Wang Z.-M.,University of Science and Technology Beijing |
Wang Z.-M.,Beijing Key Laboratory of Knowledge Engineering for Materials Science
Zidonghua Xuebao/Acta Automatica Sinica | Year: 2015
Image quality has a strong impact on human visual information acquisition. It is a key but difficult task to evaluate the quality of a distorted image without a reference image. This paper reviews the main techniques of no-reference image quality assessment (IQA) developed during the past 20 years. Firstly, some technical indexes for IQA algorithm evaluation and several public IQA databases available on network are introduced. Then, various no-reference IQA algorithms are introduced, sorted and discussed in detail. At last, several non-distortion-specific no-reference IQA algorithms presented in recent years are tested and compared on a public database. The purpose of this paper is to provide an integrated and valuable reference for no-reference IQA research. Copyright © 2015 Acta Automatica Sinica. All rights reserved.
Luo X.,University of Science and Technology Beijing |
Luo X.,Beijing Key Laboratory of Knowledge Engineering for Materials Science |
Chang X.,University of Science and Technology Beijing |
Chang X.,Beijing Key Laboratory of Knowledge Engineering for Materials Science
International Journal of Control, Automation and Systems | Year: 2015
With the increasing presence and adoption of wireless sensor networks (WSNs), the demand of data acquisition and data fusion are becoming stronger and stronger. In WSN, sensor nodes periodically sense data and send them to the sink node. Since the network consists of plenty of low-cost sensor nodes with limited battery power and the sensed data usually are of high temporal redundancy, prediction- based data fusion has been put forward as an important issue to reduce the number of transmissions and save the energy of the sensor nodes. Considering the fact that the sensor node usually has limited capabilities of data processing and storage, a novel prediction-based data fusion scheme using grey model (GM) and optimally pruned extreme learning machine (OP-ELM) is proposed. The proposed data fusion scheme called GM-OP-ELM uses a dual prediction mechanism to keep the prediction data series at the sink node and sensor node synchronous. During the data fusion process, GM is introduced to initially predict the data of next period with a small number of data items, and an OPELM- based single-hidden layer feedforward network (SLFN) is used to make the initial predicted value approximate its true value with extremely fast speed. As a robust and fast neural network learning algorithm, OP-ELM can adaptively adjust the structure of the SLFN. Then, GM-OP-ELM can provide high prediction accuracy, low communication overhead, and good scalability. We evaluate the performance of GM-OP-ELM on three actual data sets that collected from 54 sensors deployed in the Intel Berkeley Research lab. Simulation results have shown that the proposed data fusion scheme can significantly reduce redundant transmissions and extend the lifetime of the whole network with low computational cost. © 2015, Institute of Control, Robotics and Systems and The Korean Institute of Electrical Engineers and Springer-Verlag Berlin Heidelberg.
Qian W.,Nanchang University |
Qian W.,Beijing Key Laboratory of Knowledge Engineering for Materials Science |
Shu W.,East China Jiaotong University
Neurocomputing | Year: 2015
Feature selection is an important preprocessing step in machine learning and data mining, and feature criterion arises a key issue in the construction of feature selection algorithms. Mutual information is one of the widely used criteria in feature selection, which determines the relevance between features and target classes. Some mutual information-based feature selection algorithms have been extensively studied, but less effort has been made to investigate the feature selection issue in incomplete data. In this paper, combined with the tolerance information granules in rough sets, the mutual information criterion is provided for evaluating candidate features in incomplete data, which not only utilizes the largest mutual information with the target class but also takes into consideration the redundancy between selected features. We first validate the feasibility of the mutual information. Then an effective mutual information-based feature selection algorithm with forward greedy strategy is developed in incomplete data. To further accelerate the feature selection process, the selection of candidate features is implemented in a dwindling object set. Compared with existing feature selection algorithms, the experimental results on different real data sets show that the proposed algorithm is more effective for feature selection in incomplete data at most cases. © 2015 Elsevier B.V.
