Table of Content

18 May 2015, Volume 0 Issue 5

Previous Issue    Next Issue

Human Activity Recognition Based on Combined SVM&HMM

SU Hong-yu1, CHEN Qi-an1, WU Hai-tao2

2015, 0(5):  1-8+12.  doi:10.3969/j.issn.1006-2475.2015.05.001

Asbtract ( 311 )  

References | Related Articles | Metrics

Being able to recognize human activities is essential for several intelligent applications, including personal assistive robotics and smart homes. In this paper, we perform the recognition of the human activity based on the combined SVM&HMM in daily living environments. Firstly, we use a RGBD sensor (Microsoft Kinect) as the input sensor, and extract a set of the fusion features, including motion, body structure features and joint polar coordinates features. Secondly, we propose a combined SVM&HMM model which not only combines the SVM characteristics of reflecting the difference among the samples, but also develops the HMM characteristics of dealing with the continuous activities. The SVM&HMM model plays their respective advantages of SVM and HMM comprehensively. Thus, the combined model overcomes the drawbacks of accuracy, robustness and computational efficiency compared with the separate SVM model or the traditional HMM model in the human activity recognition. The experiment results show that the proposed algorithm possesses the better robustness and distinction.

A FCM-based Domain Partition Method for Time Series Data Set

LIU Che, LIU Xuan, MA Hong-xu

2015, 0(5):  9-12.  doi:10.3969/j.issn.1006-2475.2015.05.002

Asbtract ( 312 )  

References | Related Articles | Metrics

With the development of society, people increasingly demand for data prediction, fuzzy time series has been attracted much attention because it can handle the ambiguities in time series data. To improve the prediction accuracy of the model, domain partition is crucial as the first step of the prediction. In this paper, we present a FCM-based two-time domain partition method. We make the first partition according to the clustering center obtained from FCM clustering, the second partition to realize unequal domain according to the density distribution of the sample points. Finally, the feasibility of the new method is verified through the prediction of classic sample.

A Novel Assembled Text Clustering Algorithm Using Differential Evolution and SOM

JIANG Kai, YUAN Jin-hai

2015, 0(5):  13-16+20.  doi:10.3969/j.issn.1006-2475.2015.05.003

Asbtract ( 194 )  

References | Related Articles | Metrics

Self-organizing map (SOM) is an important clustering model, which can effectively improve the accuracy of search engine. But it is sensitive to the initial connection weights. After analyzing the drawbacks of the self-organizing map algorithm, a novel assembled text clustering algorithm (IDE-SOM) based on improved differential evolution and self-organizing map is proposed. Firstly, the improved differential evolution is introduced to realize coarse clustering in the document feature set with the purpose of getting an optimized initial connection weights. Then the SOM algorithm is initialized to realize fine clustering using the initial connection weights. Finally, the experiment is conducted and the results illustrate the better clustering performance of the proposed hybrid approach in terms of the value of F-measure and entropy.

Network Hot Event Mining Algorithm Based on Feature Extraction

LI Wei-yao, ZHAO Kai

2015, 0(5):  17-20.  doi:10.3969/j.issn.1006-2475.2015.05.004

Asbtract ( 291 )  

References | Related Articles | Metrics

For effectively mining the hot issues and topics concerned by people in network, improving the capabilities of data classification and the correct rate of hot tracking and detection, basing on analyzing the problems existing in the traditional unstructured mining algorithms used currently, we proposed a mining algorithm based on structured segmentation. First, by analyzing the hot events mining process, we designed a semi-structured features extraction algorithm for hot events data mining, in order to make feature segmentation for semi-structured data, generate a lot of requests, and then get hot event data allocation factors, thereby improve mining properties. Simulation results show that the algorithm is running with high efficiency, good accuracy and high robustness.

An Intelligent Particle Swarm Optimization Algorithm Based on Fitness Analysis

LI Cong1, SONG Wen-long2

2015, 0(5):  21-24+70.  doi:10.3969/j.issn.1006-2475.2015.05.005

Asbtract ( 198 )  

References | Related Articles | Metrics

When solving multi-dimensional multi-peaks function problems, the standard particle swarm optimization algorithm is not good at local search with high accuracy and its global search ability is also weak, besides, its convergence rate is slow. Therefore, an intelligent particle swarm optimization algorithm based on fitness analysis is presented, into which the “local fitness parameter”, the “global fitness parameter” and the “coordinate rotation method” are introduced. Three multi-dimensional multi-peaks functions are tested to show that this algorithm not only has the ability of local and global search, but also demonstrates faster convergence rate.

