Research Projects Supported by the Competitive Research Funding 2003

General Research
1.Principal Investigator: Hirokuni Kurokawa

RIDEE(Real-time Interactive Distance Education Environment)

Amount: ¥1,500,000

2.Principal Investigator: Ryuichi Oka

Title: Media Integration

In order to integrate motion image and sentence description, a method called two-dimensional continuous dynamic programming is developed for image spotting recognition. It consists of two algorithms. The first is based on the assumption that both the reference image and input images are rectangular. A rectangular image usually includes both the object of interest and the background, so image spotting recognition performance can be affected by differences between the background in the reference image and that in the input image. The second algorithm extends the first algorithm to enable the use of both convex and concave reference images. It was used to obtain pixel-wise correspondence between a segment of an input image and a reference image. This correspondence was guaranteed by a monotonic restriction. Experiments demonstrated that the spotting recognition worked well and that monotonicity was guaranteed. In conclusion we have developed a sophisticated method to integrate objects in an image and their interpretation described by natural language.

Amount: ¥2,000,000

3.Principal Investigator: Roman Durikovic

Title: Real-time visualization of complex effects on Japanese lacquer ware

We collected data necessary to model the Japanese lacquer ware (called in Japanese 'urushi'), which are needed for physically based simulation and rendering of the appearance of lacquer ware products. We developed an approach for simulation and rendering of different kinds of a well known kind of lacquer ware, called 'makie urushi.' This type of urushi includes both metallic finish, obtained by sprinkling metallic powders, and colored urushi lacquer. Our approach includes processing digital photo in such a way that the resulting texture has an alpha component controlling the relationship between the contributions to reflectance coming from makie metallic finish and from colored urushi. Next, we modeled a prominent kind of Japanese lacquer ware, called 'nashiji urushi.' We used explicit modeling, which allows the user to obtain realistic zoom in and out and stereo pair images. The interior of LandRover car was painted with the urushi technique which used our simulation software prior to production.

Amount: ¥1,000,000

4.Principal Investigator: Daming Wei

Title: Modeling the heart vessel system for electrophysiological simulation studies

We have already developed a computer heart model for electrocardiographic simulation studies. The purpose of this study is to extend our model to be applied to an important medical application, the electrophysiological studies and catheter ablation, by developing the functions like: -A human model with a heart vessel system (a glass model) is to be developed to simulate the real catheter operation; -A corresponding computer model of the vessel system is to be developed to be linked the catheter operation; -A computer graphics system is to be developed to create a virtual environment of the catheter operation. Our heart mode is to be modified by incorporating the heart vessel system to be able to simulate the electrophysiological processes for electrophysiological study and catheter ablation.

Amount: ¥2,500,000

5.Principal Investigator: Kazuyoshi Mori

Title: Parallel Optimization of Stabilizing Controllers for Information Systems

We have considered parallel optimization of stabilizing controllers for multidimensional systems as one of information system. To increase the effectiveness of optimization, we have proposed a parallelization of algorithm. This is, basically, to parallelize the computation of Gr measuring procedures.

Amount: ¥2,300,000

10.Principal Investigator: Qiangfu Zhao

Title: General Seedbed On-line Learning and Knowledge Extraction for Mobile Robots

