Research Project Supported by Competitive ResearchFunding 2008

General Research Principal Investigator: Yong Liu Title: A New Balance Ensemble Learning for Neural Network Ensembles Amount: Y=863,000 Generally speaking, data used in learning can fall into two groups of well-learned data and not-yet-learned data during learning depend on whether learners have learned the data or not. In the balanced ensemble learning, on those learning data where the ensemble has learned well, learning error functions for all learners in the ensemble would be changed to punish each learner from learning those welllearned data any further. Such change in learning error functions could allow the ensemble to have a little larger error values, i.e. a little larger difference between the target output and output of the ensemble in the case of the mean squared error, on those well-learned data so that the ensemble could shift its attention away from learning those well-learned data. Meanwhile, on those learning data where the ensemble has not yet learned, enlarged learning signals are given to each learner so that the individual learners in the ensemble could focus on those not-yetlearned data. Through shifting learning away from well-learned data and focusing learning on not-yet-learned data, an adaptive error balancing could be achieved in the ensemble. This research is quite different from previous work in ensemble learning. Most previous work on ensemble learning did not acknowledge or exploit the issues of how learning error functions should be adapted in the learning process, how individual learners interacts with the integration, and how such adaptation in learning error functions and interaction among learners can be used to influence and guide the learning of individual learners. Human often seem able to greatly reduce the time required to learn by communicating and working together in groups. The adaptive error functions proposed in this research provide a practical way for a neural network ensemble to communicate hypotheses and/or other information learned in each individual neural network in an attempt to converge to the learning target more rapidly. Two edited proceedings and two referred international papers have been published from this project: 1. L. Kang, Z. Cai, X. Yan, Y. Liu (eds.) Intelligence Computation and Applications, Proceedings of the third International Symposium on Intelligence Computation and Applications, Lecture Notes in Computer Science, Vol. 5370, Springer, 862 pages, 2008. Edited proceedings. 2. Z. Cai, X. Yan, Y. Liu, L. Kang, Progress in Intelligence Computation and Applications, China University of Geosciences Press, ISBN 978-7-5625-2303-1, 370 pages, 2008. Edited proceedings. Principal Investigator: Steven Tripp Title: Grammar of Space: A Semantic Approach Amount: Y=629,000 Previous research on descriptions of scenes (Tripp, 2006) indicated that students preferred absolute descriptions over relative and intrinsic descriptions when communication direction in English, but not in Japanese. The present research proposal involves three variations on previous studies. First, students navigate in the real world. Second they will discover, rather being shown, the locations that were the object of the experiment. Third they will describe motion in addition to directions. Locations will be programmed into a GPS device or a tablet computer running a navigation program. Students will be shown how to use the device to navigate to the example locations. Using the GPS device for navigation, students will discover the locations of several waypoints. After freely navigating on campus to the locations, students identified the locations by drawing them on a map, and then again by pointing to the locations from various assumed positions. Response time and direction of pointing were recorded. Finally, students were asked to describe in writing the discovered locations from other assumed locations, using either English or Japanese. Results showed that, although students found pointing to different locations from assumed locations approximately equally difficult; they showed a preference for using absolute terms of reference when writing English directions. When writing in Japanese, intrinsic and relative terms of reference were more frequent than when writing in English. Principal Investigator: Jonathan William deHaan Title: Language Learning with Computer Play and Computer Game Design Amount: Y=1,021,000 One main purpose of this project is to develop our students’s English and technology skills. Additionally, I want to understand how students learn English while playing computer games, and to understand how students use English as they learn to design computer games. There are two goals in this projects. Gial 1) Promoting and Researching Computer Game-Based Language Learning. Goal 2) Instructing and Researching Language Learning Through Computer Game Design. Principal Investigator: Victor