Linpack Test of High Performance Computing Cluster and Its Application in Big Data

HAN Fei, SUN Zan-dong, SU Hui

2015, 0(5):  25-29.  doi:10.3969/j.issn.1006-2475.2015.05.006

Asbtract ( 274 )  

References | Related Articles | Metrics

High performance computing cluster for efficient parallel computing is of very high cost performance and good scalability. How to test and evaluate the cluster system performance becomes a key problem. In this paper, we made a Linpack test based on high performance computing cluster of 6 nodes. Different matrix scale N, computational nodes, block size NB of matrix, processor grid topology and network communication were tested. Computing performance to PC and cluster were compared and cluster performance was tested. The results show that: the cluster parallel computing performance is good and is of strong scalability, but the hardware communication ability should be further improved. The cluster is applied into the computing of the actual seismic big data. It is known that the parallel computing performance of the cluster has been improved greatly.

Extension Model Modeling for Oracle Bone Inscriptions Based on HowNet

LI Zhi-yong1, GAO Feng2

2015, 0(5):  30-34.  doi:10.3969/j.issn.1006-2475.2015.05.007

Asbtract ( 249 )  

References | Related Articles | Metrics

Oracle bone inscriptions (OBI) basic information arrangement is the bottleneck of OBI information processing, and the OBI language automatic modeling is the meaningful method to resolve this problem. An automatic modeling technique of extension model is introduced to support OBI information processing. A semantic database composed of OBI and modern Chinese is built based on HowNet. There are three concept tables are used to train the OBI text. They are matter-element concept table, affair-element concept table and relation-element concept table using HowNet elements. The base-element model with a certain degree of maturation is extracted from the OBI text, and using the base-element extentics transformation rules it can realize extension reasoning. The experiment results show that the proposed OBI language model is helpful to the modern Chinese research. As well as it can provide solutions for the semantic deduction of unidentified OBI characters and the text content integration of broken Oracle bones in OBI information processing.

Application of DDS in Aircraft Collaborative Design System

BIAN Hua-xing, CHEN Dan, ZHUANG Yi

2015, 0(5):  35-39.  doi:10.3969/j.issn.1006-2475.2015.05.008

Asbtract ( 234 )  

References | Related Articles | Metrics

By introducing the concepts of data distribution services, aiming at the characteristics of aircraft collaborative design, we designed a DDS-based aircraft collaborative design data services middleware, and applied it to the aircraft collaborative design system. The middleware is of loose coupling and is easy to expand. It meets the needs of the system and realizes the information exchange among various professional software systems. In this way, it shorts the cycle of aircraft design effectively.

Single Image Haze Removal Algorithm

CHEN Wei, FAN Xin-nan, LI Min, WANG Geng-ren

2015, 0(5):  40-47+52.  doi:10.3969/j.issn.1006-2475.2015.05.009

Asbtract ( 180 )  

References | Related Articles | Metrics

At present, most of the haze removal algorithms ignore the color distortion problem, especially in areas with high brightness. This paper proposes a single image haze removal algorithm using the bilateral filtering to ensure the image detail and color information at the same time. First of all, the Butterworth low-pass filter is used to directly estimate the background air light in order to avoid the errors caused by different sizes of images and different blocks of images. Then, the bilateral filter is used to estimate the atmospheric light curtains instead of the soft matting process in He’s algorithm to obtain the optimum transmission diagram and improve the computational efficiency. Finally, the brightness gain function is used to restore color in the HSV space. Compared with several classical algorithms, simulation experimental results verify the color calibration and the effectiveness of the proposed algorithm.

Image Reconstruction Based on Improved CoSaMP Algorithm

LIU Ji-cheng1, WANG Min-ying1, LI Hao-ran2

2015, 0(5):  48-52.  doi:10.3969/j.issn.1006-2475.2015.05.010

Asbtract ( 207 )  

References | Related Articles | Metrics

Reconstruction algorithm is one of the points of compressed sensing. It determines compressed sensing whether can be applied in the real system. The paper combines compressive sampling matching pursuit (CoSaMP) algorithm with weighted block matrix technology in order to improve the accuracy of compressed sensing reconstruction, shorten the processing time and make the original algorithm more perfect. The paper compares the improved algorithm with the existing algorithms by simulation experiments which approve that the new algorithm improves the reconstruction quality under the same sparse condition.

