Search

Two-dimensional principal component analysis (2D-PCA) which is based on 2D image matrices as opposed to the standard PCA, which is based on 1D vectors, has been first successfully applied to face recognition and has higher accuracy than the latter. It was also successfully applied to other problems such as facial expression recognition, object recognition, etc. later. However, there exists still no neural network learning algorithm for 2D-PCA like those for PCA yet. In this paper, we propose a learning algorithm for 2D-PCA. Requiring no image covariance matrix evaluation and just repeatedly presenting training image samples to the single layer neural network, the desired multiple eigenvectors for 2D-PCA can be learned in the form of weight vectors of generalized linear neurons. It also profit from the parallel architecture of neural network. Simulation experiments are performed on YaleB face database, and the experimental results show that the proposed learning algorithm performs well as expected.

A novel Si LED with three terminals has been designed and fabricated compatible completely with standard n-well CMOS technology. It is composed by a combination of a forward biased p + -n junction controlling by gate voltage and a poly-Si/ultrathin oxide/Si tunneling junction. The experimental results demonstrates that: (1)The optical emitting power of LED increases with both forward p + -n junction current increasing and positive gate voltage increasing; (2) The optical emitting power of LED still increases wi th gate Voltage increasing while p + -n junction forward current at zero; (3) The spectra of the optical output power on wavelength occurs a peak near 1000nm. The results can be explained from the enhancement on the p-n junction forward current by the gate voltage induced barrier lowering effect and the S(poly-Si)OS tunneling junction theory. Keywords: Silicon LED, Light intensity controlling, Three terminals LED, SOS tunneling, Forward biased light emitting, CMOS technology, Barrier loweri ng effect, Infrared light emission

This paper aims to develop a Multiple-video-based E-learning Platform for Physical Education. The sports actions of teaching material can be shot and described by editing video into some desired clips for teaching and learning, and they would be stored into the action database. In this paper, we utilize multiple videos to display the sports action clips with three angle views which include in the front, the right front side and the left front side. The design concept is based on Instruction System Design(ISD). As a result, users can learn various sports skill within the shortest period of time by looking at the actions from our system. The system is expected to become a computerized education aid to sports action teaching and training, and to be the scientific and un-stereotyped system in Physical Education.

The more professional licenses, the more jobs for student can easily get in Taiwan. It is very important to promote some professional certification for students in vocational education. In this paper, we used the rule-space model to diagnose the cognitive skills for assessment evaluation and also considered the user's relevance feedback to improve the learner's learning achievement in ITE (Information Technology Expert) course. The rule-space model approach supports cognitive analysis of student's network technologies skill. And then to comprehensive what learning tops are weaknesses or strengths by individuals. The learner's relevance feedback provided user's evaluation that contained the item difficulty. We implemented a web testing application for ITE testing simulation. In our experiences, this approach can provide to understand the learning characteristic for e-learning achievement and to explore better educational application. The learner can fit Self-Regulation processing of the Scaffolding Theory and achieving social learning effectiveness, simultaneously.

In this paper, we propose a new inpainting algorithm based on color segmentation to preserve the structure of image. We use color segmentation to construct contour lines, and estimate the slope of contour lines and how they should extend. We also use the advantage of Bezier curves that fit the curves to the real world. The algorithm also fit to more general cases.

Learning video system is used in distance learning programs to deliver pre-recorded lectures. In this paper, we start our discussion on a multimedia presentation recording system which we had developed. We then present how the video recording concept can be further extended to realize a video segmentation system, which is based on video content information. The visual and audio information of a learning video is different. We also point out that a video classification technology can be further integrated to enhance the tool, to enable a more realistic and useful system for the discussion in distance education.

In this paper, an adaptive approach to skeleton detection is presented.First, we proposed a method to combine thinning and tracking algorithm which can find the skeleton tracking points, and the points would be reconstructed by using multiple cameras to obtain the 3D coordinates the X, Y, and Z-axis coordinates. The coordinates can be converted into three-dimensional human avatar. The joints of human avatar can be found by the thinning algorithm, and using object tracking to find the joints when the objects are disappeared. The avatar can be applied to the application such as Human computer interaction, entertainments, etc.

Martial art actions can be represented via VRML animations or extracted by video tracking. We propose an action retrieval method, which allows users to retrieve similar martial art actions. The mechanism is based on a similarity function that compares animation tracks. A representation of the human skeleton includes head, knee, elbow, wrist, etc and further aggregates important features in martial art actions. Different weights are dynamically calculated according to motion sensitivity of feature points. As a result, the system can automatically retrieve similar martial art actions. The results are tested by a professional kung fu master with good satisfaction.

With advanced technologies, computer devices have become smaller and powerful. As a result, many people enjoy ubiquitous learning using mobile devices such as Pocket PCs. Pocket PCs are easy to carry and use as a distance learning platform. In this paper, we focus on the issues of transferring the current PC based Sharable Content Object Reference Model (SCORM) to Pocket PC based. We will also introduce the Pocket SCORM Run-Time Environment (RTE) which has been developed in our lab. Pocket SCORM architecture is able to operate, even when the mobile device is off-line. It will keep the students’ learning record. When it is on line, the records will then be sent back to Learning Management System (LMS). With memory limitation, we provides course caching algorithm to exchange the course content on the Pocket SCORM.

VRML (Virtual Reality Modeling Language) is used to model the 3D object of the Web page, and become as a standard for 3D information of World Wide Web. We propose a content-based VRML 3D object retrieval system. The system can display 3D objects queries to Web browsers interactively. User may give a VRML object, and find VRML scenes that use the similar object via user interface. In this system, we propose the automatic indexing scheme of VRML database according to our Oct-tree representation method, which could filter the VRML object efficiently. Our experience shows that the system is able to retrieve 3D objects from VRML scenes satisfyingly by the proposed approaches.