Table of Content

19 December 2016, Volume 0 Issue 12

Previous Issue    Next Issue

 A Multi-label Metamorphic Relations Prediction Approach Based on RBF Neural Network

ZENG Jin-wei1, ZHANG Peng-cheng1, LI Wen-rui2， ZHOU Xin-li3

2016, 0(12):  1-6,11.  doi: 10.3969/j.issn.1006-2475.2016.12.001

Asbtract ( 158 )  

References | Related Articles | Metrics

 A multi-label metamorphic relations prediction approach based on Radial Basis Function(RBF) neural network is proposed, which uses Soot analysis tool to generate the control flow graph and the corresponding control flow graph with labels. And then, nodes and path properties extracted from functional control flow graph constitute multi-label data sets. Finally, a multi-label RBF neural network prediction model established by learning the sample data sets is used to predict the function that can satisfy multiple metamorphic relations. Experiment results prove that the method is effective.


 A Collaborative Filtering Recommendation Algorithm Based on Trust Relationship

WANG Yan, WANG Yi-zhi

2016, 0(12):  7-11.  doi: 10.3969/j.issn.1006-2475.2016.12.002

Asbtract ( 140 )  

References | Related Articles | Metrics

 With the development of network technology and multimedia technology, the scale of the teaching resources on the network becomes pretty large. How to recommend the resources to the learners according to the needs of the learners has become a hot issue in the recent years. However, because of the collaborative filtering recommendation algorithm taking the trust relationship between the users and target users so less, it is difficult to resist the recommendation of malicious users, and ensure the credibility and accuracy of the results. To solve the above problems, on the basis of traditional filtering algorithm, introducing the trust relationship between users into the algorithm, and combining the similarity of the traditional filtering algorithm with the trust degree in weighted linear method, we propose the title. The simulate experiment results show that, compared with the traditional collaborative filtering recommendation algorithm, the proposed method not only improves the accuracy of the recommendation, but also ensures the credibility of the results.


 Sort-based Particle Swarm Optimization

WANG Xu-peng, ZHENG Kai

2016, 0(12):  12-15,21.  doi: 10.3969/j.issn.1006-2475.2016.12.003

Asbtract ( 109 )  

References | Related Articles | Metrics

 Traditional particle swarm optimization is prone to low pre-search capability, easy to get into early, shocks late near the optimal solution, and so on. In this paper, adaptive updated particles inertia weight ω, sorts particles which can use more useful information from other particles, each iteration adaptively updates particle flying time optimization three optimization schemes from different angles, while the optimization schemes may be combined simultaneously. Experimental results show that three kinds of particles optimization search precision and speed have been greatly improved. In particular, the combination effect of adaptive weight optimization and sorting optimization is very obvious. Adaptive flight time adjustment programs have significantly increased the rate of particle preliminary search.


 Wireless Channel Back-off Algorithm Optimization Design #br#   Based on RBF Neural Network

ZHANG Yan-nian

2016, 0(12):  16-21.  doi: 10.3969/j.issn.1006-2475.2016.12.004

Asbtract ( 107 )  

References | Related Articles | Metrics

 Through making research for the principle of channel access protocol and back-off algorithm, and aiming at the challenges and problems for wireless networks, the paper puts forward the wireless channel back-off algorithm optimization design based on RBF neural network. The optimization design reasonably regulates the back-off time slots interval in back-off algorithm according to the whole situation of the network to reduce the possibility of nodes waiting time for channel resources in the greatest extent, and make the node have a short period of time in a state of congestion. RBF neural network is mainly to complete the memory studying of the corresponding relationship between the network status and back-off time slot control. At the same time, we establish simulation scenario and learning samples in the OPNET Modeler, and make simulation and verification for the proposed optimization design. The simulation results show that the proposed back-off method makes communication nodes in the current network conditions choose the appropriate back-off time slot, so as to reduce the unnecessary waiting time effectively, and then communication time delay is reduced by 17.6% on average, which improves the network performance. All in all, the wireless channel back-off algorithm optimization design based on RBF neural network has more prominent effect.