Image Denoising Based on Shearlet Transform and Improved Anisotropic Diffusion

LI Lei, REN Yue-mei

2015, 0(5):  53-56.  doi:10.3969/j.issn.1006-2475.2015.05.011

Asbtract ( 239 )  

References | Related Articles | Metrics

To suppress the pseudo-Gibbs in multi-scale geometric analysis denoising and alert the staircase effect in anisotropic diffusion effectively, an image denoising method combining shearlet transform and improved anisotropic diffusion is proposed. Firstly, the noisy image is decomposed by shearlet transform, and different scale coefficient matrices are obtained. Then, the improved anisotropic diffusion equation is applied to process these shearlet coefficients, which realized a new anisotropic diffusion denoising based on multi-scale analysis of the image. Finally, the clean image is obtained from the processed shearlet coefficients. Experiments show that the proposed method can suppress the pseudo-Gibbs and staircase effect, as well as reduce the noise and preserve the detail better.

An Acquisition and Display Method for Multi-view Image Based on 3D Models

LONG Ji-chang1, MA Wei-hua1, SHEN Chun-lin2

2015, 0(5):  57-61.  doi:10.3969/j.issn.1006-2475.2015.05.012

Asbtract ( 178 )  

References | Related Articles | Metrics

At the issue of rare data source and high production cost of multi-view-image in 3D model display, this paper designs and implements a new common method of data collection and display for the existing three-dimensional models. This method can be applied to any application developed based on OpenGL/Direct3D, we can get multi-view images from application by intercepting and modifying its calls. This paper also presents a synchronous rendering protocol, which guarantees the high real-time in this acquisition and display method. Experimental results show that the proposed method is of high acquisition and rendering process, the image quality is the same as the source, with the extensive application and the research values in the field of three-dimensional display.

Automatic Recognition Method of Bridges Based on Infrared Image

WEN Xiong-zhi, CHEN Xue-gong, WANG Jing, TAN Juan

2015, 0(5):  62-66.  doi:10.3969/j.issn.1006-2475.2015.05.013

Asbtract ( 159 )  

References | Related Articles | Metrics

The research on recognition of bridges is of a very important significance in military and civil. The relevant studies at home and abroad have obtained certain effects, but have also some weak points: the application range is not wide, the processing speed is not so fast and so on. Based on the characteristics of infrared image, this paper puts forward a method that uses iteration to calculate and split threshold to get the binary images. Then the target points are extracted by morphology method. Based on bridge’s linear property, the straight lines are extracted and connected from the target points by Radon transform. This algorithm is universal for bridge recognition of infrared image in water area, and it’s simple, intuitive with a high calculating speed.

A Composition Forecasting Approach of Customer Repeat Purchasing

SHU Fang, MA Shao-hui

2015, 0(5):  67-70.  doi:10.3969/j.issn.1006-2475.2015.05.014

Asbtract ( 207 )  

References | Related Articles | Metrics

SMC model was proposed by Schmittlein et al., which is very popular in customer base analyses under non-contractual settings. The SMC model assumes that customers’ purchasing follow Poisson processes until they defect. The HIPP model was proposed by Ma et al., which assumes that customers do not defect, but instead they switch constantly between active and inactive state. In reality however, both permanent and temporary defecting may exist. Therefore, SMC model may underestimate customer’s probability of re-purchasing, while HIPP model may overestimate that. Based on two models, in this paper, we propose a composition forecasting approach which combines the forecasts from both SMC and HIPP. The optimal weights of compositing are calculated by a genetic algorithm. Empirical studies show the superiority of the proposed composition approach.

Location-pricing Model and Algorithm for Waste Household Appliance Returned Logistics Network

DI Wei-min, ZHAO Dong

2015, 0(5):  71-75+80.  doi:10.3969/j.issn.1006-2475.2015.05.015

Asbtract ( 258 )  

References | Related Articles | Metrics

To acquire reasonable waste household appliance returned logistics network, a waste returning satisfaction function for consumers was constructed based on prospect theory and geometric average synthesis method, and a joint location-pricing decision model for this kind of returned logistics network was developed. With the help of the model, the following items could be determined, such as optimal sites for collection station and return center, purchase price for each collection station and favorite collection station for each consumer. To solve the model, a hybrid genetic algorithm embedded greedy algorithm and exact algorithm of linear programming was presented, and its realization steps were explained. Finally, effectiveness of the model and its algorithm were verified by an example.

