Showing papers in "Computer Engineering and Applications in 2008"

Stochastic stability analysis of fuzzy Hopfield neural networks with time-varying delays

Abstract: The ordinary Takagi Sugeno(T-S) fuzzy models have provided an approach to represent complex nonlinear systems to a set of linear sub-models by using fuzzy sets and fuzzy reasoning.In this paper,stochastic fuzzy Hopfield neural networks with time-varying delays(SFVDHNNs) are studied.The model of SFVDHNN is first established,then,the global exponential stability in the mean square for SFVDHNN is studied by using the Lyapunov-Krasovskii approach.Stability criterion is derived in terms of Linear Matrix Inequalities(LMIs),which can be effectively solved by some standard numerical packages.

Iris recognition based on score level fusion by using SVM

Abstract: In this paper,an iris recognition method using wavelet packet transformation and based on score level fusion by using SVM is proposed.First wavelet packet transformation is used to decomposition normalized iris image for 2 levels,and the feature vectors are selected from the area with the highest energy values and the second highest,then two calculated hamming distance are input to SVM in order to reduce the dimension of SVM and improve veracity.The experiment indicates that proposed approach provides a good result for the iris recognition.

Attribute reduction in concept lattices and attribute characteristics

Abstract: This paper mainly studies attribute reduction in concept lattices based on irreducible elements and constructions of attribute reducts.The judgment method of irreducible elements is first presented,and an approach to attribute reduction in concept lattices is developed based on join irreducible elements.Attribute characteristics are then analyzed by attribute reducts,and necessary and sufficient conditions of different types of attributes are provided respectively.Finally,constructions of attribute reducts are described by attribute characteristics.

P2P traffic classification method based on SVM

Abstract: A method to realize the P2P network traffic classification based on the SVM is proposed.This method uses the network traffic statistical characteristic and SVM method based on the statistical theory to classify the different P2P traffic application.Mainly research focus on four kinds of network traffic classification,which are document sharing BitTorrent,media flows PPLive,network telephone Skype and immediate communication MSN.Introduce P2P traffic classification overall framework based on the SVM,describe how gain the traffic sample and the processing method,introduce the experimental result and construct the traffic classifier.The experimental results confirm the validity of proposed method,the average precise rate is 92.38%.

Parameters selection of support vector regression based on genetic algorithm.

Abstract: This paper mainly deals with the application of genetic algorithm to parameters selection of support vector regression.The variability in predict performance of support vector regression with respect to the free parameters is firstly investigated.A conclusion is drawn that improper parameter would lead to overfitting or underfitting.Then the genetic algorithm is introduced to select parameters for support vector regression.Simulation results proved that the proposed method overcame some obstacles of traditional parameter selection methods,thus the predict precision of support vector regression is enhanced.

Multi-objective particle swarm optimization

Abstract: A new kind of filter for Pareto solutions is presented.And a kind of critertion judging the stagnation of the particles in particle swarm optimization based on the filter is proposed.Based on which,a kind of multi-objective particle swarm optimiza-tion algorithm is proposed.By using the filter for Pareto solutions can improve algorithm's ability to keep diversity.The proposed algorithm is compared with some well known multi-objective evolutionary algorithms through series of standard test functions by means of visual graphs.The results indicate that the algorithm can search the Pareto optimum more effectively.

Generic object recognition.

Abstract: Following the framework of generic object recognition system,many different interest operators and feature descriptors are discussed.Some popular recognition models and learning methods are compared.The evaluation criteria and the available datasets also introduced.Perspectives for future research work are presented in the end.

New method for sentence similarity computing and its application in question answering system

Abstract: Sentence similarity computing plays an important role in machine question-answering systems,machine-translation systems,text categorization systems,etc.Aiming at a sentence similarity model based on key words,an improved method is put forward,including the extraction of keywords ,and the induction of synonyms in sentence similarity definition.And on this basis,a question answer system based on FAQ(Frequently Asked Question) is implemented.This system involves automatically searching for candidate question set,computing sentence similarity and returning the answer to the user.This system can also automatically update and maintain FAQ.Experiments' result shows that the new method has more accuracy than the others in matching questions of question answering system.

