ISSN 0253-2778

CN 34-1054/N

2019 Vol. 49, No. 1

Display Method:
Defence mechanism against wormhole attacks based on grid
WU Yichen, ZHANG Shuguang
2019, 49(1): 1-7. doi: 10.3969/j.issn.0253-2778.2019.01.001
A wormhole attack destroys routing operations, affects the node positioning, and poses great threats to the security of the entire wireless sensor network (WSN). Aiming at resolving this issue, the grid based wormhole attack defense strategy was presented. Through additional grid properties, information was exchanged with nodes in specific grids to detect wormhole attacks, including attacks from special wormholes with the forgery and tampering ability in a network, thus greatly reducing damages brought by wormhole attacks. The simulated experiment proves that malicious nodes can be effectively detected in an attacked network if this strategy is used.
A multi-domain sentiment classification model based on sample filtering and transfer learning
QU Zhaowei, ZHAO Yanjiao, WANG Xiaoru
2019, 49(1): 8-14. doi: 10.3969/j.issn.0253-2778.2019.01.002
Most of the models for sentiment classification are trained and tested on a single dataset. However, the model parameters obtained by training on one dataset are not suitable for another dataset and the model is not generic. A multi-domain sentiment classification model (MDSC) was proposed. With sample filtering and transfer learning, the trained model can be applied to different datasets in multiple domains and the model is more applicable and expandable. Specifically, a document is first mapped to the domain distribution which is used as a bridge between domain classification and sentiment classification, and then sentiment classification is completed. In order to make the model more generic, representative data samples should be selected. MDSC constructs a domain-independent sentiment lexicon to filter sentences that belong to the same document and obtain a high-quality training dataset. At the same time, to improve the classification accuracy and reduce the training time, parameter-based transfer learning with neutral networks is used to obtain the document embeddings for classification. Extensive experiments on datasets containing 15 different domains show that the proposed model can achieve better performance compared with traditional models when applied to datasets in multiple domains.
An advanced flat predicate mechanism and compiling optimization method
WANG Xiangqian, ZHENG Qilong, ZHANG Rengao, HAN Dongke
2019, 49(1): 15-20. doi: 10.3969/j.issn.0253-2778.2019.01.003
Predicate execution is a valid method to develop instructions parallelism in programs with control flow. Predicate computation is done one by one locally in the process of classic predicate for which multi-predicate control is forbidden. This may cause some problems such as long predicate implementation computation path. To solve these problems, an advanced flat predicate mechanism is proposed, which can do predicate computation globally, and perform multi-predicate control naturally. Then, compiling method for the flat predicate mechanism is researched, and a compiling framework for the flat predicate mechanism is presented. Experiments show that the flat predicate mechanism and compiling implementation framework can better enhance the executive efficiency of multi-condition programs.
Student score prediction: A knowledge-aware auto-encoder model
SU Yu, ZHANG Dang, LIU Qingwen, ZHANG Yingjie, CHEN Yuying, DING Chris
2019, 49(1): 21-30. doi: 10.3969/j.issn.0253-2778.2019.01.004
To reduce study burden and boost efficiency, online education systems offer personalized learning experience for students. In such systems, ability assessment is a fundamental task as reflected by a basic task, named score prediction. The main drawbacks of existing prediction methods are: ① Inability unable to fully exploit the potential of big data, ② cold start problem,③ lack of reasonable explanations. A novel knowledge-aware auto-encoder model (KAEM) is proposed to address these issues. Specifically, an exercise-knowledge-graph with education experts’ prior knowledge is introduced. Then students’ performance is modeled using auto-encoders with the combination of information in knowledge graph as regularization item. By encoding and integrating the experts’ prior knowledge, KAME can improve both prediction accuracy and model robustness and deal with the cold start problem well. Furthermore, reasonable explanations for recommendations can be generated using this model. KMAE has been applied to a famous online education system. Extensive experiments on large-scale real data clearly demonstrate its effectiveness.
Pairwise interaction tensor factorization based tag recommendation
LU Yanan, DU Dongfang
2019, 49(1): 31-39. doi: 10.3969/j.issn.0253-2778.2019.01.005
The tag recommendation system is a series of tags that are most likely to be used to tag a resource for the target user. Currently, the Tucker decomposition model has better prediction quality than the traditional FolkRank algorithm, but it has high time complexity and is difficult to apply to large and medium-sized data sets. Although the time complexity of the regular decomposition model is linear, its prediction quality is not high. To solve these problems, firstly, the paired interaction tensor decomposition model PITD on the basis of improving the Tucker decomposition model is proposed. The model considers only some of the two-to-two interactions between the three characteristics of users, resources, and tags, reducing the impact of irrelevant information on model performance and efficiency. Then, the PITD model is deduced by Bayesian personalization method, and the corresponding optimization algorithm is designed. Finally, extensive experiments on real data sets show that the PITD model has better recommendation performance than the comparison algorithm.
Recognizing emotions from abstract paintings using convolutional neural network with two-layer transfer learning scheme
YANG Ziwen, CHEN Lei, PU Jianyu
2019, 49(1): 40-48. doi: 10.3969/j.issn.0253-2778.2019.01.006
In order to bridge the gap between low-level visual features and high-level emotional semantics, and to alleviate the defects inherent in small sample dataset in abstract paintings emotions recognition datasets, a two-layer transfer learning strategy is introduced into traditional convolutional neural networks and a model for recognizing emotions from abstract paintings is proposed using convolutional neural network with a two-layer transfer learning scheme. According to the hierarchical nature of deep features, a large-scale generalized image dataset is used to learn how extract universal low-level image features. Then the relevant domain dataset is utilized to learn how extract specific high-level semantic features. Finally the abstract painting emotion recognition dataset is used to finetune the network. As shown by our extensive experimental validation on MART datasets, the proposal outperforms current methods when recognizing emotions from abstract paintings.