Drawing and Optimization of Contour Lines Based on Grid DEM

GONG Heng, HE Yuan-qing

2015, 0(5):  76-80.  doi:10.3969/j.issn.1006-2475.2015.05.016

Asbtract ( 176 )  

References | Related Articles | Metrics

According to the large amount of DEM data, in order to enhance the data readability and improve drawing efficiency, we optimized the traditional process of drawing the contour lines. The entire process was divided into three operations: Data preprocessing, building height index, mapping in accordance with the index. First of all, by collecting and simply processing the original DEM (digital elevation model) data, we extracted the key information related to mapping in the data, such as the latitude, longitude, maximum height and minimum height of the endpoints left-up and right-down in the whole map. Secondly, based on data preprocessing, we analyzed the data again through the equivalent point judging, the equivalent point linear interpolation and the equivalent point tracking to establish the strongly readable index files of height. At last, according to the height index data, the data structure was constructed to reasonably store data, so it could help the computer fast drawing contour map.

Pharmacy Experiment Teaching System Based on Idea of M-learning

HU Ju-dong1, LU Tao2, LIAO Jun1

2015, 0(5):  81-84+89.  doi:10.3969/j.issn.1006-2475.2015.05.017

Asbtract ( 189 )  

References | Related Articles | Metrics

To apply the idea of m-learning to pharmacy experiment teaching, the status and characteristics of pharmacy experiment teaching were analyzed. On the basis of surveying the demand of the combination of m-learning and pharmacy experiment teaching, a pharmacy experiment teaching system was designed and developed. The system was based on the technology of Web Service and used Android as main platform. Finally, the research gave the system application example of organic chemistry experiment report in pharmacy experiment courses. It proves that the application of the m-learning idea and technology in pharmacy experiment teaching is feasible.

Neural Network Blind Equalization Algorithm Under Dynamic Correction

ZHAO Hui-qing

2015, 0(5):  85-89.  doi:10.3969/j.issn.1006-2475.2015.05.018

Asbtract ( 175 )  

References | Related Articles | Metrics

Existing neural network algorithms have the problems of slow convergence speed and low accuracy. In response to this phenomenon, we study traditional neural network blind equalization algorithms and feed forward neural networks, and present a feed forward neural network blind equalization algorithm with the effect of automatic correction. The convergence speed of the algorithm is improved by improving the cost function and iterative step factor. The acquired information is corrected by correcting the acquired target signals. Experimental results show that they are basically consistent with the expected results, and it is of the advantages of high reliability and fast convergence speed.

Joint Detection Algorithm for SDMA-MIMO Systems Based on GA

DU Juan, WANG Qi, FANG Wei-wei

2015, 0(5):  90-93.  doi:10.3969/j.issn.1006-2475.2015.05.019

Asbtract ( 197 )  

References | Related Articles | Metrics

To improve the overall system performance for SDMA-MIMO, the joint MMSE multi-user detection and signal demodulation based on genetic algorithm is proposed. The algorithm values the demodulated user information as the initial population of the genetic algorithm, so the accuracy of multi-user detection algorithm is improved, and the system performance is improved after several iterations. Simulation results show that at the modest complexity, compared to MMSE detection based genetic algorithm, MMSE-GA MJD can achieve better performance.

Network Traffic Prediction Model Based on Mixed Kernels RVM Optimized by CS Algorithm

CHEN Jing-zhu

2015, 0(5):  94-97.  doi:10.3969/j.issn.1006-2475.2015.05.020

Asbtract ( 198 )  

References | Related Articles | Metrics

In order to improve the prediction accuracy of network traffic, a network traffic prediction model is proposed based on cuckoo searching algorithm optimizing the parameters of mixed kernel relevance vector machine (CS-RVM) to solve limitations of single kernel function for relevance vector machine. Firstly, the polynomial and Gaussian kernel functions are produced to mixed kernel function for the relevance vector machine, and then the cuckoo searching algorithm is introduced to optimize the parameters of hybrid kernel function, finally network traffic prediction model is established based on the relevance vector machine using the optimal parameters. The simulation results show that, CS-RVM model is of good effect and could improve the prediction accuracy of network traffic.