Improved SIFT feature applied in image matching

Abstract: The principle of SIFT method is researched,then an improved SIFT points feature descriptor is proposed,aiming to the problem of its high dimension and complexity.Numerous experiments have been conducted for different type of images existing geometrical distortion and radiative distortion as well as noise impaction.The results show that the proposed method is more stable and fast.

Research of fault diagnostic model based on probability Petri nets

Abstract: In order to solve the influence of uncertainty about information in fault diagnosis.The concept about probability Petri nets is defined in this paper.The probability Petri nets can reflect the probability of the transition of a process and the consequence of the logic can be calculated by the incidence matrix.The design of the fault diagnostic model is based on the reduced swatch set to abate the complexity of the Petri nets.Because the complicated Petri nets will induce the combinatorial states expand quickly.And in order to get the logic between condition attributes and the results.The method about the generation of the default diagnostic rules is used to acquire the diagnostic rules.The rules can also extend the range of the diagnosis not only for the swatch set.For the programmable design of the model,the basic format rule is defined and all diagnostic rules can be transformed to a series of basic rule.The design is proved to be availability by the example about rotating machinery fault diagnosis in this paper.

Network traffic modeling and prediction based on RBF neural network

Abstract: With the rapid development,the network now has a large size and high complexity,and consequently the network management is becoming increasingly difficult,so traffic prediction play more and more important role in network management.With a large amount of real traffic data collected from the actual network,a nonlinear network traffic model based on Radial Basis Function(RBF) neural network theory was constructed to predict the network traffic.The structure design and leaning algorithm of RBF neural network is presented.The simulation results on real network traffic show that the proposed RBF_based prediction scheme is efficient,and has better precision and adaptability compared with the traditional linear model and BP neural network.

Research of Apriori algorithm in mining association rules

Abstract: A tree structure algorithm based on vertical tid-list to resolve the frequent itemsets program in Apriori is put forward.The algorithm combines the ordinal character of itemsets and makes the nodes ascending order by support count in the tree,so the sets of the candidate frequent itemsets is the least totally.The algorithm can greatly reduce the candidate frequent itemsets,keep the completion of frequent itemsets,reduce the cost of computing,and improve the efficiency of algorithm.

Flight delay propagation research based on Bayesian net

Abstract: In the industry of civil aviation,delay propagation of flight always is a main factor that impacts the flight’s delay.This paper discusses the delay propagation problem in busy hub-airport.All the data used in the experiments is come from real records of one large domestic airlines.Through producing the model of delay propagation and modeling the Bayesian Net,the impact of arrival-delay on departure-delay between related flights is discussed.The outcomes of learned Bayesian net show that the existence of propagation phenomenon from arrival-delay impacts onto departure-delay;the degrees under varied circumstance are different;and the according coping methods are different.Cancellation of flights is a drastic remedy of those coping methods.It can weaken the delay propagation to some extent.The weakened degree is related to the number of flights cancelled.A preliminary reference can be made based on this research for airports when there is large scale delays of flights happened.

Improved image edge detection algorithm based on Canny operator

Abstract: In order to solve the problem of missing edge and feigned edge using Gauss filter with the canny operator,an improved edge detection algorithm is proposed.It replaces Gauss filter with the wavelet threshold filter whose threshold obtained by the GCV theory to wipe off the image noise effectively,computes the magnitude and the direction,and detects and connects edges of the image in use of non-maxima suppression and hysteresis threshold.The experiment shows that the improved algorithm can increase veracity effectively and obtain better edge detection result.

Lightweight J2EE architecture based on spring framework and its application

Abstract: Choosing a reasonable J2EE architecture is very important for application development.The techniques Spring framework provides,such as dependency injection features component assembly,transaction and log management,etc.can sustains lightweight J2EE architecture.This paper introduces the Spring framework and compares lightweight J2EE architecture based on Spring framework with other J2EE architectures and gives a case of its application finally.

