Active Media Technology

Preface

Conference Organization

Contents

The paper introduces Web Intelligence (WI) as a new paradigm for developing business intelligence on the Web. WI research is discussed from two perspectives: (1) an intelligent Web-based business-centric schematic diagram, and (2) the conceptual levels of WI, for creating the Wisdom Web that serves its users wisely. It builds upon the fast development as well as application of various hybrid of AI and WWW technologies.

Effective human teams use overlapping shared mental models for anticipating information needs of teammates and for offering relevant information proactively. The long-term goal of our research is to empower agents with such “shared mental models” so that they can be used to better simulate, train, or support human teams for their information fusion, interpretation, and decisions. Toward this goal, we have developed a team agent architecture called CAST that enables agents to infer information needs of teammates, which further enables agents to assist teammates by proactively delivering needed information to them. In this paper, we focus on two key issues related to proactive information delivery behavior. First, we model the semantics of proactive information delivery as an attempt (called ProAssert), which extends the performative Assert in Joint Intention Theory. Second, we introduce a decision-theoretic approach for reasoning about whether to act on a potential proactive assert. Experimental results suggested that the decision-theoretic communication strategy enhances the team performance. The formal semantics and the decision-theoretic communication strategies together provide a sound and practical framework that enables further studies regarding proactive information delivery for supporting the decision making of a team involving human and agents.

Relevance of fuzzy logic, artificial neural networks, genetic algorithms and rough sets to pattern recognition and image processing problems is described through examples. Different integrations of these soft computing tools are illustrated. Evolutionary rough fuzzy network which is based on modular principle is explained, as an example of integrating all the four tools for efficient classification and rule generation, with its various characterstics. An example of rough-fuzzy case generation is also provided. Significance of soft computing approach in data mining and knowledge discovery is finally discussed along with the scope of future research.

The research theme of the Cyber Assist Research Center is the development of human-centered information processing technologies which can provide situated information that I-want-here-now through a “natural interface”. We are strengthening a variety of technologies that link the digital realm, represented by the Internet, to people who live in the real world. In this paper, we mainly focus on location-based communication by radio. This system uses physical location as its means of first contact. Subsequent communication and services given over it may or may not rely on location. We are providing a set of tool kits named UBKit to construct many kinds of user interfaces.

Almost all work in automatic facial expression analysis has focused on recognition of prototypic expressions rather than dynamic changes in appearance over time. To investigate the relative contribution of dynamic features to expression recognition, we used automatic feature tracking to measure the relation between amplitude and duration of smile onsets in spontaneous and deliberate smiles of 81 young adults of Euro- and African-American background. Spontaneous smiles were of smaller amplitude and had a larger and more consistent relation between amplitude and duration than deliberate smiles. A linear discriminant classifier using timing and amplitude measures of smile onsets achieved a 93% recognition rate. Using timing measures alone, recognition rate declined only marginally to 89%. These findings suggest that by extracting and representing dynamic as well as morphological features, automatic facial expression analysis can begin to discriminate among the message values of morphologically similar expressions.

Random agent networks have been used as important contrast roles in the study of topologies of real-world agent networks. As an important property, in this paper, we study the clustering coefficient of random agent networks. Specifically, we give the theoretical average value and lower bound of the clustering coefficient in random agent networks with n agents and m links. As a validation, we also provide some experimental results to support our theoretical results.

The foreign exchange (FX) market has many features including (1) Each trader’s payoff depends not only on his own behavior, but also on other traders’ decisions; (2) The number of traders is too large to make them all know the other dealers’ methods of decision making; and (3) The FX market has many levels. The FX market is complex because of these features. A diversity of techniques are required to deal with such complex problems. That is hybrid solutions are crucial for the FX market. On the other hand, research into the FX market has revealed that it demonstrates some characteristics of multi-agent systems such as autonomy, interaction, and emergence. To this end, an agent-based hybrid intelligent system was developed for FX trading, which is based on our proposed agent-based hybrid framework. This paper is to discuss the analysis, design, and implementation such a system. Some experimental results and comparisons with related works are also provided. The interest of this paper does not reside in improving the predictive capabilities of different FX models, but rather in how to integrate different models into one system under the unifying agent framework. The success of this system indicates that agent perspectives are very appropriate to model complex problems such as the FX trading.

