摘要:

AbstractWe propose an improved extendible hashing scheme and a bucket multi‐versioning scheme that achieve a higher concurrency. In our improved extendible hashing, the global depth and directory entries are asynchronously modified to reduce lock conflicts on the directory. Furthermore, bucket multi‐versioning enables read‐only access to a bucket which is being split. Simulation studies show that these two methods provide an increase in speed in proportion to the number of processors and that they enable concurrent insertions and retrievals to be performed without either affecting the

ISSN:0882-1666    

DOI:10.1002/scj.4690261301

出版商:Wiley Subscription Services, Inc., A Wiley Company    

年代:1995

数据来源: WILEY

摘要:

AbstractIn the multiple access to moving picture data, there must be the simultaneous realization of the guarantee for the data transfer retaining the real‐time property, the fast response to the manipulation, and high multiplicity. For this purpose, there must be a multiple access scheduling method satisfying those requirements.In the first prototype of the video‐on‐demand server constructed by the authors, the fixed time‐slot scheduling (FTSS) is used. This system attempts to satisfy those requirements by utilizing the feature of the moving picture data that the access is continuous for the time interval of approximately one scene. FTSS, however, has a problem in that the efficiency is low, since the scheduling is made based on the worst timing of the mechanical operation, such as the varying the seek time and the rotational latency depending on the position of the disk access.This paper discusses the variable time‐slot scheduling (VTSS), which executes the scheduling based on the average time of the mechanical operation, and analyzes the efficiency. As a result of the analysis, it is seen that the multiplicity of VTSS is improved more remarkably with the improvement of the disk performance. In other words, the application can be expected to the video server in t

ISSN:0882-1666    

DOI:10.1002/scj.4690261302

出版商:Wiley Subscription Services, Inc., A Wiley Company    

年代:1995

数据来源: WILEY

摘要:

AbstractText region extraction is a necessary process before character recognition is done for document images. This paper describes a new algorithm, Linear Segment Linking (LSL), for text‐line extraction from document images. The algorithm groups together the piecewise linear elements in the document images, which may be assumed to be text lines, and then extracts them from the images. The algorithm requires less knowledge about document structure and is robust for distortion of the image. The primitive rectangles are introduced for the intermediate representation of image. It is easier and faster to create them than the usual circumscribing rectangles. A method of splitting the bridges between neighboring text lines is proposed. Combining the bridge splitting process with the text line extraction, the locally touching text lines will be extracted as individual one

ISSN:0882-1666    

DOI:10.1002/scj.4690261303

出版商:Wiley Subscription Services, Inc., A Wiley Company    

年代:1995

数据来源: WILEY

摘要:

AbstractAn automatic abstract generation system including a document structure analyzer is described. From a document, the system extracts a text structure representing rhetorical relations among sentences and sentence chunks. The system evaluates sentence importance based on the analyzed structure and decides which sentence should be discarded from an abstract. It also attempts to generate an abstract consistent with the original text by replacing connective expressions.Generated abstracts were evaluated from two points of view: the cover rate of key sentences; and the quality as document presentation media. Both experimental results proved the generated abstracts to be valid.

ISSN:0882-1666    

DOI:10.1002/scj.4690261304

出版商:Wiley Subscription Services, Inc., A Wiley Company    

年代:1995

数据来源: WILEY

摘要:

AbstractWhen a natural object or phenomenon is modelled mathematically by using a fractal dimension, it is recognized that the object or phenomenon does not often have an autocorrelation. In the fractal theory, a fractal dimension is defined as a determined value (dimension) which is independent of the scale of its covering. In practice, however, the dimension often fluctuates depending on samples used (even with the same object or phenomenon) and its scale.In published works of image analyses using fractal dimensions, it has been assumed that fractal dimensions are determined and independent of scales, ignoring the practical experience. To treat a natural object or phenomenon without contradictions, this paper introduces “stochastic local fractal variables.” Using these new variables, mathematical models of dust on LSI wafers are constructed, and they are identified by using the likelihood method. Usefulness of the proposed method has successfully been confirmed with real d

ISSN:0882-1666    

DOI:10.1002/scj.4690261305

出版商:Wiley Subscription Services, Inc., A Wiley Company    

年代:1995

数据来源: WILEY

摘要:

