Showing papers by "Yuan Yan Tang published in 2002"

Stroke extraction and stroke sequence estimation on signatures

Abstract: This article addresses the problem of stroke extraction and stroke sequence estimation from a static signature. A new algorithm is proposed based on a simple but effective writer model. The results are encouraging.

Mathematical representation of a chinese character and its applications

Abstract: In this paper, a novel method to express Chinese characters mathematically is presented based on the knowledge of the structure of Chinese characters. Each Chinese character can be denoted by a mathematical expression in which the operands are components of Chinese characters and the operators are the location relations between the components. Five hundred five components are selected and 6 operators are defined to express all the Chinese characters successfully. These mathematical expressions of Chinese characters are simple, natural, and can be operated like the common mathematical expression of numbers. It makes Chinese information processing much simpler than before. This theory has been applied successfully in fonts automation, Chinese information transmission among different platforms and different operating systems on Internet, and knowledge discovery of the structure of Chinese characters. It can also be applied extensively to many areas such as typesetting, advertising, packing design, virtual library, network transmission, pattern recognition and Chinese mobile communication.

A wavelet-based method for fingerprint image enhancement

Abstract: The performance of AFIS (automatic fingerprint identification system) is heavily determined by the quality of the input image. Thus an effective method to enhance the fingerprint image is essential in each system. A new method based on the wavelet decomposition is proposed in this paper. We combine the texture filter method mostly used in nowadays with wavelet transform to achieve a more reliable and effective enhancement. Using this method, we obtain a more clarity fingerprint image, which can distinctly improve the precision of minutiae extraction module. Experimental results show that wavelet-based enhancement method is more effective and robust than the other existing methods such as filter-based and direct grey level approaches.

Multimodal interface for human-machine communication

Abstract: Introduction to multimodal interface for human-machine communication, P.C. Yuen et al. Algorithms - a face location and recognition system based on tangent distance, R. Mariani recognizing action units for facial expression analysis, Y.-L. Tian et al view synthesis under perspective projection, G.C. Feng et al. Single modality systems: sign language recognition, W. Gao and C. Wang helping designers create recognition-enabled interfaces, A.C. Long et al. Information retrieval: cross-language text retrieval by query translation using term re-weighting, I. Kang et al direct feature extraction in DCT domain and its applications in online web image retrieval for JPEG compressed images, G. Feng et al. Multimodality systems: advances in the robust processing of multimodal speech and pen systems, S. Oviatt information-theoretic fusion for multimodal interfaces, J.W. Fisher III and T. Darrell using virtual humans for multimodal communication in virtual reality and augmented reality, D. Thalmann.

Skeletonization of ribbon-like shapes with new wavelet function

Abstract: In this paper we propose a new scheme to extract the skeleton of ribbon-like shape with a novel wavelet function. It consists of two phases based on these perfect properties of the new wavelet function; and symmetry analyses of maxima moduli of wavelet transform are given. Midpoints of all pairs of contour elements are connected to generate a skeleton of the shape, which is defined as wavelet skeleton. Four basic criteria for modifying the artifacts of wavelet skeleton are presented. A corresponding algorithm is developed, and the experimental results are shown that this algorithm is capable of extracting exactly the skeleton of ribbon-like shape with different widths as well as different grey-levels.

Construction of orthogonal multiwavelet

Abstract: For compactly supported m-band (m/spl ges/2, m/spl isin/Z) orthonormal multiwavelet systems, an algorithm to construct compactly supported orthogonal multiwavelets from the associated multi-scaling functions is presented in this paper. This method is simple for computation, does not need to solve a set of nonlinear equations in unknown matrices or factorize a polynomial matrix into a special form, and is not restrained by the multiplicity of multiwavelets. It only needs to solve some linear equations and compute some matrices. As an example, the GHM multiwavelets are derived via this method.

Document segmentation using wavelet-domain multi-state hidden Markov models

Abstract: Presents a document segmentation algorithm, called context-adapted wavelet-domain hidden Markov tree (CAHMT) model, which extends the wavelet-domain hidden Markov tree (HMT) model. The proposed CAHMT can achieve more accurate quality with low computation complexity in document segmentation. In addition to further improving the segmenting performance, we combine a differential operator and the lowest frequency subband with CAHMT and produce much better visual segmentation quality than the HMT.

A novel post-classifier for search engine using hidden Markov model

Abstract: Hidden Markov models (HHMs), while well applied in fields such as speech recognition and optical character recognition, have not been used in post-classification for search engines. We explore the use of HMMs for optimization of search engines tasks, specifically focusing on how to construct a new model structure to improve the classification of web pages. We show that a manually constructed new structure model that contains only two states and two classes of observations per field can produce good classification results, and discuss strategies for learning the model structure automatically from data. We also demonstrate that the use of new structure model to classify the search results using some search engines and some different search keywords provide a significant improvement in search accuracy. Our models are applied to the task of post-classifying the web pages selected by the search engine Google, and achieve a classification accuracy of 93.4.

Order statistic filter (osf): a novel approach to document analysis

Abstract: Page segmentation is one of the important and basic research subjects of document analysis. There are two major kinds of page segmentation methods, i.e. hierarchical and no-hierarchical ones. Most traditional techniques such as top–down and bottom–up approaches belong to the hierarchical method. Though these two approaches have been used till now, they are not effective for processing documents with high geometric complexity and the process of splitting document needs iterative operations which is time consuming. A non-hierarchical method called the modified fractal signature (MFS) was presented in recent years. It can overcome the above weaknesses, however the MFS needs to calculate modified fractal signature which makes the theory very complex. In this thesis, we present a new page segmentation approach: Median Order Statistic Filter (MedOSF) — Maximum Order Statistic Filter (MaxOSF) approach which is more direct and much simpler. We use the MedOSF to remove the salt–pepper noise of the document and use the MaxOSF to do the page segmentation. In practice, they not only can adaptively process the documents with high geometrical complexity, but also save a lot of computing time.

New methodfor feature extraction basedon fractal behavior

Abstract: In this paper, a novel approach to feature extraction basedon fractal theory is presentedas a powerful technique in pattern recognition. This paper presents a new fractal feature that can be appliedto extract the feature of two-dimensional objects. It is constructedby a hybridfeature extraction combining wavelet analysis, central projection transformation andfractal theory. New fractal feature andfractal signatures are reported . A multiresolution family of the wavelets is also usedto compute information conserving micro-features. We employeda central projection methodto red uce the dimensionality of the original input pattern. A wavelet transformation technique to transform the derived pattern into a set of sub-patterns. Its fractal dimension can readily be computed, and to use the fractal dimension as the feature vectors. Moreover, a modi6ed fractal signature is also used to distinguish the distinct handwritten signatures. We expect that the proposedfractal methodcan also be usedfor improving the extraction andclassi6cation of features in pattern recognition. ? 2002 Pattern Recognition Society. Published by Elsevier Science Ltd. All rights reserved.

A novel method for harmonic geometric transformation model based on wavelet collocation

Abstract: Geometric distortion may occur in the data acquisition phase in information systems, and it can be characterized by some geometric transformation models. Once the distorted image is approximated by a certain geometric transformation model, we can apply its inverse transformation for the geometric restoration to remove the distortion. The harmonic model is a very important one, which can cover other linear and nonlinear geometric models. However, its implementation is very complicated, because it cannot be described by any fixed functions in mathematics. In fact, it is represented by a partial differential equation with a given boundary condition. In the paper a wavelet-based method is presented to handle the harmonic model. Our approach has two main advantages, the shape of an image is arbitrary and the program code is independent of the boundary. The performances are evaluated by experiments.