Traditionally, accessing databases on the Internet is based on client/server paradigm (including web-based paradigm). However, there are some drawbacks when using the techniques. Mobile agent technology provides an excellent solution that can easily eliminate those disadvantages. By conducting a series of experiments and analyzing the experimental results, this paper presents in which circumstance mobile agents can outperform traditional client/server paradigm, and how they can overcome the shortcomings in client/server paradigm in database access on the Internet.

Human-agent communication is important for many multi-agent systems. A natural language interface is desirable for such communication. In this paper, we discuss the techniques we developed to create a natural language interface between human operators and agents. We mainly address the translation from the natural language into the FIPA ACL (Agent Communication Language). The techniques are λ reduction for translating into the SL (semantic language) and inverse parsing that converts a parse tree into an ACL message.

The very challenging issue for intelligent agents is “do what they should do”. The BDI architecture is presented to solve this challenge. The main difficult for this architecture is the formalizing problem. In this paper, we discuss the procedure descriptive framework, which presents a method for formalizing BDI architecture. We a present decision model for intelligent agents in this procedure descriptive framework. This research shows how an agent to generate beliefs and make decisions by using its current beliefs.

In this paper, we contribute a framework based on multi-agent technique to manage large amounts of well-logging parameter graph information and store the data into Oracle database. We employed the idea of Gaia to design the framework, which consists of three levels, namely, role model level, agent type level and agent instance level. Four agent types, which correspond to the four roles, are proposed. We design the architectural model of agent instances with Open Agent Architecture. The experimental results show that all agents in the system can work cooperatively.

Agents in a Mobile Agent (MA) system must interact with each other using some kind of communication models to exchange data and coordinate their execution. Several communication models have been widely used in distributed systems as well as most MA systems. Typical models are Message Passing, Remote Procedure Call (RPC/RMI), and Distributed Event Handling. Communication between agents takes place by means of an Agent Communication Language (ACL). The essence of an ACL is to make agents understanding the purpose and meaning of their messages. In this paper, we will investigate these concepts and propose a new Agent-based communication model which offers a high level message passing facility and a logic-based ACL for mobile agent applications.

A distributed IDS architecture based on the machine learning, intrusion analysis mobile agents and data gathering static agents is proposed. This distributed IDS architecture emphasizes on encapsulating the machine learning algorithms related to the network security and the learned rule analysis models into the multiple mobile agents. The motivation is to cooperate to detect distributed and coordinated attacks in the network.

A metasearch engine model based on Agent is presented and introduced to improve the speed and accuracy in searching Internet information according to one’s own demand in this paper. The model adopts the structure of metasearch engine, regards Agent as the basic components and utilizes autonomy and corporation of the agents to finish personal Internet search of user. In the system design, proposed the component search engine selection strategy based on user’s favor. All these strategies in this paper can simplify the process of operation, increases the coverage of web, and promote the performance of the system.

Migration strategy is crucial to mobile agents’ migration performance. Among the various aspects of a strategy, the choice of destination is a key factor. In this paper, we present a hierarchical model for agents to locate services and to determine migration destination. The merits of this model are illustrated by the comparison between our approach and two existing paradigms.

Grid Computing is rapidly changing the computing infrastructure and has led to the possibility of using wide-area distributed computers for solving large-scale problems. However, the resources in the Grid are geographically distributed and owned by multiple organizations with different usage, cost policies, varying loads and availability. The sharing and use of resources in such a large and distributed environment is a complex task. In this paper, a multi-agent system for Grid Computing is presented to coordinate resource sharing and solve problems in dynamic, multi-institutional virtual organizations. The multi-agent system architecture and autonomous agents negotiation mechanism are discussed in detail. This multi-agent system provides decentralized resource management and sharing capability and is adaptable to changes in the environment and user requirements. It is a scalable, controllable, load balancing, and easily understandable system for sharing and use of wide-area distributed resources.