Ryzhii Title: Graphene - New Material for 21 Century Electronics: Theoretocal and Numerical Studies of Novel Graphene Devices Amount: Y=1,467,000 Features of physical processes in graphene affecting the device operation were considered, in particular, energy relaxation of nonequilibrium carriers under interband photoexcitation, effect of Coulomb scattering of carriers in graphene, and photoconductivity of graphene. Several novel graphene-based devices were proposed and evaluated using the developed physical and mathematical models. The devices under consideration include graphene-based field-effect transistors (FETs), graphene nano-ribbon FETs, and graphene terahertz lasers with optical pumping. Original software was developed for numerical calculations of device characteristics. Analytical and numerical studies of the device operation were conducted. The device characteristics were calculated and the device ultimate parameters were estimated. Principal Investigator: Qiangfu Zhao Title: A Hierarchical system for visualization, analysis, and classification of face patterns Amount: Y=1,209,000 In the previous years, we introduced the nearest neighbor classification tree (NNCTree) for face/pose recognition and proposed an algorithm for inducing the NNCTrees. There are several advantages in using the NNC-Trees. First, the NNCTrees are transparent because the decision process can be transformed directly to understandable if-then rules. Second, the prototypes used in each NNC are abstracted mental images, and can be visualized whenever necessary. Third, we can get the whole picture of the feature space by examining the tree structure. In fact, the relationships between the nodes of the tree preserve the topological structure of the feature space. To reduce the cost for inducing NNC-Trees, we investigated three methods last year. The first one is attentional learning (AL), which tries to select important data during learning. The second one is distance preservation (DP), which reduces the number of times for calculating the distances between the data and the prototypes. The third one is dimensionality reduction (DR), which reduces the number of features of the problem space. As for the AL method, we studied two approaches. The first one is called attentional learning vector quantization (ALVQ), which can find important data dynamically during learning of the prototypes. The second one is called attentional R4-rule (ARR), which selects important data before learning of the prototypes. Experimental results on many public databases show that ALVQ plus DP is more useful. As for DR, we studied three approaches. The first one is the linear discriminant analysis (LDA). The second one is a direct centroid (DC) based approach, and the third one is the centroid based approximation (CBA) approach. Experimental results show that although DC is of low cost for DR, the extracted features may not be informative enough, and the total cost for training is actually higher. It seems that LDA is most suitable for our purpose. The CBA approach, although theoretically useful, is not good for most databases. Principal Investigator: Shuxue Ding Title: Sparse Representation of Signals Amount: Y=1,150,000 Sparse representation of signals is a most general transformation in the sense that it is a shortest linear representation for the signal. This representation is based on the sparsity of signal, which says that most samples of the signal is zero or near zero valued. This property is of statistics, and the most pure video or pure audio signals hold such a property. That is, one can introduce a dictionary for representation of the signals, which includes a group of basic ”bases” or ”words”, i.e., ”atoms”. By a representation for a group of signals, one can choose several words from it and can use their combination to represent any signal. A very attractive feature is that, if a sparse signal is corrupted with noise or mixed with other signals, it will lose their sparsity or become less sparse. That is, to represent the signal one need more words from the dictionary. The sparse representation is just find a signal processing that can recover the pure signal by requiring the output with a higher sparsity. In the researches of this year, we have focused on the relation between the sparsity and the anti-sparsity, i.e. most signal samples are non-zeros. We further applied this to beamforming problem in telecommunications. In the researches of this year, we have focused on the relation between the sparsity and the anti-sparsity, i.e. most signal samples are non-zeros. We further applied this to beamforming problem in telecommunications. Title: Visual Programming and Education Environment for the High-Performance Sony Playstation 3 and Cell Architectures Amount: Y=1,041,000 The research was devoted in several directions. The first direction was in developing a WWW-based programming environment in order to support