The purpose of this project is to develop a kind of mobile robots that can learn and grow up autonomously. For this purpose, we need 1) an eAEcient learning algorithm; 2) an eAEcient method to understand the learning results; and 3) an eAEcientway to accelerate future learning based on reuse of learned knowledge. First, we studied neural network based robot learning. Since the robot considered here should be autonomous, we should use algorithms that can train the robot without a human teacher. For this purpose, we adapted the evolutionary algorithm (EA). To simplify the problem, we first studied how to evolve a robot that can reach to a goal from any start point. The problem is to find through evolution a function (represented by neural network) that can make correct decisions based on the (16) sensor inputs. We conducted many experiments based on simulation, and found that evaluation based on multiple start points can yield robots that generalize well. Next, we tried to transform the learned neural network into understandable knowledge. This problem in general is a NP-complete problem. Therefore, we tried to map the neural network into a decision tree, which is often considered as an understandable model in machine learning context. For this purpose, we need a training data set. Since we do not have a training set in EA based learning, we should have a method to generate the training data after evolutionary learning. One simple approach is to generate the data at random, and use the outputs of the learned neural network as the teacher signals. This approach, however, is not eAEcient because many data so generated are meaningless in robot control. We thus proposed several methods for generating the data eAEciently and effectively, and confirmed them through experiments. Another goal of this project is to accelerate the learning speed through reuse of learned knowledge. This poses two problems. One is which knowledge to use; another is how to use the knowledge. As described above, we can map a neural network into a decision tree. The decision tree is actually a set of decision rules. Since the decision trees obtained so far are still very large, it is diAEcult to select useful rules manually. In the future, we will try to propose method that can select rules automatically, and reuse them to speedup further learning.

Amount: ¥1,000,000

11.Principal Investigator: Minyi Guo

Title: GridLinux: Research on Architecture for a Grid Operating System

The grid is an emerging infrastructure that will fundamentally change the way we think about - and use - computing. The grid will connect multiple regional and national computational grids to create a universal source of computing power. In a computational Grid environment, the central component of the architecture is the Grid operating system (Grid OS). The Grid OS is middleware that runs on all machines constituting the Grid. Because a Grid can have machines that range from high performance supercomputers to handheld PC and networks that range from gigabit-rate fiber networks to low-speed wireless LANs, it is necessary for the Grid OS to be adaptable. This project will study a novel, highly decentralized, Grid OS architecture that borrows features from Internet routing architectures. Our Grid OS design includes several novel ideas including (i) a flexible naming scheme, (ii)a service mobility protocol, and (iii) a highly decentralized Grid scheduling mechanism. The naming scheme supports aggregation of resource names based on attributes. In our architecture, the Grid OS runs on each node that participates in the Grid. Therefore, it is essential that the Grid OS is lightweight so that the overhead is minimal and at the same time powerful enough to support different services. Our Grid OS provides a service mobility protocol to migrate/replicate services depending on the demand.

Amount: ¥2,500,000

12.Principal Investigator: Victor Ryzhii

Title: Theoretical investigation and modeling of zero-resistance and zero-conductance states in two-dimensional electron systems

The research was focused on the theoretical study of the effect of absolute negative conductivity and its consequences predicted by V. Ryzhii in 1969 and discovered in recent experiments by Mani et al. and Zudov et al. The latter caused a surge of publications worldwide. The main achievement of the project is the development of scenario of the effect explaining the results of the experemental observation. For this purpose, the original theoretical model of 60s-70s was sigtnificantly extended. As a result of the research a significant progress in understanding of the effect and its features has been achieved. The results of the project are published in the following journal papers: 1.V. Ryzhii, R. Suris, and B. Shchamkhalova, Physica E, 2004, vol. 22, pp.13-18; 2. V. Ryzhii and R. Suris, J. Phys. Cond. Mat. 2003, vol.15, pp. 6855-6869; 3.V. Ryzhii and V. Vyurkov, Phys. Rev. B, 2003, vol.68, pp.1654061-1654068; 4.V. Ryzhii, Phys. Rev. B, 2003, vol.68, pp.19934021 - 19934024; 5.V. Ryzhii and A. Satou,J. Phys. Soc. Japan, 2003, vol.72, pp.2718-2721; 6.V. Ryzhii, Physica E, 2004, in press; 7.V. Ryzhii, J. Phys. Soc. Japan, 2004, vol.73, pp.1539-1542. They also presented at several international meetings.