A typical distributed application system, which applies the Reinforcement Learning multi-agent model, is introduced, the mutual interaction between agents and dynamic environment is briefly mentioned, and the schema of fine granularity and coarse granularity are analyzed more deeply.

This paper proposes a new definition for role, and estimates role’s functionality in MAS (Multi-Agent System) analysis, design and implementation. A set of graphical notations and a method RoMAS (Role-based Modeling for MAS) are presented. A case about RoboCup simulation is studied to illustrate the applying of the modeling language.

This paper introduces the radar transmitter power measurement based on virtual instrument that is developed by us. The hardware frame of intelligent radar transmitter power measuring system is given. The structure and principle of virtual instrument card are also given. The guiding software that automatically measures radar transmitter power is described after measuring method of radar transmitter power with this system is discussed.

A new way for the comprehension of natural language is proposed in this thesis. It is the L-Agent (language-agent). First the problem for a pronoun in a dialogue is analyzed. The L-Agent is the interactive interface of human and computer, especially for the Chinese comprehension. There are three mechanisms in it. One is the Basic Meaning of Natural Language (BMNL). One is the language’s forming basing on the comprehension model of BMNL. The other is the behavior and activity’s forming basing on the cooperative semanticizing comprehension model of the BMNL. Its’ theory basing on the new language’s theory of BMNL for the computer’s natural language’s comprehension.

Military C2 (Command and Control) systems can be viewed as Complex Adaptive System (CAS), inspired by ideas from complexity theory, we introduce the multi-agent technique to cope with complexity and improve the “intelligence” and adaptiveness of C2 systems. In this paper we describe the architecture and the development process (analysis, design and implementation) of multi-agent C2 systems (MAC2S). We also present a testbed based on High Level Architecture (HLA) simulation, which supports the testing and evaluation of MAC2S.

This paper proposes quality of service (QoS) metrics for multimedia network applications based on human factors and technology attributes. Human factors describes human perception to different kinds of media, such as audio and video. Technology attributes represents the different technological attributes of multimedia applications, such as time-dependence and symmetry. Both of these key factors lead to varying requirements for quality of service. With these metrics, we can provide criteria for resource allocation for differentiated multimedia traffic in the realization of QoS on computer networks.

The investigation of the human interactive characteristic between visual and auditory systems is very important for designing active computer systems and brain computer interfaces. Up to now, human visual and auditory mechanisms have been investigated as two separate systems. In this study, we pay attention to human interactive characteristic of the visual and the auditory calculation processing, and the electroencephalograms (EEG) are measured. Based on the measurement results, the authors discuss the difference between the visual and the auditory information processing, and design a brain computer interface and an active computer system for handicaps.

Coupled with the development of the science and technology is the rapid growing demand for the parallel computation that is just the target of MPI (Message-Passing Interface). In this paper, we begin with the basic concepts of MPI, and focus on how to implement MPI in Java. This new implementation of MPI is an application interface targeted for an MPP (Massively Parallel Processor) and a cluster of PCs or workstations, and is able to provide high performance at a low cost.

This paper presents an agent-based application model that can support flexible, dynamic deployment in Bluetooth piconet. Applications can be dispatched over wireless media. By mobile agent, portable devices can perform complex task without concerning about own limited capacity because applications can be effectively and flexibly summoned on demand. Finally, we outline the architecture of our model and its component.

In this paper we provide two extensions to the FIPA standard. One extension provides a generic behavior model for agent types. Another extension is the provision of a security model for agent applications. We illustrate the correspondence between the abstract behavior models and the FIPA architectural standard.

Tutoring system is very important to Tele-Education. The current tutoring systems only matches the keywords of the questions in the question-database according to users’ inputs, so they can’t provide a friendly interface to the users. To solve this problem, this paper presents an agent-based intelligent tutoring system by applying the semantic-net principle. The necessity of establishing an intelligent tutoring system is discussed and the tutoring model and its implementation based on the constrained fields are also given.

Although the Semantic Web describes a wonderful perspective for intelligent and precise information searching and gathering, it is still a challenging issue of how to implement information query on the Semantic Web. For the purpose, we propose an automatic information gathering and querying architecture on the Semantic Web, and give a kind of prototype implementation of this architecture by existing technologies. In particular, we introduce our Q language to bridge the gap between end-users and query agent system.

