Seiichiro Hangai

Motion compensated frame rate conversion using normalized motion estimation

Frame rate conversion is effective to solve degradation of motion picture quality at liquid crystal display. It is also useful for low picture rate video signal such as movie. In case of conversion, basic motion compensation causes gap at covered area and uncovered area. To avoid the gap, we propose the motion compensation centered at the position of target. Fully motion compensated interpolation is possible in this method. To realize this, error normalizing by block activity is used for motion estimation. We also propose the adaptive interpolation by using four frames. Finally, we compare the original picture with the interpolated picture converted from decimated picture. The proposed conversion method gives 5 dB better at the maximum PSNR than the conventional method.

Signature verification adapting to intersession variability

In signature verification using on-line hand written data, we have cleared that the pen inclination data improves the verification rate[1]. However, the influences of intersession variability on the verification rate have not been considered yet. In order to avoid making error caused by a mismatch between his/her signature and his/her references, we applied some renewal methods to make the reference. From experimental results using signatures over 9 weeks, a reference renewal using authenticated data is effective to maintain high verification rate with 98.5%.

Improvement of On-line Signature Verification System Robust to Intersession Variability

In signature verification using on-line hand written data, we have shown that the pen inclination data improves the verification rate[1]. For the data of 9 weeks, we applied some reference renewal method to reduce the influence of intersession variability and obtained some improvement[2]. In this paper, we propose a threshold adaptation method, in which a reference renewal using the matching error of last five authenticated data is done. And, we show that the proposed method is effective to maintain verification rate with 91.2% using genuine and forgery signatures over 9 weeks.

Genre classification method for home videos

In this paper, we present a pioneering study on genre classification for home video. Analyzing home video is a challenging problem because it is generally unstructured and of low production quality in audio and video. Our approach is to define a set of genres referring to those in the actual video sharing site, and to extract salient low level features from MPEG compressed data which are robust to low production quality content. Experimental results based on ensemble learning show that our proposed method achieves around 0.7 to 0.8 F-measure values with 37.5 times faster processing than real-time playback for QVGA resolution home video.

A Study of Information Hiding Performance Using Simple Dot Pattern with Different Tile Sizes

There are several proposed techniques on the information hiding for printed document. Most of them detect embedded information using directional components which consists of a line element or group of dots. Therefore those techniques have serious problems regarding a visual quality and information capacity. In this paper, we use single dot pattern instead of the line element or group of dots. It is turned out that information capacity is improved for its small amount of dot and compact embedded area. We acquire possible information capacity 50688 bits in business card size with 93.45% information extraction accuracy. This information capacity is more than six times bigger than the conventional technique.

Information Hiding Method Utilizing Low Visible Natural Fiber Pattern for Printed Documents

Digital watermarks provide the capability to add extra information to various media, such as still images, movies and audio, by utilizing features of its content. Several methods for printed document have already been proposed using features of the text or images. In this paper we treat the paper media itself as content and propose a new information hiding methodology which utilizes the fiber pattern of paper. The proposed method is expected to improve the quality of watermarked documents and the geometric sensitivity in the watermarked documents. In this paper we show the principle of the paper-like pattern as a watermark, its possibility, advantages in terms of visual quality and geometric sensitivity.

Development and Evaluation of a Keyword-Accessible Lecture Video Player

It is desirable to build up a lecture video library to enable students to view past lectures at any time and from anywhere on their PCs. For this purpose, we have developed a lecture video player/maker system (Yoshida, 2002, 2003). In developing this system, we considered the usability for students and operability for teachers. The player includes a keyword access function, which enables the student to jump to scenes where one of the registered keywords was spoken. For this purpose, the maker realizes automatic index generation after continuous speech recognition of the whole lecture stream. In this paper, we discuss the structure and functions of an ideal lecture video player and the importance of the index corresponding to the scenes in which the related keywords are spoken. We also will present experimental results regarding keyword extraction from three lecture streams. Evaluations of the lecture videos and the player by students are also discussed.

A proposal of writer verification of hand written objects

From a biometric view point, it is preferable to recognize the writer from not only signature but also hand written characters or pictures. In on-line signature verification, we have shown that the combination of pen position, pen pressure and pen inclination information realized a high verification rate (see Hangai, S. et al., IEEE ICME2000, 2000; Yamanaka, S. et al., IEEE ICME2001, 2001). So, we apply the method to the writer verification of arbitrary written characters and pictures. In the verification, the information of direction and inclination of each segmented stroke are mainly used as feature parameters. From experimental results using hand written numeric characters, figures, and pictures by 20 persons, we can realize a verification rate of 92.59%.

An Improvement of Robustness Against Physical Attacks and Equipment Independence in Information Hiding Based on the Artificial Fiber Pattern

Digital watermarks provide the capability to insert additional information onto various media, such as still images, movies and audio, by utilizing features of the content. Several methods for printed documents have already been proposed using features of the text or images. In order to overcome the disadvantages of existing methods, we have proposed a new information hiding scheme for printed documents; namely, the artificial fiber pattern by using the features of the paper instead of those of the contents. The method has features of rotational invariance and low visibility. In this paper, we improve the algorithm aimed at equipment independence and the robustness against physical attacks.

An Identification of Japanese Numerical Characters on a Biometrical Smart Pen System

Biometrical smart pen (BiSP) prototypes were still used for handwriting signature verification and identification as well as for biometrical PIN identification based on Arabian number. In this paper, we apply a biometrical smart pen prototype to biometrical identification of PINs given by four handwritten Japanese numeric characters extracted from ten characters. As a result, it is estimated that the identification rate is 88% with one numeric character. And, it is found that the error occurs when the numbers, which have similar stroke time and similar direction in writing, are not distinguished clearly.