Amount: ¥1,600,000

13.Principal Investigator: Jung-pil Shin

Title: Pen-Based Intelligence System

The purpose of this project is to develop the pen-based handwriting recognition and understanding algorithms with high performance, with themes such as: 1. Developing the stroke order- and number-free recognition algorithm, 2. Developing the signature evaluation, 3. Developing the original font generation algorithm reflected personal writing style.

Amount: ¥800,000

14.Principal Investigator: Nadia Berthouze

Title: Analysis and modeling of human gestures

We proposed a set of posture descriptors to support the creation of systems that can recognize human gestures. Two scenarios were considered: affective human gestures and skill levels of baseball batting motions. Discriminant analysis was used to evaluate the saliency of the descriptors in both scenarios. Using the same descriptors, we designed a categorization architecture, based on CALM network, that incrementally learns to recognize affective postures.

Amount: ¥800,000

15.Principal Investigator: Junji Kitamichi

Title: Project of Research for a High-level Design Method of Dynamically Reconfigurable Architecture Amount: ¥1,800,000

16.Principal Investigator: Jerold A. DeHart

Title: LIFT Project: Thesis Writing Textbook

The idea of the LIFT Project was to create a writing text to lift the writing abilities of students preparing their graduation thesis. Students need to spend most of their time writing their graduation thesis, and do not neet to spend the majority of their time reading about how to write the thesis. This text reduces reading and increases writing by presenting materials with visuals and text. The textbook is language oriented (L), instructs in the technical writing processes (I), considers documentation format of formal technical documents, (F) and is task-based (T) as learners work through various writing segments to develop their writing skills. The book has been written and incorporates many visuals to Lift the second language learner's comprehension level and composition skills.

Amount: ¥1,000,000

17.Principal Investigator: Robert H. Fujii

Title: Multi-Agent Robot Control Components and Optimization of Such With Probabilistic Algorithms

Objective: to investigate methods
to achieve cooperative behavior with autonomously operating robots. As an example of a cooperative behavior, the game of soccer played by a number of Aibo robots was chosen.
1) Algorithms were developed for accurate robot motion control and self-localization.
2) Co-evolutionary genetic algorithms were put forward which could be used to optimize robot motion. The asymptotic convergence of scaled genetic algorithms to global optima was analyzed.
3) Participated in two national Aibo robot competitions.

Amount: ¥1,500,000

18.Principal Investigator: Hisako Murakawa

Title: Discovering the Criteria of Spoken Data Taken From American Freshmen and University of AizuStudents to Prepare for the TOEFL Spoken Test Planned for 2005

The goal of this project was to determine the new criteria to effectively evaluate speaking ability tested in the TSE, which will be included in the TOEFL test starting in 2005. In order to obtain a high score on the TSE,an examinee must process the following language skills: listening, memorizing, summarizing, pronouncing, and many others. The title of this papers was Discoverying the Criteria of Spoken Data Taken From American Freshmen and University of Aizu Students to Prepare for the TOEFL Spoken Test Planned for 2005. The results were found according to the different sections of the test.

Amount: ¥2,000,000

19.Principal Investigator: Minoru Ueda

Title: 3D Web Environmental Science e-learning as an application of 3D geographical Information system

Web3D technology might be a good propaganda measure for appealing a better environmental management. 3D CG animation in two forms, Shockwave 3D and Viewpoint(VET) of Web 3D was verified, and their present problems were investigated. Using the Aral Sea shrinking problem in Uzbekistan as a content,animation was made using 3D CG software called Lightwave. Furthermore, the reproducibility of an object and a light source and the comparison of performance were also investigated. Neither Shockwave 3D nor VET produced the expected results. Viewpoint had more restrictions than Shockwave 3D regarding object and light source reproducibility. VET was superior for visitors, and Shockwave 3D was superior for Web 3D contents creators and webmasters. It is diAEcult to make an ideal animation with the present Web 3D from.