designing, debugging and implementing different applications basetd on new multi-core architectures as well as educating and training students in practical work with the PS3 architecture. The other direction was devoted to design visual programming tools to improve the Cell programmer ’s productivity as well as allow the effective program code generation and debugging. As the basis of investigations, we used the“ movie-based ”visual programming technology representing an algorithm as an animation movie. The third research direction was in investigating and developing high-performance parallel Ce ll algorithms for the Tsunami Wave Propagation Modeling. The Tsunami Wave Propagation module of the Method of Splitting Tsunami was adapted for different parallel environments including Shared memory systems using OpenMP, Distributed memory systems using MPI, and the IBM CELL BE. This kind of problem can be characterized as rather difficult for parallel fine-grained implementation because of necessity to use intermediate matrix transpositions in order to keep data streams on a regular level. Nevertheless, we showed the possibility of a significant acceleration of this problem solution time for different programming platforms including a modern Cell environment with non-standard high-performance equipment like Sony PlayStation3. Principal Investigator: Naohito Nakasato Title: A Compiler for High Performance Adaptive Precision Computing Amount: Y=1,084,000 Recently, a number of arithmetic units in a processor is increasing while a growing speed of clock frequency of a processor is slowing down or even in plateau phase. As a result, it is normal that a usual office PC has dual core CPU on it. Furthermore, new processors used for special purpose such as computer graphics, game console and high performance computing (HPC) reside 8 – 100 computing cores. This recent trend opens us a new way of computing on such new processors with multiple computing cores (many-core SIMD computers). To take an advantage of a SIMD computer, we should program it with a special programming language or libraries. In addition, each SIMD computer has its own programming language or special techniques that we need to cope with. To tackle these difficulties, we propose and implement a compiler program for high performance adaptive precision computing.In this project, we investigate and implement new compiler software for easily utilizing many-core SIMD computers. Principal Investigator: Hiroshi Saito Title: An Asynchronous Processor for Embedded Systems Amount: Y=1,104,000 In this project, we designed an asynchronous MIPS processor using Verilog HDL. Then, we evaluated the processor in terms of latency and area targeting FPGA implementation using FPGA design tools. Finally, we proposed optimization methods to improve performance. Principal Investigator: Junya Terazono Title: Construction of WebGIS-based total geospatial data search environment Amount: Y=968,000 Mapping enviroment is an essential base for the location-based research in planetary science as well as terrestrial studies. Web-based geographical information system (Web-GIS) is now becoming a standard platform for displaying location information. As almost of all scientific resources are closely related with its location, the mapping environment should be a platform for not only information display but scientific production. We started to construct a Web-GIS based platform for collaboration environment used by planetary scientists. As a first step, we made some questionnaire for the planetary scientists to know their demands and requirements for our system. Then, we concluded that the first implementation for the system should be the security enhancement. In our implementation, security module ”Gridsite” was used to realize server-wide security capability. Using this module, Web-GIS based environment has additional security enhancement without any additional requirements for users’ individual terminals or web browsers, with flexible security application which is required for scientific research. Principal Investigator: Song Guo Title: Network Architectures and Protocols for Ultra-Wideband Wireless Ad Hoc Networks Amount: Y=1,013,000 Ultra-Wideband (UWB) wireless transmission is characterized by the use of radio signals with an extremely large bandwidth. Recent advances in high-speed integrated circuitry and signal processing ignited a renewed interest in the development and commercialization of UWB systems for personal communications. While UWB technology and its potential have been demonstrated by several companies and research labs in the recent past, the design of cost-effective and low-power UWB radios that provide high data-rate and/or long range and flexible networking capabilities remains a technical challenge. Conventionally, more bandwidth is required for higher data-rate transmission. However, due to spectral limitations, it is often impractical or sometimes very