AbstractNeural networks trained via backpropagation are now widely applied in a pattern recognition method. However, since it becomes much more difficult for a network to accomplish its task when the number of categories increases, research on multiple network combination is active. Novel learning and recognition processes are proposed here. It is shown that by combining small‐scale neural networks, the proposed method allows exploitation of the potential capabilities of the networks. In the learning process, multiple networks are trained with patterns organized in overlapping groups. During the recognition process, response is obtained by making the networks compete with each other. In experiments involving recognition of individuals from various facial images and different expressions, a recognition rate of higher than 96 percent was obtained for 20 individuals and 131 images. Furthermore, results of simulations in which noise was added confirmed that the proposed method is robust with respect to pattern change

ISSN:0882-1666    

DOI:10.1002/scj.4690261306

出版商:Wiley Subscription Services, Inc., A Wiley Company    

年代:1995

数据来源: WILEY

摘要:

AbstractThis paper proposes a new method of contour extraction using active contour models for the purpose of image structuring. Active Contour Models, such as “snakes,” are based on energy minimization, which takes a great deal of computational time. In the proposed method, the active contour models consist of sampled contour points, and each sampled contour point is guided and moves in parallel by pressure exerted perpendicularly to the contour, attraction between neighbor boundary points, and repulsion from edge. Since each contour point is independently guided by a resultant force, the proposed method has the following characteristics: noise tolerance; parallel execution; and low computational cost. Experimental results on real pictures show the effectiveness of this met

ISSN:0882-1666    

DOI:10.1002/scj.4690261307

出版商:Wiley Subscription Services, Inc., A Wiley Company    

年代:1995

数据来源: WILEY

摘要:

AbstractSpeech conversion can be considered as the most comfortable means of information transmission between the human and the machine. To realize such a mechanism, there must be a technique for efficient natural language processing.This paper proposes a new topic‐control mechanism based on the statistical technique as a means to ameliorate the complexity of the natural language processing. The input speech is analyzed using the predictive CFG. The sequence of generated grammatical rules in used to drive HMM, and the topic is identified as the bias of the state distribution. Conversely, the bias of the distribution is reflected dynamically on the probabilities of the grammatical rules, thus, recognition search space is narrowed. Using the large‐scale dialogue database, the reduction effect for the perplexity to the text data input is evaluated, and it is verified that the proposed method has a high power to narrow the search sp

ISSN:0882-1666    

DOI:10.1002/scj.4690261308

出版商:Wiley Subscription Services, Inc., A Wiley Company    

年代:1995

数据来源: WILEY

摘要:

AbstractA method is proposed to verify seal impressions automatically. It has the following characteristics, and its effects are proved: (1) efficiency to judge genuine seals, even if the seals are degraded to certain extent, as in blur, scratchy or spotty impressions; (2) safety to reject false seals rigorously. The authors precise verification method uses small block information of seal conditions to cope with the above contradictory characteristics. Also, seal impression models are proposed to evaluate the trade‐off by simulation. Parameters in the method are tuned using the models, and by experiments it is proved that the method has both efficiency and safet

ISSN:0882-1666    

DOI:10.1002/scj.4690261309

出版商:Wiley Subscription Services, Inc., A Wiley Company    

年代:1995

数据来源: WILEY

摘要:

AbstractThis paper presents a neural network algorithm called SOLAR (Supervised One‐shot Learning Algorithm for Real‐valued inputs) that can complete learning by a single presentation of the real‐valued inputs.The similarity matrix is introduced in SOLAR, which can measure the Euclidean distance between training instances represented by a real number. Based on the geometrical structure of the similarity matrix, the structure of the network, the connection weights, the learning parameters and the linear discriminant function are determined. Then, the learning is executed by a single presentation of the training set. The nonlinearly separable case can be reduced to the linearly separable case by decomposing the set of supervisor signals into linearly separable sets.The main contribution of this paper is that SOLAR realizes a high‐speed learning by a single presentation of the input data composed of real values and a large‐scale complex set of data can be learned. This has been difficult in previous models.In addition, since the number of hidden units and the parameters are determined by the algorithm, there is no problem such as trial‐and error.Simulations show the effectiveness of the proposed method by learning a relatively large‐scale problem, such as the backpropagation algorithm, which has been difficult based on the gradient d

ISSN:0882-1666    

DOI:10.1002/scj.4690261310

出版商:Wiley Subscription Services, Inc., A Wiley Company    

年代:1995

数据来源: WILEY