An Image Watermarking Algorithm Based on Multi-scale Space

LI Ai-hua, WANG Li-bin

2015, 0(5):  98-102+126.  doi:10.3969/j.issn.1006-2475.2015.05.021

Asbtract ( 197 )  

References | Related Articles | Metrics

The existing image watermarking algorithm is of good robustness when against some common image processing attacks such as Gaussian noise, median filtering and lossy compression, etc. However, the robustness of watermark is poor when against some geometric attacks such as cropping and scaling, etc. Therefore, in order to improve the robustness of watermark, an embedded watermarking method is proposed based on invariant feature by combining the scale invariant feature transform and visual model. In the detection scheme, the probability density function of wavelet coefficients which in mid-frequency sub-band is modeled by the generalized Gaussian distribution model and the watermark decision threshold is determined by using the binary hypothesis theory. Furthermore, the receiver operating characteristic (ROC) curve between the probability of false alarm and the probability of detection is derived. Experimental results demonstrate that the proposed method is of good robustness when against some attacks such as JPEG compression, additive white Gaussian noise, scaling, cropping and rotation etc.

Verifiable and Secure Protocol for Outsourcing Matrix Determinant Computation to Malicious Cloud

SHEN Yin-jie

2015, 0(5):  103-106.  doi:10.3969/j.issn.1006-2475.2015.05.022

Asbtract ( 241 )  

References | Related Articles | Metrics

Computation outsourcing to the cloud has become a popular application. This computing paradigm brings in some new security concerns and challenges, such as input/output privacy and result verifiability. A protocol is designed to enable secure, robust cheating resistant, and efficient outsourcing of matrix determinant computation to a malicious cloud. Client encrypts the original matrix and then transfers it to cloud server to compute. Analysis shows that the proposed protocol is correct, secure, and robust cheating resistant. Extensive theoretical analysis also shows its high-efficiency.

An Embedded Dynamic Security Networking Technology Based on Quick Jump and Trust

ZHANG Shu-feng

2015, 0(5):  107-111.  doi:10.3969/j.issn.1006-2475.2015.05.023

Asbtract ( 201 )  

References | Related Articles | Metrics

Safe and efficient networking of embedded systems was the integration development bottleneck for the next generation of embedded systems and network systems. In conventional networking technology of embedded systems, either the network security or network efficiency was ignored, so a better performance cannot be achieved. An embedded dynamic security networking technology based on quick jump and trust was proposed, and the multiple network nodes based on a random beating of fast switching were used to achieve optimal network search. When the Web search was completed, the trust mechanisms were taken to protect the security of the network system access, while achieving rapid networking and security. A set of 10-node was used to test the system, the result shows that with quick jump and trust, the systems security risk distribution can be accurately detected, enabling secure network, so it is of good network system security application.

Combination Conformal Spherical Parameterization Method Based on Specific Boundary

XIAO Xiong, LIU Hui-yi, ZHU Jie

2015, 0(5):  112-116+120.  doi:10.3969/j.issn.1006-2475.2015.05.024

Asbtract ( 173 )  

References | Related Articles | Metrics

This paper proposes a combination conformal spherical parameterization method based on the specific boundaries for 0-genus manifold triangular mesh model. The method uses mean value coordinates to calculate each vertex related weight on the original triangular mesh model, and then uses the weights to amend the vertices on the spherical domain to get new vertex coordinates which are of conformal character, and then transforms the new generated spherical domain barycenter to the center of sphere by Möbius transformation, after that the areal distortion degree can be reduced, finally we can get a higher accuracy initial spherical domain. Experimental results show that this method is stable and efficient with minor distortion.

Quad-rotor Aircraft Attitude Control System Based on PID Control

YE Shu-qiu, ZHAN Lin

2015, 0(5):  117-120.  doi:10.3969/j.issn.1006-2475.2015.05.025

Asbtract ( 259 )  

References | Related Articles | Metrics

The performance of quad-rotor flight has been largely determined by flight control system. By analyzing a flight model, and on the basis of this model, a kind of flight attitude control system based on PID control method was designed and implemented. The whole control system includes quaternion attitude resolving of the first stage (PI) and throttle calculation of the second stage (PID). By fusing the outputs of the acceleration sensor and velocity sensor with PI method, the current flight attitude was calculated out. By fusing the current attitude angle and target attitude angle with PID method, the output values of motors' throttle were set up. The host computer receives flight’s data through Bluetooth. Experimental results show that this system can well maintain the stable attitude of quad-rotor.

Access Database Login Authentication Systems Based on Various Types of Macros

ZHANG Wei-wei, GUO Feng-ying, HAN Ai-qing

2015, 0(5):  121-126.  doi:10.3969/j.issn.1006-2475.2015.05.026

Asbtract ( 232 )  

References | Related Articles | Metrics

For the login authentication problem in the development of the Access database application systems that non-computer professionals encountered, this paper describes the design ideas and methods, which using various types of macros to complete multi-user multi-role login authentication. Data macro function that Access 2010 newly added is first introduced to solve such problems.