expensive to increase bandwidth. In this case, we shall explore the scheme using the multiple transmit and receive antennas for spectrally efficient transmission. In particular, we studied the multicast lifetime capacity of energy-limited UWB-based wireless ad hoc networks using multiple directional antennas by formulating and solving the corresponding optimization problem. Through both theoretical and simulation studies, we evaluated the tradeoff between the costs, in terms of the maximum number of antennas that the networks should support for each node, and the lifetime improvements by using the multiple antenna technology. The experimental results show that using two-beam directional antennas can exploit most lifetime capacity of the networks for both multicast and broadcast communications. Principal Investigator: Jung-pil Shin Title: A Study of Pen Computing Amount: Y=1,085,000 This research is about the usage of electronic-pen with computer and developing the algorithms for, including oriental brush implementation, signature/writer verification, and handwriting synthesis.The brush implementation is to realize oriental brush in computer. Through this system a user can interact with electronic-pen and computer as if it is the real brush.The signature/writer verification is an authorizing tool. People can prove him/herself through this system only by writing own name or just a few characters. We reduced error rates of conventional system dramatically using newly developed algorithms.The handwriting synthesis is a system artificially generating a person’s handwriting. Through this system, people can use their handwritten style characters in personal computers. It is also possible to characterize personal web-pages with the handwritten style characters. Principal Investigator: Keitaro Naruse Title: ARFID Based Position Detection System for Agricultural Mobile Robots Amount: Y=1,167,000 The objective of this project is to develop a position detection system using RFID for autonomous mobile robots which are applied to agricultural works. Our results are summarized as follows: (1) We have implemented the prototype of RFID based position detection system and we have evaluated it for the applicability to the agricultural mobile robot. As the results, it shows that the prototype is very good in functions and remaining problem is economical cost. (2) We have developed and evaluated a group control method of multiple mobile robots, in which each of them determines its motion autonomously communicating with others to make a stable group. The method shows a good stability and robustness. Principal Investigator: Hiroyuki Sagawa Title: Microscopic Study of Unstable Nuclei Amount: Y=1,208,000 The purpose of this project is to understand physical phenomena from quantum femto system to astrophysical systems by using microscopic models. The scale of the system is from femto meter to the size of the universe. 2008 was the second year ofthe same project. Pairing correlations is one of the most important subject in microscopic quantum physics. We continue to study the pairing correlations in unstable nuclei by Hartree-Fock(HF)-BCS mode and also HF-Bogolyubov model solving in the coordinate space. Especially the isospin dependence of the pairing interaction was applied for the study of excited states in unstable nuclei. We also pursue possible di-neutron correlations in neutron-rich nuclei. To this end, we solve the response function ofelectric quadrupole excitation and spin excitation taking into account the coupling to thecontinuum. It is known that the response function is very sensitive to the separation energy of valence neutrons and shows a huge enhancement just above the threshold when the last neutrons have a very small separation energy less than 1MeV. We study further correlations among several nuclear matter properties in the Skyrme HF and relativistic mean field (RMF) models. Linear correlations are found between neutron skin thickness , the pressure of the neutron matter and also the symmetry energy coefficient using various Skyrme Hamiltonian and Relativistic mean field Lagrangians. We will study electric dipole and charge exchange spin-monopole excitations of 90Zr and 208Pb are in order to extract the neutron skin thickness and to determine the neutron matter equation of states. Principal Investigator: Gennady Nikishkov Title: Atomic-scale modeling of nanostructures Amount: Y=968,000 Atomic-scale finite element procedure for modeling of self-positioning nanostructures has been developed. Our variant of the atomic-scale finite element method is based on a meshless approach and on the Tersoff interatomic potential function. The developed algorithm is used for determination of equilibrium configuration of atoms after nanostructure self-positioning. Dependency of the curvature radius of nanostructures on their thickness is investigated. It is found that for thin nanostructures the curvature