Huifang Deng

South China University of Technology, Shengcheng, Guangdong, China

Are you Huifang Deng?

Claim your profile

Publications (20)3.47 Total impact

  • Weimin Peng, Huifang Deng
    [Show abstract] [Hide abstract]
    ABSTRACT: One mission of feature fusion is to obtain a complete yet concise presentation of all existing feature data by detecting and fusing the duplicate feature data. In contrast to the already developed feature fusion methods which have shown their limitations, this paper applies the theories of quantum information to feature fusion. Further, a novel and effective step-wise quantum inspired feature fusion method, which detects the duplicate feature data based on maximum von Neumann mutual information and fuses the duplicate feature data using the operations on quantum state, is developed. This same idea is also used for feature dimensionality reduction, and the corresponding models are investigated. For comparison, another quantum inspired feature fusion method based on average quantum phase is presented here. The experimental results show that the quantum inspired feature fusion method based on von Neumann entropy gives better results on completeness and conciseness than the method based on average quantum phase.
    Information Fusion 07/2014; 18:9–19. DOI:10.1016/j.inffus.2013.10.003 · 3.47 Impact Factor
  • Caifeng Zou, Huifang Deng, Qunye Qiu
    [Show abstract] [Hide abstract]
    ABSTRACT: A new model of hybrid cloud computing architecture based on cloud bus is proposed. The system is based on local private cloud, combined with one or more type(s) of public cloud(s). The internal structures of private cloud and public cloud are the same, including infrastructure and virtualization layer, cloud platforms layer, cloud bus layer, cloud application layer, the management center and storage centers. The layer of infrastructure and virtualization is designed to incorporate the underlying hardware resources into a virtual cluster, providing a variety of virtual resources to the upper layer. The layer of cloud platform is used to run Web applications or services, and carry application-specific development and application integration through its open interfaces. The cloud bus layer, consisting of a control bus, a number of node buses and adapters, is designed to manage and monitor the various services of the cloud platform layer. The proposed model of the architecture can accelerate the migration of the existing IT environments to cloud computing environments, reduce the investment, and make full use of IT resources.
    Proceedings of the 2013 IEEE 9th International Conference on Mobile Ad-hoc and Sensor Networks; 12/2013
  • Huifang Deng, Hao Cheng
    [Show abstract] [Hide abstract]
    ABSTRACT: This paper proposes a model which incorporates the traveler dynamics into the cell transmission model (CTM) by decoupling the travel demand from the traffic demand, in order to simulate traffic dynamics in a more controllable way. The model takes travelers as the source of the traffic load and formulates the relationship between these two demands mathematically. With the proper design, the model can be used as an enhancement to any existed cell transmission model. For implementation in this paper, an open source macro-simulation tool, called Aurora Road Network Modeler (AuroraRNM), is adopted. The experiments are conducted on a real road network with the help of this tool, and the results show the traveler oriented model can reflect the traffic dynamics properly and reasonably. Besides, based on the simulation results, the model also provides a way to find out the manageable status in a nonequilibrium traffic. This is more practical as though user optimal or system optimal solutions present optimal equilibrium traffic, such equilibrium status remain analytical and hard to apply to the real world management due to limited resources or other reasons.
    2013 International Conference on Connected Vehicles and Expo (ICCVE); 12/2013
  • Source
    Tao Chen, Huifang Deng
    [Show abstract] [Hide abstract]
    ABSTRACT: To deal with the two problems in image retrieval, i.e., the small number of query images, the ambiguity of an image - the image consists of many regions with different semantic meaning, in this paper, we proposed a novel method for image retrieval based on Bayesian multi-instance learning using unlabeled data, termed as Bayesian-MIL method, which treats the image retrieval as a binary classification problem. In this method, to obtain an approximate estimation of the class-conditional probability of positive images, a multi-instance learning algorithm is adopted to filter out background regions in positive images, and then a Bayesian classifier is constructed to rank the images from a large digital repository according to their score of posterior probability. Finally, the ranking top k images will be returned to users. Experimental results on COREL image data set have demonstrated the effectiveness and efficiency of the proposed approach.
    Proceedings of the 2nd International Conference on Computer Science and Electronics Engineering; 03/2013
  • Chunhui Deng, Huifang Deng, Qiping Ma
    Annual International Conference on Computer Games, Multimedia and Allied Technology; 05/2012
  • Huifang Deng, Zhen Liang, Chunhui Deng
    [Show abstract] [Hide abstract]
    ABSTRACT: With widespread applications of RFID (Radio Frequency Identification) technology, RFID data are growing at an exponential rate. More and more companies use Data Mining to find information which is the most relevant to the company business from mass of data. K-Means algorithm is one of the main algorithms of Data Mining. CUDA (Compute Unified Device Architecture) introduced by NVIDIA for GPGPU (General-Purpose computation on Graphics Processing Units) programming is a very powerful tool to accelerate K-Means. In this paper, we proposed a new set of parallel solution of K-Means algorithm called BG K-Means (Batch GPU-based K-Means) to implement K-Means based on GPU with CUDA. Compared with existing GPU-based K-Means algorithm, by taking so-called “batch” approach, BG K-Means makes full and rational use of CUDA's memories (shared memory, global memory, and constant memory) and reduces the access to data set. Hence, the speed of BG K-Means could reach as high as 55 times that of the CPU-based K-Means. Finally, we designed a system based on RFID data, and applied BG K-Means to data analysis.
