Catálogo de publicaciones - libros

In this paper, we propose an algorithm for extracting text regions from images in spam-mails. The Color Layer-Based Text Extraction(CLTE) algorithm divides the input image into eight planes as color layers. It extracts connected components on the eight planes, and then classifies them into either text regions or non-text. We also propose an algorithm to recover damaged text strokes in Korean text images. There are two types of damaged strokes: (1) middle strokes such as ‘⌉’ or ‘—’ are deleted, and (2) the first and last strokes such as ‘∘’ or ‘□’ are filled with black pixels. An experiment with 200 spammail images shows that the proposed approach is more accurate than conventional methods by over 10%.

Human-Artifact interaction in real world situations is currently an active area of research due to the importance foreseen of the social capabilities of near future robots and other intelligent artifacts in integrating them into the human society. In this paper a new paradigm for mutual intention in human-artifact interactions based on the embodied computing paradigm called is introduced with theoretical analysis of its relation to the embodiment framework. As examples of the practical use of the framework to replace traditional symbolic based intention understanding systems, the authors’ preliminary work in a real-world agent architecture (IECA) and a natural drawing environment (NaturalDraw) is briefed.

We propose a speech-based interactive guidance system based on document retrieval and presentation. In conventional audio guidance systems, such as those deployed in museums, the information flow is one-way and the content is fixed. To make the guidance interactive, we prepare two modes, a user-initiative retrieval/QA mode (pull-mode) and a system-initiative recommendation mode (push-mode), and switch between them according to the user’s state. In the user-initiative retrieval/QA mode, the user can ask questions about specific facts in the documents in addition to general queries. In the system-initiative recommendation mode, the system actively provides the information the user would be interested in. We implemented a navigation system containing Kyoto city information. The effectiveness of the proposed techniques was confirmed through a field trial by a number of real novice users.

Mobile devices are treasure boxes of personal information containing user’s context, personal schedule, diary, short messages, photos, and videos. Also, user’s usage information on Smartphone can be recorded on the device and they can be used as useful sources of high-level inference. Furthermore, stored multimedia contents can be also regarded as relevant evidences for inferring user’s daily life. Without user’s consciousness, the device continuously collects information and it can be used as an extended memory of human users. However, the amount of information collected is extremely huge and it is difficult to extract useful information manually from the raw data. In this paper, AniDiary (Anywhere Diary) is proposed to summarize user’s daily life in a form of cartoon-style diary. Because it is not efficient to show all events in a day, selected landmark events (memorable events) are automatically converted to the cartoon images. The identification of landmark events is done by modeling causal-effect relationships among various events with a number of Bayesian networks. Experimental results on synthetic data showed that the proposed system provides an efficient and user-friendly way to summarize user’s daily life.

This paper proposes a new forecasting model based on neural network with weighted fuzzy membership functions (NEWFM) concerning forecasting of turning points in business cycle by the composite index. NEWFM is a new model of neural networks to improve forecasting accuracy by using self adaptive weighted fuzzy membership functions. The locations and weights of the membership functions are adaptively trained, and then the fuzzy membership functions are combined by bounded sum. The implementation of the NEWFM demonstrates an excellent capability in the field of business cycle analysis.

In recent years, some methods have been presented for dealing with the concept map construction to provide adaptive learning guidance to students. In this paper, we present a new method to automatically construct concept maps based on fuzzy rules and students’ testing records. We apply the fuzzy set theory and fuzzy reasoning techniques to automatically construct concept maps and evaluate the relevance degrees between concepts. The proposed method provides a useful way to automatically construct concept maps in adaptive learning systems.

In this paper, the performance of the proposed fuzzy logic parallel interference cancellation (FLPIC) multiuser detector is evaluated for frequency-selective fading channels in wireless CDMA communication systems. A modified fuzzy logic system (FLS) with an adequate scaling factor (SF) is proposed to infer adequate partial factors (PFs) for the PIC scheme. Simulation results show that the proposed FLS can adapt to the large variations of users’ fading effects. Therefore, the FLPIC outperforms the conventional PIC (CPIC) and constant weight PIC (CWPIC) over two-path and three-path time-varying frequency-selective fading channels especially at heavy system load in DS-CDMA systems.

With the exponential growth of digital materials in this age, the protection of Intellectual Property Right (IPR) becomes an important and urgent topic. In this paper, we propose a novel digital watermarking scheme based on the Singular Value Decomposition (SVD) method and the Distributed Discrete Wavelet Transformation (DDWT) method. Our scheme transforms original image data from the spatial domain into the frequency domain by using the multi-scale DDWT technique, and then applies the SVD technique by modifying singular values of two sub-bands with watermark data and the DDWT watermarking embedding process on the rest two sub-bands. Thus, watermark information is embedded into the four sub-bands of the last scale. We exploit both of the advantages of the DDWT method, which is robust against one kind of geometric attack, i.e. the cropping attack, and the SVD method, which is robust against other geometric attacks and non-geometric attacks. Experimental results show that the quality of the stego-image is superior and the embedded watermark has high resistance against a variety of common geometric and non-geometric attacks.

This paper proposes a context-based image retrieval system based on color, area, and perimeter intercepted lengths of segmented objects in an image. It characterizes the shape of an object by its area and the intercepted lengths obtained by intercepting the object perimeter by eight lines with different orientations passing through the object center, and the object color by its mean and standard deviation (STD). Recently, we reported that the color-shape based method (CSBM) is better than conventional color histogram (CCH) and fuzzy color histogram (FCH) in retrieving computer-generated images. However, its performance is only fair in the retrieval of natural images. For CSBM, object color is treated as uniform by reducing the number of colors in an image to only 27 colors. In this paper, we improve the performance by representing the color features of an object with its mean and STD. During the image retrieval stage, -value is calculated based on the color features of two images, one in the query and the other in the database. The result shows that the proposed method achieves better performance in retrieving natural images compared to CCH, FCH, and CSBM. In the future, the proposed technique will be applied for the retrieval of digitized museum artifacts.

In general, we can easily determine the manufacturing step that does not function properly by referring to the flaw type. However, a successful segmentation of flaws is the prerequisite for the success of the subsequent flaw classification. It is worth noticing that, different segmentation methods are needed for different types of images. In the study, a mechanism that is capable of choosing a proper segmentation method automatically has been proposed. The mechanism employed artificial neural networks to select a suitable segmentation method from three methods, i.e., Otsu, HV standard deviation, and Gradient Otsu. The selection is based on the four features extracted from an image including standard deviation of background image, variance coefficient, the ratio of the width to height of both foreground and background histograms. The results show the success of the proposed mechanism. The high segmentation rate reflects the fact that the four carefully selected features are adequate.