 Multi-objective Swarm Intelligence Evolutionary Algorithm #br#   for Software Requirement Selection

SUN Hou-ju, YANG Shi-da

2016, 0(12):  22-28,33.  doi: 10.3969/j.issn.1006-2475.2016.12.005

Asbtract ( 111 )  

References | Related Articles | Metrics

 The requirement choice is of a great difficulty when updating the software release, and it will have a certain degree of influence on the cost and customer satisfaction of the next release of the software. For software developers, two contradictory goals need to be achieved at the same time: first, the cost of the selected requirements in the next release of the software should be the lowest; second, those customers of different importance to the company and having their own different views toward this requirement, ought to be satisfied as far as possible. Much attention should be paid to the multiple interaction constraints between requirements. This problem is known as “Next Release Problem” (NRP). To solve this problem, the academic literature, which has been published, lacks comprehensive consideration of all types of constraints, and in most researches, this problem is reduced to a simple single-objective optimization problem. We present a multi-objective swarm intelligent algorithm for software requirement optimization. It makes up for the deficiency of the existing researches and draws the bird mating strategies. Besides, it can adjust the parameters according to the scale of the problem in order to better apply it to the different scale of the problem. The results show that this algorithm can produce high-quality and effective solutions. By comparing with other different algorithms, it can be learned that the solution produced by this algorithm is better than the others relating to this field are.


 Improvement on Activation Functions of Recurrent Neural Network Architectures

YE Xiao-zhou, TAO Fei-fei, QI Rong-zhi, ZHANG Yun-fei, ZHOU Si-qi, LIU Xuan

2016, 0(12):  29-33.  doi: 10.3969/j.issn.1006-2475.2016.12.006

Asbtract ( 164 )  

References | Related Articles | Metrics

  Recurrent neural network has the advantage of learning long term dependencies, in contrast with other deep learning network architectures. However, the problems of vanishing and exploding gradients seriously obstruct the transmission of information over time, resulting in the deviation of learning long term dependencies. Hence, a great deal of studies focus on the adaption of classical recurrent neural network architectures. In this paper, we analyse the effect of two types of activation function for basic RNN and RNNs with gating mechanism. An improved model based on the basic RNN structure is proposed. The improved gating mechanisms of LSTM model and GRU model are proposed. Experiments on PTB classical dataset LMRD feeling classified dataset show that the improved models are advanced than traditional models and greatly improve the learning ability of the models.


 Optimal Tracking Control Based on Gradient Estimation Algorithm

YAO Qing-hua, HE Yong-jun, GUO Zhen-jiang, FU Dong-xiao

2016, 0(12):  34-37.  doi: 10.3969/j.issn.1006-2475.2016.12.007

Asbtract ( 131 )  

References | Related Articles | Metrics

 Based on the gradient estimation algorithm and adaptive dynamic programming, this paper solved the optimal control problem online. At first, regarding to the nonlinear system, a performance index was proposed. Then a Hamiltonian(HJB) function was constructed and a neural network(NN) was used to approximate the performance index. Another neural network was proposed to approach the actor, and both critic and actor NN weights are estimated based on gradient estimation online and simultaneously. Furthermore, steady-state control and robust term were designed to obtain the robust optimal control. At last, simulation results proved the effectiveness of the proposed methods.


 Recommendations Based on Logistic Regression by Exploiting Life-stage Change of Users

FU Quan-xing, HAN Li-xin, YANG Yi

2016, 0(12):  38-41.  doi: 10.3969/j.issn.1006-2475.2016.12.008

Asbtract ( 101 )  

References | Related Articles | Metrics

 Recently, the large impact of life stage on consumer’s purchasing behaviors has not taken into consideration. In this paper, we introduce a recommender system based on life-stage. Firstly, we label the life stage of the consumer by the model of lif-stage, then rank the candidate products with the predicted life stage by joint probability model and then recommend proper products. In the domain where the gap of the life stage is deterministic, we develop an efficient solution to do prediction using multinomial logistic regression algorithm. Our experiment results show that the proposed approach improves the accuracy of the recommendation.