radius is considerably smaller than predicted by continuum mechanics equations. Curvature radius variation with varying orientation of crystallographic axes is also modeled and results are compared to finite element continuum anisotropic solution. Principal Investigator: Mohamed Hamada Title: Collaborative Web-based Active Learning Tools with Learning style Index Amount: Y=583,000 In this project, a research using web-based active and collaborative learning in the theory of computation and related fields is introduced. The twofold contributions of this project are a novel use of existing technology to improve learning and a longitudinal quasi-experimental evaluation of its use in context. In addition to introducing a learning style index to enable students finding their learning preferences. Principal Investigator: Kazuyoshi Mori Title: Paramentrization of Strictry Causal Stabilizing Controllers Amount: Y=683,000 Since stabilizing controllers of a plant are not unique in general, the choice of the stabilizing controllers is important for the resulting closed loop. In the classical case, that is, in the case where the given plant admits coprime factorizations, the stabilizing controllers can be parametrized by the method called Youla-Kuceraparametrization. However, this parametrization may include stabilizing controllers which are not causal and may result a direct loop. We have established alternative parametrizations of stabilizing controllers, that is, the parametrizations of all strictly causal stabilizing controllers. These parametrizations include neither any non-causal stabilizing controller nor any direct loop. Partial result of this result will be published in IEEE Transactions on Automatic Control (to appear). Principal Investigator: Hirokuni Kurokawa Title: Skill Evaluation Support in Nuclear Power Plant Operator Training Amount: Y=1,236,000 This research is to support BWR Operator Training Center Corp. (BTC), an organization that develops and implements training of operators of BWR (Boiling Water Reactor)-type nuclear power plants, to realize quantitative evaluation of its trainees. One issue in quantitative evaluation is the automatic recognition of trainee’s behaviors and actions while dealing with an accident. The current step of this research is to find a way for this issue by using computer vision technology. Principal Investigator: Hitoshi Oi Title: Power-Efficient Virtual Machines for Wireless Sensor Networks Amount: Y=904,000 Virtual Machines (VMs) have been proposed as an efficient programming model for Wireless Sensor Network (WSN) devices. However, the processing overhead required for VM execution has a significant impact on the power consumption and battery lifetime of these devices. In this project, we explore the design space for the virtual machine architectures for wireless sensor node in the current and future implementation technologies. We also investigate the virtual machine applications that can utilize the proposed design. This includes the development of library for interfacing the proposed hardware modules and the main microcontroller on the sensor network node. Principal Investigator: Tomoyuki Yamakami Title: Designing Quantum Protocols to Protect Users’ Confidential and Private Information Amount: Y=821,000 This research is intended to design and test quantum cryptosystems and bring in a new efficient protocol, which expects a real-life implementation of a privacyprotection protocol. I will employ, for instance, the notions of quantum list decoding, limited quantum computing device, algorithmic argument, and ranom access coding. Moreover, the research will contribute to the advancement of quantum information technology. This research will ultimately lead us to the better understandings of quantum mechanics and thus nature itself. Another benefit comes from applications of my results in different fields of interest, including “locally decodable codes” in coding theory and “random self-reduction” in computational complexity theory. Principal Investigator: Irina Khmyrova Title: Computer simulation of heterodtructure devices for THz applications based on equivalent circuit approach Amount: Y=1,304,000 The project was focused on the study of fringing effects as a possible factor limiting resonant frequency of plasma oscillations in heterostructure devices similar to high-electron mobility transistor (HEMT). Analytical model based on the concept of ”gate extension” has been proposed and resonant plasma frequencies in the presence of fringing have been estimated. Cascaded transmission line model developed has been used in IsSpice simulation. The results obtained are in good agreement with experimental data. Principal Investigator: Maxim Ryzhii Title: Computer Simulation of novel nanoelectronis devices Amount: Y=835,000 Computer models and original software were developed for numerical simulations of novel nanoelectronic devices such as High-Electron-Mobility