Research and application of transaction complement mining algorithm

Abstract: An algorithm of transaction complement mining is presented,which uses transaction complement search strategy to generate candidate-items,the number of which is reduced by pruning subsets of frequent item sets and superset of non-frequent item sets,and further uses binary logic operation and transaction character to count support,its efficiency is improved.The algorithm is used to extract transverse spatial association rules,and its experiment indicates that the efficiency is fast and efficient.

Improvement and application to weighting terms based on text classification

Abstract: Text representation has been the fundamental problem in Information Retrieval.tf.idf(term frequency,inverse document frequency) as one of term weighting schemes in Vector Space Model is a good text representation,Which is popular and make good results in the field of Information Retrieval.The difference of the proportion of distribution of terms in text collection is one of the most important factors of expressing the content of text.But the calculation of IDF,don’t consider the information of distribution about terms among classes,and don’t consider the more term weighting for the terms of the relative distributed balance inner classes.The improved TFIDF are used to select feature,KNN algorithm and genetic algorithm are used to train the classifier.and proves that the improved TFIDF method is feasible.

Solving nonlinear systems of equations based on Social Cognitive Optimization

Abstract: The solving problem for nonlinear systems of equations is transformed to the function optimization problem,and then,Social Cognitive Optimization,a new Intelligent Optimization Algorithm,is introduced to achieve the optimal result for the problem.The experiment results manifest that Social Cognitive Optimization can be feasible and effective on solving the nonlinear systems of equations problems.

New method of curve fitting based on cloud transform

Abstract: Cloud transform is a method to discretize continuous data based on cloud modelThis paper proposes a new method of curve fitting based on cloud transformSome experiments validate the proposed method

0-1 integer programming problem based on RecA-mediated triple-stranded DNA structure

Abstract: Since triple-stranded DNA can be formed readily at any location in double-standed DNA with a homologous deoxyoligonucleotide in the presence of RecA-protein,it can be explored the possibility and feasibility to compute by this unique structure.In this paper the author shows that triple-stranded DNA structure mediated by RecA protein can be used for solving 0-1 integer programming problem.It can be solved the integer programming problem and satisfiability problem by this model.It can be transformed from integer programming problem and satisfiability problem to 0-1 integer programming problem.

Fingerprint separation based on morphological component analysis

Abstract: Based on the characteristic of the fingerprint,after making some changes on the morphological component analysis and combining the MCA with the harris corner detector,it puts forward a fingerprint separation algorithmThis algorithm is based on the basis pursuit denosing algorithmFirstly,it uses two same dictionaries to represent sparsely the overlapping fingerprint image or the mixture of the finger and the texture background,then shrinks the sparse coefficients with the soft threshAfter this,minimizes the harris-like operator of the two separating texture images respectively using gradient descending algorithm,to make the coners of the two images minimizedLastly,it implements total variation regulation on one of the two separating images to achieve separationExperimentation results indicate that this algorithm can realize fingerprint separation

New research on multi-classification based on Support Vector Machines.

Abstract: Support Vector Machines(SVMs) is originally designed for binary classification.How to extend it for multi-category classification is one of hot research issues.This paper emphasizes on analyzing decomposing and reconstitution methodology of SVM multi-classification algorithms,and discusses two key factors which affect performances of categories in detail:decomposing strategy and composing strategy.The paper does experiments to validate authors' opinions and then comparesall kinds of SVM multi-category classification,including M-ary SVMs and fuzzy SVMs.

Access control model for enhancing survivability

Abstract: Traditional access control models have disadvantages when they are applied in a survivability system.An access control model for enhancing survivability is proposed,which named TTC.The model can protect critical services and data,response to intrusion alarm in real time and help recovery process even when the system is compromised.It contains three rules of trigger,track and control.The trigger and track rules trace activities of intruder in system by using attack tree.The control rule forbids the traced subjects to tamper critical services and data.The value of TTC for enhancing survivability is attested by formal proving, application demonstration and comparing with traditional access control models.