  • Huifang Deng, Chunhui Deng, Jingjing Li
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, we first discussed the video decoding standard and its architecture, and then analyzed the decoding complexity of each process. By using the benefit of the CUDA programming model, and taking advantages of GPU to optimize the decoding process of MC (motion compensation) and CSC(color space conversion) that are very time consuming, we proposed a MC accelerating method based on CUDA, and a CSC accelerating method based on CUDA and OpenGL shader. The experiments show that it is feasible to decode high definition video in real time using GPGPU-CUDA.
    Intelligent Information Hiding and Multimedia Signal Processing (IIH-MSP), 2012 Eighth International Conference on; 01/2012
  • Source
    Yuan-Dong Lan, Huifang Deng, Tao Chen
    [Show abstract] [Hide abstract]
    ABSTRACT: Data with high dimensionality often occurs, which will produce large time and energy overheads when directly used in classification tasks. So, as one of the most important fields in machine learning, dimensionality reduction has been paid more and more attention and has achieved a prodigious progress in the theory and algorithm research. Linear Graph embedding (LGE) model is an efficient tool for dimensionality reduction. According to the problems of supervised dimensionality reduction with Non-Gaussian data distributions and at the same time consider neighborhood preserving relations among samples, a novel subspace learning method, neighborhood preserving and marginal discriminant embedding (NP-MDE), is proposed based on LGE and marginal Fisher analysis in this paper. NP-MDE could minimize the within-class scatter and meanwhile maximize the margin among different classes. Moreover, the neighborhood structure with each class is preserved. Experiments on Yale face image data sets show that after dimensionality reduction using NP-MDE, the average classification accuracy is very good.
    Procedia Engineering 01/2012; 29:494–498. DOI:10.1016/j.proeng.2011.12.749
  • Yuan-Dong Lan, Huifang Deng, Tao Chen
    [Show abstract] [Hide abstract]
    ABSTRACT: With an intensive study of the existing density-sensitive distance measures, we proposed a new distance measure for graph-based semi-supervised learning. The proposed measure can not only effectively amplify the distance between data points in different high-density regions, but also reduce the distance among data points in a same high-density region. Then, a graph-based semi-supervised clustering algorithm is presented based on the proposed distance measure. Experimental results on some UCI data sets show that the proposed method has obvious advantages than the old one.
    Machine Learning and Cybernetics (ICMLC), 2011 International Conference on; 08/2011
  • Huifang Deng, Junbin Chen
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, we analyzed RFID middleware technology and its application in logistics customs clearance process. Based on this, we discussed the design tactic of business logic components in RFID midware for logistics customs clearance. Using data filtering, event aggregation and other technology, business logic components can turn the raw RFID data to valuable business information.
    E-Learning, E-Business, Enterprise Information Systems, and E-Government, 2009. EEEE '09. International Conference on; 01/2010
  • [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, a new SOA service bus model (SSB) is proposed. This model, based on SOA architecture, takes advantages of Web service and software bus technology as well as the message passing mechanism in order to enhance the interoperability between heterogeneous systems. SSB integrates the existing resource of IT systems and middlewares, making full use of the data and service resources of these systems. Therefore, a large integrated platform system, which has the features of good scalability, high flexibility, weak coupling, excellent reusability and convenient management, can be constructed. Finally, SSB has been successfully applied to RFID Logistics-Customs Clearance Service Platform(LCCSP).
  • Huifang Deng, Haiyan Kang
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, ONS (Object Naming System) architecture is studied, and the necessity of high performance RFID code resolving service is discussed. Then, ONS distributed network architecture is proposed, along with ONS-based distributed database, synchronous query, storage and calculation using ID as a primary key. High performance RFID coding resolving service system is designed and implemented. The simulation is done for the testing. The experimental results show that our model gained significant performance improvement and achieved millisecond-level time-consuming in resolution.
    Third International Symposium on Intelligent Information Technology and Security Informatics, IITSI 2010, Jinggangshan, China, April 2-4, 2010; 01/2010
  • [Show abstract] [Hide abstract]
    ABSTRACT: Three-Dimensional Electro-Magnetic Particle Model (3DEMPM), based on the equations of Maxwell and Newton-Lorentz, takes advantages of the Finite-Difference Time-Domain (FDTD) and Particle-In-Cell (PIC) to trace a large quantity of particles in order to gain insight into the physics of them. Although MPI alone can be used to parallelize with 1D decomposition along x-direction, the efficiency decreases with increase of CPUs because there are more communications involved. We combine MPI and OPENMP to reduce the communications in the PC cluster, one node of which shares memory with duel-core. The whole domain is decomposed into several sub-domains, the same number as the nodes. Between the nodes we use MPI to realize the communications, and inside each node the OPENMP is applied to do the parallel computing with no communication. In this way, the higher speed-up is achieved while the communication is reduced.
  • Huifang Deng, Meng Xie, Li Zhang, Zhihong Yao