Transistor (HEMT) structures for terahertz radiation detection and generation, Graphene Nano-Ribbon field effect transistors (GNR-FET) and structures, and detectors based on quantumwell structures Analytical and numerical studies of the device operation were conducted and the device ultimate characteristics were estimated. The project results were published in 5 papers in refereed journals and presented at international conference. Journal papers: V. Ryzhii, M. Ryzhii, and T. Otsuji, ”Thermionic and tunneling transport mechanisms in graphene field-effect transistors”, Phys. Stat. Sol (a) 205, p.1527, 2008. V. Ryzhii, M. Ryzhii, A. Satou, and T. Otsuji, ”Current-voltage characteristics of a graphene nanoribbon field-effect transistor”, J. Appl. Phys. 103, p.094510 (2008). V. Ryzhii, V. Mitin, M. Ryzhii, N. Ryabova, and T. Otsuji, “ Device model for graphene nanoribbon phototransistor, ”Appl. Phys. Express vol. 1, p. 063002, 2008. V. Ryzhii, M. Ryzhii, and T. Otsuji, ”Tunneling current-voltage characteristics of graphene field-effect transistor”, Appl. Phys. Exp. 1, pp. 013001, 2008. T. Otsuji, T. Suemitsu, H.-C. Kang, H. Karasawa, Y. Miyamoto, H. Handa, M.Suemitsu, E. Sano, M. Ryzhii, and V.Ryzhii, “ Epitaxial graphene grown on Si substrate and its applications to electron devices ”, IEICE Technical Report, ED2008-224, SDM2008-216, (2009-2). Conference papers: M. Ryzhii, A. Satou, T. Otsuji, and V. Ryzhii, ”Dynamic characteristics of graphene nanoribbon field-effect transistors”, Ext. Abst. of the 2008 Int. Conf. on Solid State Device and Materials (SSDM2008), Tsukuba, Japan, September 24-26, 2008, p.598. UBIC Research Principal Investigator: Wenxi Chen Title: Estimation of Female Menstrual Cycle from Heart Rate during Sleep Amount: Y=1,448,000 The purpose of this study is to develop an automatic network-based system for monitoring heart rate during sleep, and to explore an Hidden Markov Model-based approach to estimate biphasic property throughout the menstrual cycle from a series of measured heart rate. Principal Investigator: Takafumi Hayashi Title: Information Integration using Messaging Network Amount: Y=1,880,000 This project will study an approach to construct an SOA-based system using a messaging network. The proposed solution uses a messaging networking, which is a type of overlay networking. The proposed approach can reduce the complexity and the maintenance cost of the system, which both makes an information system vulnerable. Therefore, the proposed approach can contribute to the construction of a secure information system. The requirements of a messaging network for an SOA-based system will be examined by using a test-bed system of our laboratory. The project will study of ”Mash-up” of web services by using messaging network technologies. We have already constructed the following test-bed systems: i) Weather-sensors information integration, and ii) Integration of sight-seeing information in Aizu area. A ”Messaging Network” refers to a form of network centric approach to enterprise application integration based on the extensive use of asynchronous messages together with a collection of well-documented reusable design patterns related to message handling. This project proposes the use of messaging networks as a foundation of sensor-data integration. The proposed scheme provides (a) Authentication/authorization of users and sensors, (b) Data-format conversion using message mediation, (c) Interface mediation for integration of devices, and (d) Mash-up mechanisms for Web services. The project studies an application of the messaging network architecture to the design of sensor networks. The key observation is that the network-centric approach the messaging network can provide is very suitable for the integration of sensor data in that the above issues are in many cases easy to resolve. Principal Investigator: Toshiaki Miyazaki Title: Die-hard Sensor Network for Monitoring Disaster-hit Region Amount: Y=2,464,000 In the case of disaster, quick and detail surveillance of the damaged area is very important to make a rescue plan to save human life. However, the detail surveillance is usually done by hands in actual rescue scenes because there are no effective aids. Our goal is to develop a wireless sensor network (WSN) system, so called ’die-hard sensor network’, which can monitor the disaster-hit region automatically by only scattering many sensor nodes in the region. The WSN has the following advantages; (a) no individual setup action is required in the field, (b) whole WSN system is never down and continuously monitors the disaster-hit region even if some sensor nodes involved in the WSN die, and (3) it has so scalable that we can expand the monitoring area and improve the sensing quality by only deploying extra sensor nodes as many as needed. In addition to these features, this system has a good cost performance compared to other surveillance systems such as a remotecontrolled