 Solution for High Performance of Distribution Network Line Loss Big Data

SUN Li-hua1, HU Mu1, MENG Qing-qiang1, QIAN Ya-kang1, WONG Song2

2016, 0(12):  42-46,50.  doi: 10.3969/j.issn.1006-2475.2016.12.009

Asbtract ( 133 )  

References | Related Articles | Metrics

 In allusion to the problem in the distribution network line loss analysis, this paper proposed a solution for data of sales and distribution integration based big data technology, and introduced the key technology for sales and distribution data synchronizationss, the line loss data storage and transformer line loss data computing. At the same time, aiming at Hadoop analyzing the line loss data analysis not suitably, we proposed a solution of data update, data feedback, and large data interactive query and so on. Tests proved that the systems performance increased 6 times which can provide the technology refer for Hadoop big data technology applying in distribution network line loss computing and other types of power fields.


 Speeding K-NN Classification Method Based on Data Block Mixed Measurement

DENG Xi-hui, ZHAO Li

2016, 0(12):  47-50.  doi: 10.3969/j.issn.1006-2475.2016.12.010

Asbtract ( 112 )  

References | Related Articles | Metrics

 This paper presents a K-Nearest Neighbor (KNN) method based on data block mixed measurement, called KNN_DBM², in order to solve the problem that the low training efficiency and cannot solve the large scale problems of normal K-NN because it needs compute the distance between the sample to be tested and the labeled samples in the new sample classification prediction process. By introducing the data block mixed measurement into the prediction process of K-NN, this method divides the labeled samples into many various data blocks, and the mixing degree and center of these blocks are computed. When the new sample to be tested is produced, all the distances between this sample and all the centers of data blocks are computed and the nearest k data blocks are extracted. If all these k data blocks are purity, then label the sample to be tested according to the centers label and adopting a minority to obey the majority. But if the mixed data blocks are existed in these blocks, the distance between the sample and all the samples in mixed data block is calculated, and the distance from the center of the other pure data blocks is also calculated, then label of the sample to be tested by the k nearest sample or centers. By this data block dividing and mixed measurement method, it reduces the number of distances between the sample to be tested and the other labeled samples and obtains the high prediction efficiency synchronously. The experiment results demonstrate that the proposed KNN_DBM² model can obtain the high learning efficiency and testing accuracy simultaneously.


 Key Technologies and Applications of Distributed Memory Power Grid

ZHANG Chun-ping1,2, YANG Bo3, YANG Zhi1,2

2016, 0(12):  51-56.  doi: 10.3969/j.issn.1006-2475.2016.12.011

Asbtract ( 153 )  

References | Related Articles | Metrics

  With the power information system integration degree gradually increasing, the management scope further extended to the low pressure so as to result in complex resource model and massive resources data, which made traditional grid model and storage based on relational database be unable to meet the development needs in grid resource access efficiency, data calculation, power grid history section management, and multiple aspects. Due to the universal design of the current mainstream distributed memory computing products, when they are applied directly to the business systems of State Grid Corporation, they are unable to play the biggest advantage, even cannot solve the problem of engineering application. Distributed memory grid is based on memory computing, distributed computing and other cutting-edge technology, through deeply fusing network business, we research and construct a set of distributed memory computing technology and products which meet the grid resources processing requirements and are of high reliability, strong extensibility, in line with the business characteristics of State Grid Corporation. The experimental results show that the distributed memory grid can improve the processing power of the grid resource data, and greatly improve the processing efficiency.


 Research on High Performances and Parallel Processing for Network Packet

HUANG Yi-bin, JIN Qian-qian, JI Yuan

2016, 0(12):  57-61.  doi: 10.3969/j.issn.1006-2475.2016.12.012

Asbtract ( 165 )  

References | Related Articles | Metrics

 When using multicore processor to deal with high-speed network packets, it often faces poor performance and low CPU utilization. By analyzing the bottleneck of the packet processing on high-speed network, a model which can parallel process network packet is proposed. This model uses multi-queue NIC, along with parallel TCP/IP stack, multicore and multithread technology, and lock-free program technology to achieve a completely parallel processing in the whole network packet path. Test results indicate that this method can utilize CPU completely, and greatly improve efficiency of network packet processing.