Amount: ¥300,000

UBIC Research
1.Principal Investigator: Subhash Bhalla

Title: Accessing Dynamic Contents in a Hospital Environment

Most users of the web are highly skilled at referring to tabular data. A natural inclination exists among the users for object-by-object traction to find information. We propose a convenient approach for the web information systems users. The users are assumed to be not skilled at using traditional database management system query interfaces such as the SQL language. The proposed interface introduces simplicity and avoids communication ambiguities. Based on the experiments,- A framework of high-level Query language facility has been developed for a selected application domain using the Hospital environment. The present study considers users, who need information contents as skilled users of all types of information based on available data contents. The present study has identified 'user-objects' approach and 'step-by-step navigation' approach. An effort is being made to implement these approaches for testing by users.

Amount: ¥1,900,000

2.Principal Investigator: Nobuyoshi Asai

Title: Design and Optimization of Running Water Bathtub with 1/f Functuation for Nursing Facility

Numerical simulations for Running Water Bathtub with 1/f Fluctuation peaked 10Hz have been examined. As the result of simulations, the fluctuation from motor will not disappear and the bathtub shape is not so important.

Amount: ¥1,2000,000

3.Principal Investigator: Incheon Paik

Title: Supporting the Automating Design in Supply Chains Based on Semantic Web Services

The integration of four design attributes (component-cost, quality, function, and technology) streamlines product design by providing a multilateral view to designers that leads to cost savings in the supply chain. Since the existing information infrastructure, based on server-side components for these design attributes, does not provide interoperable Web interfaces or semantic service descriptions for agents, improved architecture for the semantic Web service environment where automating agents can work more eAEciently was designed. Based on an improved ontological paradigm which integrates product design attributes organically to improve agent eAEciency, the building of a Web service environment for agents in DAML-S, which wraps the existing server-side software components to access attribute instance data, was carried out. A prototype of an application-centric agent for design support in this environment and its evaluation are also developed to illustrate the effectiveness of this Web service environment.

Amount: ¥1,600,000

4.Principal Investigator: Takafumi Hayashi

Title: High Precision measurement and High performance communication by Zero-Correlation-Zone sequence sets

Amount: ¥1,000,000

5.Principal Investigator: Carl Vilbrandt

Title: HAVEN-Heritage Aizu Virtual Education Network

The objective of the research is to establish a continuous, transparent learning environment associated with interactive, immersiveWeb based environments supporting community cultural heritage. After researching wireless technology for multimedia type systems, we selected the Sony Airboard LF-X1 hardware based on a thin client server approach as closest to meeting our needs including Japanese language support and a unique touch screen interface for a public interactive/multimedia kiosk based on local cultural heritage. Communitymembers have been trained in the use of digital media techniques and started recording their local history. Technical contributions have been made by researchers from West Virginia UniversityUSA on a US National Science Foundation grant to laser scan the Sanjusan Kannon of the Aizu area. Researchers from the University of Bordeaux France have optimized 3D imaging processing for handling the laser scan data. Fukushima Museumcollaborated inscanning of archeological models of haniwa figurines. Data collection and content creation for the educational kiosk are ongoing within the local community and with the West Virginia NSF supported team, who is continuing laser scanning of the Aizu temple and shrine architecture.

Amount: ¥1,500,000

6.Principal Investigator: Wenxi Chen

Title: Mental and Physical Health Evaluation Through Specific Voice Phrases in Daily Mobile Phone Use

Health is usually monitored using specific vital signs, such as body temperature, heart rate, as well as blood pressure. However, these parameters are not always sensitive enough to detect a minor change in health status. Furthermore, it is necessary to use medical devices to obtain such vital signs. This study aims to find more sensitive parameters from voice features in order to detect a subtle change in the mental and physical health at very early stages of the ailment development. While a user is talking on the mobile phone, health assessment is performed automatically and unawares by analyzing some specific voice phrases.

Amount: ¥1,000,000