Security is a major problem of mobile agents. In this paper, we firstly discuss both the security threats and the protection issues. After that we introduce a new mechanism for protecting mobile agents based on detail analysis of data categories of mobile agents. This active-prevention mechanism following software-protecting way is much more secure than many other mechanisms, as it can not only prevent the Tamper-attack, but also prevents the Read-attack.

Mobile agents show promise as a new distributed programming paradigm. Generally speaking, a mobile agent is an executing program, an independent unit of computation that can migrate, at times of its own choosing, from machine to machine, in a heterogeneous network, to accomplish its user’s given work. However, because of the complexity and changeful of network, how to implement a mobile agent system so that agents can move in network efficiently, becomes the essence of mobile agent technology. In this paper, we propose a navigation model of mobile agent, and then analyze this model.

The metadata standard of data in Geography Information System (GIS) has been created, and plays an important role in data’s organization and management, but for the Model and Method Metadata, which is very important to the usage of these method, has not been described in details, thus it is very difficult to query and index the existing models and mathematics methods in existence. With the research to the GIS, we put forward a set of metadata database architecture that is suit to the GIS and other information systems which have relationship with geographic data, so that it can be used to realize the efficient management of the models and methods. In this paper, we also have some discuss on the Agent-based architecture of the Model and Method Metadata services, put forward the prototype of the metadata Query Agent which will quickly query in the extraordinary big metadata database. Finally it draws a conclusion that it is feasible to realize the Query Agent based on the Model and Method Metadata, and it will be very efficient to find the models and methods with Query Agent in GIS.

Mobile Agent technology is a beneficial technology for conceptualizing, designing, and implementing distributed web software systems. Digital City Spatial Information Service System is a typical distributed system with huge volume of spatial data to acquire and analyze. In this article, we firstly introduce the Mobile Agent technology and the importance to implement Digital City Spatial Information Service System. Then the necessities for Mobile Agent technology to implement Digital City Spatial Information Service were described from two aspects, i.e. the advantage of Mobile Agent technology and the need of Digital City Spatial Information Service System. The architecture of Mobile Agent based Digital City Spatial Information Service we put forward was introduced in the next part. Lastly we expatiated the realization of Mobile Agent based Digital City Spatial Information Service System using Java and DCOM technology. A conclusion and future works on the Mobile Agent based Digital City Spatial Information Service is also discussed.

Since agent has an action model similar to human being, it finds its right place in the application of Geo-DVE. In this paper, we firstly analyse and point out the four-level features of multi-user interaction in a Geo-DVE; then illustrate the feasibility and methods of joining MAS with Geo-DVE; further more, outline the architecture of an agent based multi-user interaction Geo-DVE system, and discuss the message transport service in detail where an XML object is highly appreciated to model the message content; finally, introduce our prototype and reach the conclusion.

According to [4], a real-time task needs to be executed in parallel on different nodes in distributed systems, and it gives a fault-tolerant scheduling algorithm and the result of simulation. In this paper, we discuss a built-in security scheduling (BISS) and show an example of our application. This paper, first proposes a hierarchy model of BISS for a distributed real-time system, and introduces the criteria of each layer. Then, gives the example for BISS in our practical system. At last, we discuss the efficiency and the future work of BISS.

Digital Video Recorder (DVR), which is a key part of digital monitoring system, goes on real-time storage and treatment to the digital video information material mainly and has highly intelligent, synthetic. This text introduces some basic information of DVR and tries to analyse the extant question mainly. Put focal point on research about design principle of embedded DVR and the current situation, exposition of development in future.

This paper designs an IP intelligent agent development platform based on μC-5471, and analyzes the characteristic of its hardware dual-core (ARM+DSP) architecture for which various peripherals can be configured to realize intelligent product with specific purpose. DSP can be used to realize real-time speech code or complex control algorithm in its high operation speed, and ARM can manage keyboard, monitor and Ethernet peripherals etc. Real time multitask operation system μC/OS-II embedded in CPU can build up an intelligent agent development platform with powerful capability. The digital speech classroom terminal on this platform has been developed. Embedded software design, multimedia, microprocessor and network technology are used in the IP intelligent agent. It also can act as Internet network terminal interface.