 Research on Network Public Opinion Supervision Forecast Based on Big Data Analysis

WANG Guo-hua

2016, 0(12):  62-66.  doi: 10.3969/j.issn.1006-2475.2016.12.013

Asbtract ( 137 )  

References | Related Articles | Metrics

 Accurate regulation and prediction for network public opinion is the key technology of big data network information statistics and investigation. It is of great significance in terms of keeping network order stability. In view of the traditional method of using correlation feature matching method to forecast public opinion, as network public opinion data distribution scope is bigger, interference is stronger, there is a problem of poor regulatory prediction performance. The paper proposes a network public opinion supervision forecast algorithm based on large data semantic feature analysis, builds wide stationary time series model of network public opinion, uses binary representation methods of semantic information to the semantic information characteristics sequence fitting, constructs and matches thesaurus of network public opinion, adopts big data analysis method to improve the semantic feature analysis and extraction of the network public opinion, implements the improvement of prediction algorithm. Simulation experiments show that the method of network public opinion prediction is of higher accuracy, good convergence and high real-time performance, improves the regulation ability of network public opinion.


 Classification Analysis and Discussion of Collaborative #br#   Filtering Recommendation Technology

LIANG Xiang-yang, ZHANG Bo-lun

2016, 0(12):  67-72.  doi: 10.3969/j.issn.1006-2475.2016.12.014

Asbtract ( 134 )  

References | Related Articles | Metrics

 Facing with the worsening problems of information overload, recommendation system is widely utilized on Internet. Collaborative filtering is one of the most successful recommendation technologies in recommendation system. According to the classical theories and the latest researches, there exist three key problems, namely cold-start, sparsity and scalability in collaborative filtering recommendation. For these limitations, we described and concluded the research status, which based on users cold-start, items cold-start, rating forecast, similarity calculation, social recommendation, matrix factorization, clustering, cloud computing, etc. Finally, the research hotspots of collaborative filtering were predicted.


 An Adaptive Visual Odometry Algorithm Based on RGBD Sensors

DUAN Shan-shan1, LI Xin2

2016, 0(12):  73-77.  doi: 10.3969/j.issn.1006-2475.2016.12.015

Asbtract ( 142 )  

References | Related Articles | Metrics

 For the problem of three-dimensional path estimation about mobile robots in unknown environment, a kind of calculation method of scene adaptive visual odometry based on RGBD sensor was proposed. Firstly, the masking edge feature point, RGB edge feature points, feature point of ORB adaptively was selected by the texture information of the scene in the density. Secondly, three equation systems were constructed by the KTL algorithm matching depth information of the corresponding feature points in the target frame and the reference frame. In order to improve the pose accuracy, the LM algorithm is used to minimize the projection error of the corresponding feature points. Finally the bundle adjustment is applied to optimize the robots overall trajectory. The experimental results show that, in the indoor scene with rich texture, the performance of the method is equivalent to RGBD SLAM, whereas in the scenes with sparse texture (FR2 slam, FR2 slam2) and no texture(FR3 str_notxt), the proposed method outperforms the RGBD SLAM algorithm. The offset A-RGBD SLAM algorithm of Transl. RMSE, Rot. RMSE, ATE RMSE were (54%, 66%, 54%; 43%, 77%, 31%; 60%, 43%, 81%) of RGBD SLAM algorithm.