Li Y.,University of Science and Technology Beijing |
Li Y.,Beijing Key Laboratory of Knowledge Engineering for Materials Science |
Hu C.,University of Science and Technology Beijing |
Minku L.L.,University of Birmingham |
Zuo H.,University of Science and Technology Beijing
Genetic Programming and Evolvable Machines | Year: 2013
Learning aesthetic judgements is essential for reducing users' fatigue in evolutionary art systems. Although judging beauty is a highly subjective task, we consider that certain features are important to please users. In this paper, we introduce an adaptive model to learn aesthetic judgements in the task of interactive evolutionary art. Following previous work, we explore a collection of aesthetic measurements based on aesthetic principles. We then reduce them to a relevant subset by feature selection, and build the model by learning the features extracted from previous interactions. To apply a more accurate model, multi-layer perceptron and C4.5 decision tree classifiers are compared. In order to test the efficacy of the approach, an evolutionary art system is built by adopting this model, which analyzes the user's aesthetic judgements and approximates their implicit aesthetic intentions in the subsequent generations. We first tested these aesthetic measurements on different artworks from our selected artists. Then, a series of experiments were performed by a group of users to validate the adaptive learning model. The study reveals that different features are useful for identifying different patterns, but not all are relevant for the description of artists' styles. Our results show that the use of the learning model in evolutionary art systems is sound and promising for predicting users' preferences. © 2013 Springer Science+Business Media New York.
Shu W.,Beijing Jiaotong University |
Qian W.,Nanchang University |
Qian W.,Beijing Key Laboratory of Knowledge Engineering for Materials Science
Knowledge-Based Systems | Year: 2014
Attribute measures, used to evaluate the quality of candidate attributes, play an important role in the process of attribute reduction. They largely affect the computational efficiency of attribute reduction. Existing attribute measures are acted on the whole universe in complete decision systems. There are few studies on improving attribute reduction algorithms from the perspective of attribute measures in incomplete decision systems, which motivates the study in this paper. This paper proposes new attribute measures that act on a dwindling universe to quicken the attribute reduction process. In particular, the monotonicity guarantees the rationality of the proposed attribute measures to evaluate the significance of candidate attributes. On this basis, the corresponding attribute reduction algorithms are developed in incomplete decision systems based on indiscernibility relation and discernibility relation, respectively. Finally, a series of comparative experiments are conducted with different UCI data sets to evaluate the performance of our proposed algorithms. The experimental results indicate that the proposed algorithms are efficient and feasible. © 2014 Elsevier B.V. All rights reserved.
Chen Y.,University of Science and Technology Beijing |
Chen Y.,Beijing Key Laboratory of Knowledge Engineering for Materials Science
Sensors and Transducers | Year: 2014
At present, most of research works in mobile network focus on the network overhead of the known path which exists between the sender and the receiver. However, the trend of the current practical application demands is becoming increasingly distributed and decentralized. The Delay and Tolerant Network (DTN) just comes out of such background of the conflicts between them. The DTN could effectively eliminate the gap between the mobile network and the practical application demands. In this paper, a Self-Adaptive Context Aware Routing Protocol (SACARP) for the unicast communication in delay and tolerant networks is presented. Meanwhile, according to the real-time context information of DTN, the Kalman filter theory is introduced to predict the information state of mobility for the optional message ferrying node, and then gives the optimal selection strategy of the message ferrying nodes. The simulation experiments have shown that, compared to the familiar single-copy and multi-copy protocols, the SACARP proposed in this paper has better transmission performance and stability, especially when the network is free, the protocol would keep a good performance with fewer connections and less buffer space. © 2014 IFSA Publishing, S. L.
Yu Y.,University of Science and Technology Beijing |
Wang Q.,University of Science and Technology Beijing |
Wang X.,Beijing Key Laboratory of Knowledge Engineering for Materials Science
China Communications | Year: 2013
The clustering of trajectories over huge volumes of streaming data has been recognized as critical for many modern applications. In this work, we propose a continuous clustering of trajectories of moving objects over high speed data streams, which updates online trajectory clusters on basis of incremental line-segment clustering. The proposed clustering algorithm obtains trajectory clusters efficiently and stores all closed trajectory clusters in a bi-tree index with efficient search capability. Next, we present two query processing methods by utilising three proposed pruning strategies to fast handle two continuous spatio-temporal queries, threshold-based trajectory clustering queries and threshold-based trajectory outlier detections. Finally, the comprehensive experimental studies demonstrate that our algorithm achieves excellent effectiveness and high efficiency for continuous clustering on both synthetic and real streaming data, and the proposed query processing methods utilise average 90% less time than the naive query methods. © 2013 IEEE.