Tire impurity defect detection based on morphology and projection histogram
YU Xiangru, DING Jianpei, LI Jinping
2019, 49(1): 49-54. doi: 10.3969/j.issn.0253-2778.2019.01.007
A method based on morphology and projection histogram is proposed, which aims at detecting impurities in the sidewall and shoulder of the tire. Firstly, image preprocessing is used to segment the sidewall and shoulder, decreasing the negative influence caused by the tread pattern. Secondly, the binarized images are obtained by using OTSU in the local area, while impurities are extracted from the background by morphological operation. Then, vertical filtering is performed to remove burrs. Finally, the impurity is located by vertical and horizontal projections. The position of the impurity is random, and the projection curves of impurities are consistent with the square wave model, so the impurities are filtered according to the above characteristics of impurity. The experiments show that the proposed method can effectively detect impurities in the sidewall and shoulder outside the tread pattern, and that it can also meet the system requirement of real-time performance.
Robust stability of reduced-order linear ADRC for first-order plants with large time-delay
WANG Yongshuai, CHEN Zengqiang, SUN Mingwei, SUN Qinglin
2019, 49(1): 55-62. doi: 10.3969/j.issn.0253-2778.2019.01.008
Aiming at first-order inertial plants with large time-delay, the stability and robustness of RLADRC are studied by combining it with the Smith predictor. The stable feasible region of parameters is obtained according to routh criterion together with the verification of the numerical analysis. Besides, the phase margin range in the feasible region is analyzed according to the frequency response. In the end, the predictive RLADRC is compared with single RLADRC on robustness when parameters of control plants have some perturbation, and the results prove that predictive RLADRC has better dynamic performance and stronger robustness based on Monte Carlo experiments. These conclusions can be used to design parameters of the Smith predictor and RLADRC controllers.
Salt&pepper noise detection algorithm for color images based on Spearman rank correlation
JIA Xiaofen, GUO Yongcun, HUANG Yourui, ZHAO Baiting
2019, 49(1): 63-70. doi: 10.3969/j.issn.0253-2778.2019.01.009
To avoid the operation on non-noise pixels, and preserve the original feature information of the image, which can protect the edge details and obtain high quality color images, a noise detection method based on Spearman rank correlation coefficient is designed for color images. Firstly, each pixel of the color image is taken as the point to be detected, and two six-dimensional vectors are constructed using the spatial and color correlation of the color image between the detection point and the adjacent pixel on the left side. Secondly, the Spearman rank correlation coefficient of the two vectors is calculated, and the extreme points of the image are selected by setting the threshold T1, which is used to make comparisons with the Spearman rank correlation coefficients. Then, the comparative results are used to determine all the possible noise and edge points, called the extreme points. The primary point is determined as the extreme point if the Spearman rank correlation coefficient is smaller than the threshold T1. Finally, eight Spearman correlation coefficients between the extreme point and its surrounding eight pixels are calculated, and the threshold value T2 is set, then, comparisons are made between the eight Spearman rank correlation coefficients and T2 to determine whether the extreme point is noise point or edge point. The simulation results show that the reference values of the two parameters optimized by particle swarm optimization algorithm are 0.28
Development of smartphone based magnetic resonance elastography simulation and information processing system
LIANG Xiao, SHI Mian, SHAN Xiang, WU Jie, LI Bingnan, GAO Rongke
2019, 49(1): 71-78. doi: 10.3969/j.issn.0253-2778.2019.01.010
To meet the requirement of portability and intelligentization when medical doctors, researchers and patients use magnetic resonance elastography (MRE) information processing system, a magnetic resonance elastic image simulation and information processing system was designed based on the smartphone platform. By analyzing MRE and studying processing algorithms, software based on the html5 + css3 + javascript language in AppCan (hybrid mixed mobile development platform) platform was developed. The design of software interface and acquisition of MRE data were projected by html5+css3 language. The simulation and reconstruction of the MRE data was designed by javascript language, and it can be shown on the smartphone platform in real time. The significance of the software lies in the advantage of operating easily by touch screen on the smartphone platform. The simulation and reconstruction process of MRE is more intelligent and portable, indicating its application value and potential in popularizing smart medical diagnosis and personalized diagnosis.
Monte carlo simulation of secondary electron emission from wave-type structure
KHAN Muhammad Saadat Shakoor, ZOU Yanbo, LI Chao, DING Zejun
2019, 49(1): 79-86. doi: 10.3969/j.issn.0253-2778.2019.01.011
Monte Carlo (MC) simulation techniques for the study of electron interaction with solids have been successfully applied to obtain the line-scan profiles in critical dimension scanning electron microscopy (CD-SEM). However, previous studies have been mostly concerned about the sample of simple geometries having sharp edges. The simulation was extended to the study of wave-type structures with smooth curved shapes. The MC model is based on using the Mott cross-section for electron elastic scattering and the full Penn algorithm in a dielectric function approach to electron inelastic scattering. The CD-SEM line-scan profiles of wave-type structures have been calculated by taking into account such experimental factors as primary beam energy, geometry parameters and material property. It is shown that by decreasing the height of the structure, the double side peaks can shrink to merge into a single peak. This characteristic will pose a challenge to the CD characterization for the smoothed line structure.