Traditionally, when mobile Agents move from one node to the other across a network, they must move as a whole. If a mobile Agent performs more than one function, the codes which are not necessarily run on the destination node must move with it, too. These can affect the efficiency of a network. So we introduce the Code-Forwarding Agent model, and discusse its implementation and application from its definition and kernel function.

With the rapid development and popularity of the Internet and multimedia, computers have become an important communication medium for long-distance people besides telex, fax and telephone. How to process Emotion information and how to realize the natural and harmonious human-centered human-machine interface are great challenges faced by artificial intelligence today. The research of emotion information retrieval is an intersectional research. Based on the research products and methods of psychology, painting, a content-based emotional image retrieval model has been proposed in this paper. First, based on the idea of “dimension” from psychology, a common emotion space is constructed and the measurement of similarity in the space is analysis. Then, dominant colors, moment invariants, color and gray compositions, which can stimulate human emotion very easily, have been extracted from images to construct the feature space. After that, emotion annotation idea has been provided, support vector machines are used to map images from the low level feature space to the high level emotion space, and automatically annotate unevaluated images based on user’s common emotion. Furthermore, an emotional image retrieval algorithm, which including common emotion retrieval and individual emotion retrieval are proposed. The common emotional image retrieval has been indexed in the “common” emotion space, while the visualized interactive genetic algorithm has been used to realize the individual emotional image retrieval. Based on image content, an emotional image retrieval system has been realized. The experimental results demonstrate the effectiveness of our approach.

In this paper, a digital image dual watermarking scheme is presented. In the scheme, robust watermark is integrated with fragile watermark. Robust watermark is the copyright symbol of ditigal, image and fragile watermark is a validating authenticity measurement to confirm the tamper status of digital image. Both of two watermarks are embedded in spatial domain. Double keys, embedding and correlative recovering ways of non-plane and some characters based on HVS property are introduced during embedding robust watermark, while fragile watermark, adopting spatial domain low bit plane random embedding way aimed at pixels. The scheme is an effective measurement to determine the authenticity and integrality of an image.

It is often necessary to construct discrete-time Hopfield neural networks in arbitrary waveform synthesizer design. According to the need, an effective method is presented in this paper. Under constraints of fixed points, limit cycles of iteration sequences, the method simultaneously acquires both the topology and weights for Hopfield neural networks by solving inequalities. Experimental results demonstrated clearly the effectiveness of the method.

The coordinate axes of the feature space of linear discriminant analysis (LDA) are not orthonormal, and LDA’s performance is often affected by the variance of train set. In this paper, We put forward an orthonormal version, named orthonormal component discriminant analysis (OCDA), which does not increase computational time obviously. The experiment on Yale and ORL database shows that OCDA is reliable and has a superior recognition rate.

In this paper, a new variable step-size CMA blind equalization algorithm is introduced. The time-varying step-size sequence is adjusted using the kurtosis of the estimation error. The convergence properties of the algorithm are analyzed, and the selection principle of parameters is discussed. Simulation results confirm the algorithm’s improved performance.

In many practical situations, e.g., in manufacturing, it is important to test the structural integrity of material structures, in search for defects. Integrity testing is often very expensive. In mechanical testing, samples are taken during the manufacturing process to be destroyed in order to determine the properties of the part that has just been manufactured. On the other hand, Non Destructive Testing method (NDT), is a field of engineering that tests objects without destroying them in the process.

In this implementation, ultrasound is used as NDT. It works on the receiver transmitter principle, that is, some energy sent from the transmitter passes through the test object and ends in the receiver. A multiprocessor system performs the Fast Fourier Transforms (FFT) on data sets produced by sensors, from the energy that is measured by the receiver, to determine the properties of the test object.