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, an improved LSB (Least Significant Bits) information hiding algorithm is proposed according to human visual theory and implemented in the temporal-spatial domain by using C#, and based on this, a system is developed for hiding and extraction of multimedia files. Compared with the traditional LSB method implemented by Matlab, the improved one has obvious advantages in the environmental configuration requirements, the time-consuming in running and ease of use. To be specific, the system realized on .net platform takes less running time and resources, has better extensibility and greater ease-to-use, and other distinct features. Besides, the system developed for hiding and extracting multimedia files based on the method has a wide range of adaptivity and good applicability
  • Huifang Deng, Fengzhe Chen
    [Show abstract] [Hide abstract]
    ABSTRACT: 3D human motion analysis system is gaining more and more popularity and importance in sports training, game simulation and many other areas. Particle filter algorithm, as a powerful optimized method, can be applied to 3D human motion analysis system with more accurate results delivered and assured. An improved (hybrid) particle filter algorithm (IPFA) is proposed in this paper which integrates the advantages of partitioned particle filter algorithm (PPFA) with annealed particle filter algorithm (APFA). The results show that, the hybrid algorithm (IPFA) gives rise to more accurate results with less computational time consumed, compared to PPFA and APFA, and improves tracking efficiency and accuracy of 3D human motion substantially.
    Information Science and Engineering, 2008. ISISE '08. International Symposium on; 01/2009
  • Huifang Deng, Wen Deng
    [Show abstract] [Hide abstract]
    ABSTRACT: A full examination of the content security requirements of logistics-customs clearance service platform (LCCSP) based on radio frequency identification (RFID) are conducted in this paper. The unified identity authentication (UIA) module is designed and then implemented. Within this module, a new control transfer method based on an improved Kerberos-based authentication approach is proposed, and, therefore, the tickets' sharing problem among different servers at different departments is solved.
    Fifth International Conference on Intelligent Information Hiding and Multimedia Signal Processing (IIH-MSP 2009), Kyoto, Japan, 12-14 September, 2009, Proceedings; 01/2009
  • Huifang Deng, Lizhen Li
    [Show abstract] [Hide abstract]
    ABSTRACT: The widespread applications of RFID technology have brought new challenges to the traditional data management method, as the traditional storage schema has serious shortages in RFID data warehouse environment. More and more enterprises urgently need an effective method to make use of the massive data. In this paper, based on the nature of RFID data, especially the relationship between objects in motion, we proposed a new encoding method which is called object-packing (OP) schema and dual path coding (DPC). OP schema can significantly improve the effectiveness of data compression. DPC method not only removes the constraint that there cannot be a loop in the path, but also does not decrease the encoding efficiency of object tracking. They can be an effective solution to the problems the enterprises encounter in managing and using RFID data.
  • Huifang Deng, Guosheng Lin
    [Show abstract] [Hide abstract]
    ABSTRACT: Radio Frequency Identification (RFID) is playing a more and more important role in our life. How to analyze and discover knowledge from RFID data sets is an urgent and challenging research field. Each tracking object will form a path when it moves through different locations. We present a novel algorithm called PDSC (Path Division and Segments Clustering) to cluster such path data. Considering that there may be some common segments among paths although the full paths are not so similar in general and the common segments may reveal some interesting patterns, we focus on segments clustering in this paper. Firstly we develop an algorithm to divide paths into segments. Secondly a novel similarity definition and algorithm are proposed to measure the similarity of two path segments. Finally we develop a robust clustering algorithm to discover segment clusters. An experimental system is developed to visualize data in every phase. Experimental results demonstrate that PDSC correctly discovers the common path segments.
  • Huifang Deng
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, the author first gave a thoughtful analysis of the necessity of why there is an urgent need for China to quickly train the internationally qualified software talents and then reviewed the current situation of China software talent market and software industry. By categorizing the software professionals on the sector, a definition of high-level software talents with global suitability (or global reach capability, or adaptivity) is described and the characteristic of such talents is discussed. The objective and subjective conditions for such talents to grow out are pointed out. Finally the systematic way of how to effectively train such talents is suggested. Hopefully, the suggestions made in this paper will be of general interest to the policy makers of the government, IT entities and educational (training) institutions.
    Young Computer Scientists, 2008. ICYCS 2008. The 9th International Conference for; 12/2008
  • Huifang Deng, Ruji Jiang
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, a technique based on body features is employed to develop a system for sports training, healthcare and medical diagnosis through reusing huge amount of video resources, and extracting useful information from them. This technique realizes the motion capture by using existing video of human motion. The predefined human model is matched and identified with a sequence of videos. The interested feature points are tracked and calculated using relevant features of the human body. During the tracking, the human model is used for the conditional constraint and corrections. The approach is easy and quick. The efficiency of capture, stability and accuracy of the tracking are improved greatly with a low requirement for hardware specification. The 2D motion data can be generated easily and conveniently for 3D motion reconstruction in this system.