitive to background clutter, the color histogram is introduced to achieve complementary appearance representation. The statistical histogram is used as the independent judgment basis to improve the accuracy of the algorithm in judging the candidate target. The experimental results on video sequences demonstrate that the proposed algorithm has the ability of anti-occlusion and anti-background interference. 2018 Vol. 31 (7): 612-624 [Abstract] ( 406 ) [HTML 1KB] [ PDF 3763KB] ( 436 ) 625 Scene Classification with Adaptive Learning Rate and Sample Training Mode CHU Jun, SU Yawei, WANG Lu In scene classification based on convolutional neural network, over-fitting is caused due to the large number of network training and poor convergence performance with the small training dataset. To eliminate the negative effect, an algorithm for scene classification with adaptive learning rate and sample training mode is proposed. The network learning rate is adaptively adjusted on the framework of convolutional neural network according to the variation of the error function in the network training. When the error function changes slightly, the learning rate of the batch is unchanged. When the error function changes more remarkably, the variation of the learning rate is inversely proportional to the variation of the error function. Meanwhile, according to the network output, the sample training mode is switched, and the inaccurately recognized images are emphatically trained. The experimental results on Scene-15 and Cifar-10 scene datasets show that the proposed method improves the convergence of neural networks and effectively improves the classification accuracy, especially the classification accuracy of complex scenes such as indoor scenes. 2018 Vol. 31 (7): 625-633 [Abstract] ( 469 ) [HTML 1KB] [ PDF 1167KB] ( 328 ) Researches and Applications 634 Semi-supervised Short Text Stream Classification Based on Vector Representation and Label Propagation WANG Haiyan , HU Xuegang , LI Peipei The huge volume of short text streams produced by social Network is fast, high-volume and it contains concept drift, short length of texts and massive unlabeled data. Therefore, a semi-supervised short text stream classification algorithm based on vector representation and label propagation is proposed in this paper to classify short text stream with a few labeled data. Besides, to adapt to the concept drift, a concept drift detection algorithm based on clusters is proposed. Experimental results on real short text streams show that the proposed algorithm improves the classification accuracy and the macro average compared with classical semi-supervised classification algorithms and semi-supervised data stream classification algorithms, and it adapts to the concept drift quickly in data stream. 2018 Vol. 31 (7): 634-642 [Abstract] ( 437 ) [HTML 1KB] [ PDF 1801KB] ( 368 ) 643 Model Decision Tree: An Accelerated Algorithm of Decision Tree YIN Ru, MEN Changqian, WANG Wenjian, LIU Shuze The decision tree algorithm is constructed in a recursive style. Therefore, the low training efficiency is yielded and the over-classification of decision tree may produce overfitting. An accelerated algorithm called model decision tree(MDT) is proposed in this paper. An incomplete classification decision tree is established via the Gini index on the training dataset firstly. Then a simple model is utilized to classify impure pseudo leaf nodes, which are neither leaf nodes nor in the same class. Consequently, the final MDT is generated. Compared with DT, MDT improves the training efficiency with smaller loss of classification accuracy or even no loss. The experimental results on benchmark datasets show that the proposed MDT is much faster than DT and it has a certain ability to avoid overfitting. 2018 Vol. 31 (7): 643-652 [Abstract] ( 608 ) [HTML 1KB] [ PDF 869KB] ( 385 ) 653 Daily Behavior Recognition with Single Sensor Based on Functional Time Series Data Modeling SU Benyue, ZHENG Dandan, SHENG Min In inertial sensor based human activity recognition, the periodic and temporal characteristics are often ignored in the traditional algorithms, and there are corresponding requirements for the size of the sliding window to extract features. In this paper, a recognition algorithm based on functional data analysis and hidden Markov model for periodic behavior is proposed with a single wearable sensor placed on the waist for human daily activities. Firstly, the functional data analysis method is used to fit the motion capture data of periodic daily activities, and then the single cycle data are extracted after fitting. Secondly,based on the single periodic behavior data, a hidden Markov model describing each daily behavior process is established. Finally, human activities are classified with the maximum likelihood .Compared with the multisensor human activity recognition methods, the proposed method is able to effectively classify 8 daily activities via single sensor with high recognition rates in both user dependent mode and user independent mode. 2018 Vol. 31 (7): 653-661 [Abstract] ( 469 ) [HTML 1KB] [ PDF 830KB] ( 315 ) 662 Speech Recognition Based on Semi-supervised Data Selection via Decoding Multiple Candidate Results WANG Xilou, GUO Wu, XIE Chuandong For speech recognition of low resources, a selection strategy for semi-supervised learning with a large number of unlabeled data is proposed, and this strategy is applied to both acoustic modeling and language modeling. After a small amount of data is used to train the seed model, the unlabeled data is decoded using the seed model. Firstly, high-confidence sentences are selected by using a combination of confidence measure and perplexity in the decoded best candidate results. Then, the high-confidence sentences are used to train acoustic model and language model. Furthermore, the decoded lattice is transformed to obtain multiple candidate texts for language model training. In the Japanese recognition task, the proposed method obtains a better recognition rate than the method of selecting data based on confidence measure. 2018 Vol. 31 (7): 662-667 [Abstract] ( 387 ) [HTML 1KB] [ PDF 628KB] ( 411 ) 668 A Parallelized Multi-objective Particle Swarm Optimization Algorithm Based on MPI GENG Wenjing , DONG Hongbin, DING Rui To improve the efficiency and accuracy of speed-constrained multi-objective particle swarm optimization(SMPSO), a parallelized SMPSO algorithm based on Message Passing Interface(MPI)(M-SMPSO) is proposed. The master-slave mode of MPI is used in the proposed algorithm. The entire population is divided into several sub-populations. Then, these sub-populations are evolved independently. In addition, an adaptive global optimal solution selection strategy is proposed to balance the distribution and convergence. Several standard test functions are adopted to verify the performance of the proposed algorithm. The experimental results show that ompared with other multi-objective algorithms, M-SMPSO obtains a higher speedup ratio and it converges quickly. 2018 Vol. 31 (7): 668-676 [Abstract] ( 712 ) [HTML 1KB] [ PDF 915KB] ( 314 )

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