Digital watermarks can be used to control and trace the original source of unauthorized copies. But by present protocol, because of the seller’s mastery of the watermarked multimedia, the buyer can claim that it is the seller who resells the copy of it. To solve this problem, Nasir Memon proposed an interactive buyer-seller watermarking protocol. But a third party is introduced in that protocol, which masters the embedded watermark, so he must be fully trusted. In this paper, combined with the advantage of “protocols for watermark verification” proposed by K.Gopalakrishnan, we present a watermark protocol based on the semi-trusted third party, in which the third party only takes part in the dispute resolution and doesn’t know the content of the watermark.

Recently the performance of parallel simulation has attracted a great attention in computer science. Most of the works have devoted to minimize the time consumption in simulation processing, whereas this paper stresses the advantages of the parallel discrete-event simulation of VHDL based on COW. In the paper it describes the advantages of COW and MPI and focus on the key points in the PDES and what the improvement we do on Time Warp mechanism to decrease the time in communication processing.

It’s hot and highlighted now to applying Micro-Electro-Mechanical Systems (MEMS) technique to manufacture broadband microstrip antennas. The Finite-Difference Time-Domain (FDTD) method presented in this paper is efficient and accuracy for modeling and analyzing the MEMS multiplayer microstrip patch antenna.

This paper proposes a novel high-order Associative Memory System based on the Newton’s Forward Interpolation, which is capable of implementing error-free approximations to multi-variable polynomial functions of arbitrary order. The advantages it offers over conventional CMAC-type AMS are: high-precision of learning, much smaller memory requirement without the data-collision problem and also the advantages of much less computational effort for training and faster convergence rates than that attainable with multi-layer BP neural networks.

Multilayer neural network based on BP algorithm is being applied in many fields, but in spite of BP algorithmic shortcomings of low convergence rate and proneness to yield to local minimum. Many analyses have revealed that the adjusted weights of various layers differed much, so a new algorithm of weight balance layer by layer was brought forward to calculate the weights’ influences on neural network. The results indicated the new algorithm converged faster than the standard BP algorithm. And a control system simulation using the new algorithm in Synthetic Integration for Meteorological Forecast is to be a good approach.

An efficient method for postprocessing of the directional field is presented. Utilizing the coherence of the squared gradients, our algorithm firstly clears off the background/noise block in the directional field and then smoothes the directional field. Through this method, we can get accurate directional field. The efficiency of this method is illustrated by experiment.

This paper proposes a wavelet fuzzy clustering algorithm that textured and non-textured images can be automatically classified and retrieved. At first, wavelet transform is performed on the images. Secondly, fuzzy clustering is performed in the lowest frequency wavelet subband of the images according to the color and texture features. Thirdly, the images are judged as textured or non-textured images by using the region scatterness. Finally, the features can be extracted and the similarity of images can be obtained according to the similarity criterion. The algorithm is confirmed by the test that can get good classify and retrieval performance.

Character recognition is an important research area. The recognition of printed Mongolian characters is a brand new research orientation. We adopt syntactic structural recognition as the main method and implement the recognition of multi-font printed Mongolian, In order to enhance the efficiency of human-machine interaction interface, The whole system is based on multi-agent architecture. Practice demonstrates that the whole correctness rate is up to 98.2%, human-machine interaction environment is also greatly improved.

In this paper we utilize the universal technology frame (mainly the distributed computation technology) to build the distributed simulation application. We bring forward the concept to utilize the universal technology to build web based IDE for distributed simulation system. Additionally, we bring forward a simple technology frame about the standard of the coming distributed simulation system.

Viseme is an audio-visual model for speech-driven talking head animation. In this paper, a viseme HMM based speech recogntion system is built to drive a talking head. Triseme is used to take mouth shape contextual information into account to achieve accurate models. As models mushroomed, to get robust models using the limited training data, decision tree based state tying is adopted in the triseme modeling. Similarity of mouth shapes (SMS) is brought forward to design visual question set in the tree building process. Experimental results show that SMS is a good measurement of mouth shape contexts. Decision tree is a feasible way to get robust model parameter estimations.