 A New Digitization Method of Hydrological Yearbook

CHEN Wan-wan1, LI Shi-jin1, HU Jin-long2, GAO Xiang-tao2

2016, 0(12):  78-82,86.  doi: 10.3969/j.issn.1006-2475.2016.12.016

Asbtract ( 173 )  

References | Related Articles | Metrics

 Flood forecasting analysis requires to input a lot of images. Due to the gradual wearing and aging of paper hydrological yearbooks as well as the high-cost and high error rates of traditional methods, the research on the digitalization of hydrological yearbook data is of a great importance. We propose a digital recognition method with multi-feature fusion based on layout analysis, and provide a post-correction mechanism according to the relevance of time series. For hydrological yearbook pictures, experimental results show that our method achieves high recognition accuracies and is practical to restore the true values of misrecognitions.


A Multi-sensor Image Fusion Method Based on Contourlet Transform

YU Xiang-qian

2016, 0(12):  83-86.  doi: 10.3969/j.issn.1006-2475.2016.12.017

Asbtract ( 107 )  

References | Related Articles | Metrics

 Single sensor can not fully describe image information, and multi-sensor image fusion can get more image information. Aiming at the edge fuzzy, important information loss and other issues in image fusion methods, a multi-sensor image fusion method based on Contourlet transform is put forward. Firstly, the paper uses Contourlet transform decomposes the acquired image in multi-scale and multi-orientation, then use different rules respectively to fuse the low-pass and high-pass sub, and finally the fused image is reconstructed by using inverse Contourlet transform. The results show that image of the proposed method is more natural, and image quality and fusion result are better than other image fusion methods.


 Scene Text Localization Algorithm Based on Region Feature and SVM 

PENG Yan-bing1,2， GUAN Yun-zhu1

2016, 0(12):  87-91.  doi: 10.3969/j.issn.1006-2475.2016.12.018

Asbtract ( 115 )  

References | Related Articles | Metrics

 Scene text localization and recognition problem has received great attention in recent years. This paper proposes a method based on the combination of region and connected domain for the character of the scene. First, the algorithm combines the maximally stable extremal regions (MSER) with stroke width transform (SWT) to extract candidate text connected components. Then, part of the non-text region is filtered by using the heuristic rules of the candidate components. At last, the histogram of oriented gradient (HOG) and the uniform local binary patterns(LBP) of the candidate text connected components are extracted to be the input of SVM classifier. In that way, text components and no-text components can be distinguished. The experimental results show that the algorithm has good performance in text localization, and the recall rate is high.


 Improvement of ZigBee Routing Algorithm and Its Application in Smart Home

JING E-lin

2016, 0(12):  92-96,101.  doi: 10.3969/j.issn.1006-2475.2016.12.019

Asbtract ( 146 )  

References | Related Articles | Metrics

 A smart home environment monitoring system is proposed in this paper based on the ZigBee technology, namely the ARM-Linux is used as on-site control center, through the ZigBee network technology, the equipments of the intelligent home system consist of a tree topology. The real-time monitoring of the situation in the home is carried out through the sensor and CMOS camera on the terminal node. In addition to obtaining the environment parameters in the home, this system also needs to observe whether there is a stranger into the family or not through real-time monitoring. In the data transmission of traditional ZigBee single path routing, control packet overhead and network delay are relatively large. When the load is large, there will be problems such as network congestion, especially in the transmission of video information, the larger network transmission delay will not guarantee the quality of video service. Based on original ZigBee tree routing algorithm, we proposed a multipath routing algorithm: transmitting data between source node and coordinator node at the same time through multiple paths, effectively increases the data transmission rate and the system throughput, and reduces transmission delay and data packet loss. In addition, the energy consumption of the nodes is balanced by pre-setting the node energy threshold, so that the failure of the whole network does not happen because of premature death. Experimental results show that the algorithm can effectively solve the problem of data loss and energy consumption in the home monitoring system, which can guarantee the transmission quality of the video, and prolong the network lifetime.


