Abstract:In this paper, a method of facial surface modeling in which the facial surface is designed in layers and in normal logical lattices is presented. Furthermore, on the basis of FACS(facial action coding system) and the characteristics of facial anatomy and mechanics of facial tissue, the principle of developing muscle actuators is presented as well as generating facial expression animations by using and coordinating corresponding muscle actuators.

Abstract:Abstract The rapid advances in distributed multimedia systems pose new challenges to the QoS(quality of service) management mechanisms. In this paper, the drawbacks of the QoS management mechanisms in the existing computer networks are introduced, and then the basic viewpoints about the QoS in the distributed multimedia systems are presented. The QoS parameters at each layer and the QoS management functions are discussed in detail. An example is also described. Finally, some conclusions are reached.

Abstract:A unification data model of vector and raster UNM is proposed in this paper. UNM divides the 0-, 1- and 2-dimension space into irregular sub-spaces. Relationships among these sub-spaces are explicitly built and these sub-spaces are organized as hierarchy. Features on the map are expressed as sets of some of these sub-spaces. Relationships among features can be inferred and calculated through their sub-spaces. UNM supports those analysed functions of both vector and raster structures.

Abstract:This paper proposes an EDHM(event-driven hypermedia model), which focuses on the task of defining and realizing hypermedia link, object synchronization and user interaction involved in a multimedia presentation. EDHM can be divided into three layers, which are data layer, container layer and presentation layer. Based on an event-driven scheme, timeline is applied to instinctively specify temporal and synchronous relationships between objects, and interactive objects together with an embedded simple script language——EasyScript handle the synchronization and action of objects. With an efficient hypermedia link solution, EDHM is able to support very large scale distributed demonstration system. Multiple authors can work simultaneously on a large project, and EDHM is suitable for inconstant authoring environment. An authoring tool called “Ark” based on EDHM has been used successfully.

Abstract:In order to extend distributed computing environment and support system integration, the authors are doing some research on multi-agent distributed integrated middleware CSE/MA to provide a general developing architecture of C/S applications for application developers and support developers to bulid various C/S applications easily. Firstly, this paper provides the architecture of CSE/MA, and discusses the components, principal and process of the applications developing with CSE/MA. Secondly, it introduces the services and implementing algrithoms about the management kernal of CSE/MA——register, and provides implementation framework of service agent and requuest agent and their design and implementation. Finally, it concludes the technical features and importance of CSE/MA。

Abstract:In this paper, the author takes parallel matrix multiplication as an example to see how their parallel implementations are under PVM, EXPRESS and NX environments. By analyzing this simple problem, the strategies of parallel programming are summarized for general purposes. The given parallel programming method is specially useful for parallel porting of a large scale application. The purpose of this paper is to provide a reference for scientists who want to write parallel programs.

Abstract:In the object-oriented dynamic programming environment, dynamic modification of a class will cause a problem that active objects of the class are difficult to handle. In this paper, a method for modifying dynamically classes based on cloned classes is proposed, and the comparison with other methods is given.

Abstract:This paper presents a parallel method of converting boundary to region quadtree. The method based on MIMD model has been implemented on DAWN1000 by comparing sequential results with the parallel results. The algorithm can be finished in O((B×logB)/P), where B is the number of chaincodes, and P is the number of processors. The algorithm can be applied widely in image processing, computer graphics, and pattern recognition etc.

Abstract:There is a conspicuous gap between dynamic rapid prototype development and high efficient product generation in the software developing environment, which leads to repetitive developing processes——develop rapid prototypes in the dynamic environment at first, and then code them in efficient static programming languages in order to generate high-performance products. As a consequence, it results in both waste of time and expenses. To fill the gap, the freeze／melt mechanism is proposed in this paper, which functions as smoothing the transition from dynamic prototypes to high efficient products.

Abstract:M. S. Chen has put forward heuristics G_MC and G_MR, which are used to produce a join bushy tree with less total cost. On the basis of his work, the paper gives an improved algorithm with complexity of O(n²), by means of analysing relationship between the order of join operations and computing costs. The algorithm can reduce more total cost of a join bushy tree than G_MC and G_MR, which benefits from the following: the operation order of two arbitrary adjacent internal nodes (join operations) is optimum.

Abstract:This paper presents a version control model of open hypermedia systems among other features, which supports version sets, permits exploring and managing alternate configurations, maintains document histories and supports cooperative work. The concepts of context node and version association table are proposed to solve the problem of inconsistency between structure version and content version effectively.

Abstract:This paper proposes an algorithm of automaticallytagging the POS(part of speech) of Chinese words which is based on integration of the statistical technique and the rule technique with the priority of the quantitative statistical analysis. The confidence intervals in the estimation of parameters is employed in the algorithm, and this makes the high-accuracy quantitative statistical technique as the top priority of tagging a corpus. Then the untagging part of the corpus is tagged in terms of rules, and some errors by statistics can be corrected by rules. Both closed and opened tests indicated that the accuracies of the algorithm are 98.9% and 98.1% respectively without consideration of both unknown words and segmentation errors.

Abstract:Scheduling is an important problem for many computer application areas. Job-Shop scheduling is a typical kind of difficult problems among them. It usually comprises several parallel goals, methods and resources available to realize the goals. Based on a practical Job-Shop problem model, the paper presents a chromosome representation method for genetic algorithm based on the scheduling problem. To fully use the knowledge of the problem, the authors also propose resource-space oriented and plan-space oriented genetic operators. The experiment result demonstrates that the performance of the method is satisfactory, and it also shows that the pre-mature problem can be solved by dynamically changing the probability of genetic operators or by scaling the fitness values of chromosomes.

Abstract:Automatic data distribution is a key technique in the area of automatic program parallelization for MPP(massively parallel processing) systems. Data distribution schemes directly influence the parallel execution performance of application programs in MPP systems. Taking MPP Fortran as an example, this paper discusses techniques of automatic data distribution in detail such as alignment analysis, shared array distribution, performance estimation for shared array distribution schemes, and shared array redistribution. The corresponding algorithms are presented and they will be implemented in these automatic program parallelization tools for MPP Fortran.

Abstract:This paper presents a learning algorithm of Hopfield neural network based on evolutionary programming with forgetting. The algorithm can avoid local minima by forgetting some individuals. Under constraints of fixed points, limit cycles or iteration sequences, the algorithm simultaneously acquires both the topology and weights for Hopfield neural network by solving inequalities. It copes with the limitations of evolving Hopfield learning algorithm. It can also find several optimal solutions. The experimental results also demonstrate the effectiveness of the algorithm.

Abstract:In this paper, an algorithm based on heuristic function for minimum set-covering problem is presented, as well as the definition of complete strategy concept and the method to structure heuristic function. The rationality, the time complexity and the precision of the solution are discussed for the algorithm. The basic idea is to structure heuristic function with the given heuristic strategies. The method can apply to other NP hard problems. As application of the algorithm, this paper presents a new algorithm of learning rule from the examples based on heuristic function.