The parameters of coal excavator have great non-linearity (dead area, space and soon), time-variation and pure-delay time in the process of cutting along with the differences of traction speed, depth of cutting, and characteristics of coal and rock. In addition, the cutting curves cannot be any kind of form, because of restriction of hydraulic brackets. So the control system of stored memory cutting is developed based on the above- mentioned and the practical work model. The previewing control principal will be applied to the process of automatic height adjustment for the roller of coal excavator. Memory track will be fully utilized as previewing information. The complex control problem of will be solved radically. The control effect of simulation is better. Also performance index will arrive into optimization at the same time.

This paper introduce a fruit classify system based on pattern recognition. Three key technology are discussed: 1. double vidicons and multi fruit’s pictures in one image are adopted, thus the effective pictures of one fruit are increased, this can improve the measure precision; 2. A new arithmetic based on pixels statistic is brought forward for the purpose of quick measure the fruit’s diameter; 3. Using some intelligent rule to eliminate fruit’s spot and the glisten influence.

A new adaptive watermarking embedding and detection scheme of digital image is presented. The local embedding strength of the watermarking is determined by the statistic feature of the original coefficients of the image which allows the strength of the embed signal is vary with the local Dates, The robustness against dates analysis of the watermarking is improved. The estimation of the correlation peak value is used in the detection process. Finally, the simulate results is followed.

A novel scheme of fragile watermarking construction, embedding, detection and authentication is presented, it inlays the watermarking information for copyright authentication in the original image by quantizing the corresponding DWT coefficients. The watermarking is sensitive to the tamper occurred in localized space and frequency regions, and the credible evidence for the image authentication is provided by assessing the degree of the changed watermarking. On the other hand, a Watermarking Authentication System constructed by Zero Knowledge Proof Protocol (ZKPP) based on the discrete logarithm is used for copyright verification of digital products. In this scheme, the owner could disclose the correlation information to show his/her Id, but the secret copyright information is not disclosed during the authentication process, and this allows the authentication of the digital copyright is more safe and credible.

Many researchers focus on classifying E-Mail. This paper discusses how to combine rules-based classifier with VSM-based ones to filtering mails. In this paper, a prototype system—EMCMS integrated with Microsoft Outlook is designed based on a notion that a model classifying E-Mail should embed into a commercial E-Mail client. In addition, a new classifying committee is provided. Experiment results shows that the design is effective and promising.

This paper designs an embedded web media development platform based on low cost TMS320DSC21. The character of its dual-core (ARM+DSP) architecture is analyzes. It is described how various peripherals are configured to realize web visiting. DSP can be used to real-time run audio and video encoding algorithm. Embedded operating system μCLinux has been ported to ARM to manage keyboard, monitor, Ethernet card and H.323 application etc. The embedded web media on this platform can be easily developed. Embedded software design, multimedia, microprocessor and network technology are used in the adaptive web media.

This paper expounds the semantic conflicts existing in distributed information systems. In order to iron out the conflicts between data coming from different data sources and facilitate the interoperability among distributed, heterogeneous data sources, we put forward a novel method: using ontology to describe concepts in different sites and structuring the information. It allows for a more disciplined design of database, and it facilitates sharing and reuse. Then we also put forth an interoperable and sharing model for remote education. Though this model is put forward for remote education, it also can be applied to other application domains, such as Web-based databases, digital library, data warehouses, data mining, etc.

With the rapid development and spread of the Internet technology, there have been significant progresses in the Internet information retrieval and active services. However, the traditional Internet search engines and active services are all based on simple similarity of key word or bag of words, and documents features and users’ personal interests aren’t well depicted due to the lack of semantics. XML is a newly emerged standard for the www data exchange and description. XML-based information retrieval system has the rapid development because it not only describes the contents of the documents, but also takes the documents’ structural information. In this paper, a newer personal active information service system is discussed, under which the heterogeneous, distributed data spreading on the WEB will be integrated to XML data, and users’ personal information is described through XML queries. We have proposed the general architecture of this system, and several key technologies for the establishment of such a system are also presented.

This paper presents a model of Add-Mult Fuzzy Neural Network (AMFNN) and the model’s architecture as well. Here, Error Back Propagation algorithm for AMFNN is presented according to the Gradient Descent Method. The result Compared with six representative fuzzy inference methods shows that AMFNN has high reasoning precision, wide application scope, strong generalization capability and easy implementation characteristics. Consequently, AMFNN has vast application prospect.