 Remote Control Solution for Multi-function Intelligent Device Based on WebSocket

CHEN Jian-gang, HUANG Guo-wei, CAI Hong-xin, TAN Guo-long

2016, 0(12):  97-101.  doi: 10.3969/j.issn.1006-2475.2016.12.020

Asbtract ( 184 )  

References | Related Articles | Metrics

 This paper presents a remote control solution for mobile intelligent devices (such as smart cars, robots) based on the open source hardware—WRTnode with WebSocket protocol and WeChat public accounts H5. WorkerMan framework is used for WebSocket Server. Intelligent devices, installed modified LibWebSocket plugin, is connected with WebSocket Server. The WeChat users whose OpenIDs bind to the device can control the device through the H5 interface in WeChat public accounts. The control instructions include moving instruction and switching command, video image display, sensor data (such as temperature value) reporting. Separate transmission of video data and control instruction data are completed through dual WS Server.


 Filtering and Application of Aged Information Based on AdaBoost Algorithm 

CHENG Guang-yang1, LIAN Bin2

2016, 0(12):  102-106,110.  doi: 10.3969/j.issn.1006-2475.2016.12.021

Asbtract ( 125 )  

References | Related Articles | Metrics

 Facing attention to the needs of older persons in the information society for aged information, this paper uses Web crawler technology based on Python to crawl the news and official documents from Ministry of Civil Affairs website. Aiming at the characteristics of news on portals, the paper optimizes data fetching process as well as the training set, uses Adaboost algorithm to train a given collection of text and get filtering model. And the paper provides an effective feature selection method which uses the χ2statistic principles, effectively reduces the feature dimension, and then uses this model to filter the collection information to get aged information. Finally, the results of information filtering are analyzed. The experimental analysis results show that the proposed method can effectively filter the aged information and meet the elderly demand of aged information acquisition in the practical application.


 Planting Management APPs Based on Mobile Intelligent Terminal

QIU Feng1,2， LIU Bo-ping1,2, YANG Guo-qiang1,2, FU Kang1,2, WANG Lei1,2

2016, 0(12):  107-110.  doi: 10.3969/j.issn.1006-2475.2016.12.022

Asbtract ( 132 )  

References | Related Articles | Metrics

  Mobile intelligent terminal gets rapid popularization and application in various fields because of its characteristics of intelligence, convenience and not limited to time or space. Aiming at the application of mobile intelligent terminals in planting management, a set of planting management APPs based on mobile intelligent terminal is designed and implemented by using Restful Web Service and Qr code technology. At last, the paper introduces the design and implementation method of the software.

 Design and Implementation of Fire Inspection System Based on WeChat Public Platform

LI Xiao-lei1, LYU Ming2, XU Jin3, ZHAO Gao-peng1, ZHANG Jie1

2016, 0(12):  111-114,121.  doi: 10.3969/j.issn.1006-2475.2016.12.023

Asbtract ( 183 )  

References | Related Articles | Metrics

 Through analyzing the police requirements of Nanjing Xuanwu public security bureau, this paper adopts .NET, HTML, jQuery, CSS3 and other Web technologies to design a kind of fire inspection system based on WeChat public platform. The system has realized the data upload from the mobile end, information query, message delivery, and other functions, as well as the PC business audit, and other functions. This paper introduces the framework, function and structure of fire inspection system, and then introduces WeChat public interface docking platform, the construction of server-end, the interaction of frontend and backend data, and backend database query process, and concrete implementation method, etc. This system effectively strengthens the security surveillance of the police for enterprises or institutions, and reduces the workload of the police.


  Design of 8 Channels VGA Long-distance Transmission System Based on FPGA

SHI Zhong-meng, HU Xiao-ji

2016, 0(12):  115-121.  doi: 10.3969/j.issn.1006-2475.2016.12.024

Asbtract ( 117 )  

References | Related Articles | Metrics

 Based on the need of multi-channel long-distance military transmission system, this paper studied the long-term transmission function and multi-switching function. By studying the current problems in military switch matrix, and proceeding from the chip-level design, a set of sending and receiving devices is developed. To analyze the current problems of switch matrix in depth, we studied the differential cable transmission technology. Based on the use of FPGA hardware platform and the differential transmission technology, we combined the long-term transmission with the differential switching matrix innovatively to design an eight channels VGA long-distance transmission system. The results showed that the system improved the performance of switch matrix and the signal quality in long-distance transmission.