Guo C.,University of Science and Technology Beijing |
Guo C.,Beijing Key Laboratory of Knowledge Engineering for Materials Science |
Zheng X.,University of Science and Technology Beijing |
Zheng X.,Beijing Key Laboratory of Knowledge Engineering for Materials Science
Proceedings - 2014 IEEE International Conference on Granular Computing, GrC 2014 | Year: 2014
Feature subset selection, as an important processing step to knowledge discovery and machine learning, is effective method in reducing irrelevant and or redundant features, compressing repeated data, and improving classification accuracy. Rough set theory is an important tool to select feature subset from high-dimensional data. In this work, feature subset selection based on fuzzy rough set is introduced, and the efficient measure of feature significance is designed. Based on the fuzzy rough set model, a quick feature subset selection approach is presented, which can efficiently identify relevant features as well as redundancy among all features. In addition, the KNN-based classifier based on the proposed approach is constructed. The experimental results show that the proposed feature subset selection approach achieves better classification on UCI datasets. © 2014 IEEE.
Yu Y.,University of Science and Technology Beijing |
Yu Y.,Beijing Key Laboratory of Knowledge Engineering for Materials Science |
Wang Q.,University of Science and Technology Beijing |
Wang Q.,Beijing Key Laboratory of Knowledge Engineering for Materials Science |
And 3 more authors.
Computer Science and Information Systems | Year: 2013
Trajectory data streams contain huge amounts of data pertaining to the time and position of moving objects. It is crucial to extract use-ful information from this peculiar kind of data in many application scenarios, such as vehicle traffic management, large-scale tracking management and video surveillance. This paper proposes a density-based clustering algorithm for trajectory data stream called CTraStream. It contains two stages: trajectory line segment stream clustering and online trajectory cluster updating. CTraStream handles the trajectory data of moving objects as an incremental line segment stream. For line segment stream clustering, we present a distance measurement approach between line segments. Incremental line segments are processed quickly based on previous line clusters in order to achieve clustering line segment stream online, and line-segment-clusters in a time interval are obtained on the fly. For online trajectory cluster updating, TC-Tree, an index structure, which stores all closed trajectory clusters, is designed. According to the linesegment-cluster set, the current closed trajectory clusters are updated online based on TC-Tree by performing proposed update rules. The algorithm has exhibited many advantages, such as high scalability to process incremental trajectory data streams and the ability to discover trajectory clusters in data streams in real time. Our performance evaluation experiments conducted on a number of real and synthetic trajectory datasets illustrate the effectiveness, efficiency, and scalability of the algorithm.
PubMed | Beijing Key Laboratory of Knowledge Engineering for Materials Science, University of Science and Technology Beijing and Communication University of China
Type: | Journal: Journal of food science | Year: 2017
Growing evidence shows that consumer choices in real life are mostly driven by unconscious mechanisms rather than conscious. The unconscious process could be measured by behavioral measurements. This study aims to apply automatic facial expression analysis technique for consumers emotion representation, and explore the relationships between sensory perception and facial responses. Basic taste solutions (sourness, sweetness, bitterness, umami, and saltiness) with 6 levels plus water were used, which could cover most of the tastes found in food and drink. The other contribution of this study is to analyze the characteristics of facial expressions and correlation between facial expressions and perceptive hedonic liking for Asian consumers. Up until now, the facial expression application researches only reported for western consumers, while few related researches investigated the facial responses during food consuming for Asian consumers. Experimental results indicated that facial expressions could identify different stimuli with various concentrations and different hedonic levels. The perceived liking increased at lower concentrations and decreased at higher concentrations, while samples with medium concentrations were perceived as the most pleasant except sweetness and bitterness. High correlations were founded between perceived intensities of bitterness, umami, saltiness, and facial reactions of disgust and fear. Facial expression disgust and anger could characterize emotion dislike, and happiness could characterize emotion like, while neutral could represent neither like nor dislike. The identified facial expressions agree with the perceived sensory emotions elicited by basic taste solutions. The correlation analysis between hedonic levels and facial expression intensities obtained in this study are in accordance with that discussed for western consumers.