In this paper, we present the experiment results of three adaptive equalization algorithms: least-mean-square (LMS) algorithm, discrete cosine transform-least mean square (DCT-LMS) algorithm, and recursive least square (RLS) algorithm. Based on the experiments, we obtained that the convergence rate of LMS is slow; the convergence rate of RLS is great faster while the computational price is expensive; the performance of that two parameters of DCT-LMS are between the previous two algorithms, but still not good enough. Therefore we will propose an algorithm based on H₂ in a coming paper to solve the problems.

In the process of Web data mining, determining User Session is one important aspect for data preprocessing. According to the former four models of determining user session, the paper will introduce a method of uniting the two of the four models and offer the inferential algorithm to perfect it. Therefore, the method will be able to offer more logical data in the further application of data mining algorithm.

This paper designs a classifier by increasing new features and using mixed features extracting approach. The results show that the new added features can improve the performances of the classifier obviously and make the rate of recognition higher than before. If the probabilistic structure of problem is completely known, the new features will not increase the Bayes risk.

This paper briefly introduces the ISARITHMIC -- algorithm of congestion technique and analyzes its weakness, and presents a dynamic method to adopt two-level permit. In order to guarantee that throughput of network is of stability, this algorithm dynamically manages the permit -- length in a node and the totality of permit in the subnet. Performance (average of waiting time, optimal number of the 0-level permit) of this algorithm analysis is done, and then the corresponding mathematical model is shown in this paper. Finally, the future work is proposed.

The proliferation of signal processing is creating a pressing need for high-accuracy interpolation methods. The focus of this paper is to produce an interpolation approach based on discrete Fourier transform (DFT) that can greatly enhance accuracy of interpolation. The essential technique of the new method is to insert corresponding samples in frequency domain and the inverse DFT follows. What makes it very attractive is its undistortion to the foregone information in frequency domain. In addition, it is no need for the new method to define any interpolation functions. Numeric examples have been provided to illustrate the superiority and the effectiveness of the new interpolation method.

Path planning has received considerable attention in the research literatures. The A* method has long been used for planning purposes. It can find the smallest cost path if the path exists. But it is extremely slow for large numbers of cells to be searched. In this paper, we introduced an improved A* method. It increases the searching speed, and maintains the optimality and less memory requirement.

This paper proposes a novel CMAC neural network based on the Discrete Taylor Series (DTS-CMAC), including interpolation algorithm and training algorithm, which is capable of implementing error-free approximations to multi-variable polynomial functions of arbitrary order. The advantages it offers over conventional CMAC neural network are: high-precision of learning, much smaller memory requirement without the data-collision problem and much less computational effort for training and faster convergence rates than that attainable with multi-layer BP neural networks. Based on the novel CMAC, the author designs a type of a truly general manipulator intelligent control system, the simulation results verify that the control strategy is effective.

Triangle mesh is one of the important representations of 3-Dimension geometry and is broadly used in computer animation, and computer game. To efficiently represent and transmit the 3-D graphics, geometry compression is introduced to reduce the volume of geometric data. Connectivity compression is an important component of all geometry compression algorithms. In this paper, the relations between vertices of a triangle mesh are maintained by an adjacency matrix, based on which a connectivity coding strategy for 2-manifold triangle meshes is proposed. With this method, all the triangles in a mesh can be organized into a set of star strips which may be represented by a binary bit stream. The connectivity of a triangle mesh is encoded with no loss of information to an average of less then two bits per triangle. If the binary bit stream is compressed further by entropy coding using, less one bit per triangle can be reached.

A new method for fault diagnosis of nonlinear systems, which uses multi-step prediction of time series based on neural network, is presented in this paper. By using recurrent neural network, direct multi-step predictions for sampling series from multiple sensors are proceeded simultaneously, then two residual series, namely historical residual series and predicting residual series, are separately constructed from predicting series and sampling series. Lastly, several evaluation indexes used to detect whether exist fault(s) or not in a nonlinear system are derived. Simulation results show that the method is effective, and can strengthen the fault information.

Author Index