Novel approach for vehicle license plate tilt correction based on Radon transform

Abstract: Tilt vehicle license plate have a bad effect on its character segmentation and recognition.In this paper,a novel approach for number plate tilt correction based on Radon transform is presented.Radon transform in the range of -20°,20° is implemented on the number plate images.Then the accumulative total of absolute value of difference of the results is calculated,respectively.The angle of the Radon transform corresponding to the maximal accumulative total is confirmed as a skew angle of the number plate.Bilinear interpolation rotation correction is performed to the plate that is tilt to the horizontal line,and bilinear interpolation offset correction is made to the plate that is tilt to the vertical line.Experimental results show that the method can be implemented easily and offers robustness when dealing with dirty license plates and license plates in variant lighting conditions.

New extension of cubic Bézier curve and its applications

Abstract: In order to construct Bezier curves with shape control parameters,two classes of polynomial basis functions of 3th and 4th degree are presented in this paper.They are both extensions of cubic Bernstein basis functions.Properties of the proposed basis functions are analyzed and the corresponding polynomial curves with shape control parameters α,γ and α,β,γ are constructed accordingly.The constructed curves not only inherit the outstanding properties of the cubic Bezier curve,but also are adjustable in shape and fit close to the control polygon.In addition,the continuity condition and some applications in curve and surface design about the new curves are discussed,and two definition of Extension-surface are also presented.Some examples illustrate the new curves are very valuable for the design of curves and surfaces.

Color image segmentation algorithm based on sub-block region growing

Abstract: A color image segmentation algorithm based on sub-block region growing techniques is presented.The original image is first partitioned into sub-blocks which are not overlapped.Applying color and texture feature of one sub-block,a color clustering algorithm based on perceptual uniformity within the sub-block is performed to classify these sub-blocks,and a block growing algorithm is employed to achieve segmentation for the whole image based on growing rules.The experimental segmentation results hold favorable consistency in terms of human perception,and confirm effectiveness of the algorithm.

Design of multi-routing protocol based on energy of wireless sensor network

Abstract: Based on the traditional network routing protocol,DSR,and the idea of routing expected life in ESDSR,we have designed a Multi-Routing protocol of energy which works well in the wireless sensor network.In the design of this protocol,we have dealt with the three kinds of routing situation in the network structure.In the course of broadcasting the data,the route can be changed with this protocol.By simulation,the network’s life is proved to be prolonged effectively.

Algorithm of sound transformation from relational schema to ontology

Abstract: Ontology technology can achieve the capacity of interoperation and interchange at semantic level,so ontology learning from relational data model is becoming one of the most important issues in the construction of Semantic WebWe present an algorithm which stems from a method of translating a database model into a conceptual model through database inverse engineeringThis algorithm for transformation from relational schema to ontology expressed by OWL DL,we prove to be soundFurthermore,a prototype system to verify the result is developed

Channel assignment algorithm based on immune strategy

Abstract: Based on restriction and characteristic of channel assignment,an immune strategy algorithm(ISA) is presented to solve the dynamic channel assignment problem in cellular telecommunication systemsThe algorithm can assure greatest compact assignmentAnd focused on the degradation of genetic algorithm,immunization strategy algorithm is introduced for channel assignment,which can resolve degradation preferablyThe experiment results demonstrate that the callblocking probability of the immune strategy algorithm for channel assignment(ISACA) is lower than all of the fixed channel assignment(FCA),dynamic channel assignment(DCA) and improved genetic algorithm dynamic channel assignment(IGADCA)

Dual channel power controlled backoff based on geographic location in Ad Hoc networks

Abstract: In this paper,an energy optimization with dual channel power controlled backoff based on geographic location is proposed.The basic idea is control channel backoff from low power to high power,data channel "backoff" from high to low.And using the benefit of geographic information,the problem of asymmetric links is solved better.In addition backoff algorithm is introduced and the successful transmission probability is analyzed briefly in this paper.The simulations result shows it can meet the design of Ad Hoc networks of short distance priority.