helicopter. Thus, it can be applied to many critical disaster-hit regions, and we believe that our system must contribute not only to the rescue processes, but also to protecting the rescuers from secondary disasters. Principal Investigator: John Brine Title: International Business Writing using a Learning Management System and Flexible Language Acquisition Tools Amount: Y=1,029,000 This project employed EFL-based business writing pedagogy using a server-based learning platform and computer-supported collaborative writing techniques in order to equip students with business writing skills. In an effort to develop such a learning platform, we used Moodle, the widely popular learning management system, and Flexible Language Acquisition Tools (FLAX) to create a systematically structured and student-customized web-based business writing course for free public access. As part of this course design, our research provided formative feedback in the design and use of this software. We also explored the software needs for computer-supported collaborative writing. Further, we carried out a preliminary survey-based usability study to explore how this software can be used for genrebased language acquisition and content development with open source materials. The research provided a basis to prepare workshops designed for participants who expect to write English professionally. Successful employees know how to communicate clearly and effectively, changing writing style and content for varying audiences and purposes. The resulting workshops focused on the difficult task of meeting readers’ needs while simultaneously representing the employee ’s and employer’s best interests. To meet that end, the workshop assignments covered a variety of tasks required at work (e.g., e-mail messages, presentations, reports, proposals, minutes of meetings, etc.). Students who took the workshop improved their writing, especially in terms of clarity and precision. An important benefit of our approach has been the control of systematic group work while students made use of text analysis software tools. Mid-Term Plan Research Principal Investigator: Ryuichi Oka Title: Mult-Media Search System Amount: Y=1095,000 We have developed a new algorithm called two-dimensional Continuous Dynamic Programming (2DCDP) for multi-media search. The algorithm can perform optimally full pixel matching between images which enables to calculate robust similarity measure between images which is used for media search. The algorithm is applicable to many research themes. One of promising ones is free-viewpoint TV. Using 2DCDP three video cameras are enough for realizing a frame-wise 3D image which is viewed from arbitrary viewpoint of a user. This is called free-viewpoint TV which is regarded as the next generation TV. Principal Investigator: Michael Cohen Title: Cryptosteganographic affordances for multimodal interfaces Amount: Y=967,000 “Cluspi” is a watermarking system capable of yielding rectangular coordinates and orientation when a camera captures a section of appropriately printed material. We have deployed textsccluspi-encoded materials, starting with panoramic photographs, as an affordance to our locally-developed multimodal display suite. Because of the cyclic graphical nature (2 textscd depthlessness) of a “pano,” it can be displayed just as easily, if somewhat unnaturally, inside-out. In particular, such panos can be printed on cluspi-watermarked background and mounted on a cylindrical accessory, a conveniently sized light-weight manipulable, about the size of a can or bottle. Since the cluspi codes are subtle, they do not interfere with watermarked panos. We experiment with textsccluspi encoding as a tangible interface for virtual tour applications, especially when deployed in conjunction with a multimodal interface suite featuring panoramic browsing, spatial sound, optional stereographic display, and rotary motion platform. Keywords: Cryptosteganography, multimodal interface, panoramic interface Principal Investigator: Stanislav Sedukhin Title: 3D Toroidal Array Processor and Orbital Algorithems Amount: Y=1,703,000 We have designed and evaluated the performance of the 2- and 3-dimensional torus array processors and corresponding orbital algorithms for multidimensional transforms, including forward and inverse Fourier, Cosine/Sine, Hartley, Hadamard, and other separable transforms. 3-D Array Processor was implemented in FPGA for the 3-D video codec which is based on the forward and inverse 3-D Discrete Cosine Transforms. Principal Investigator: Vitaly Klyuev Title: Query-Specific Summarization of Web Pages Amount: Y=578,000 The role of the Internet as a source of any kind information has been changed dramatically over the past decade. There are a lot of tools to support users when they serf the Web but finding appropriate information is still difficult. The key reasons for such a situation can be characterized as follows. Texts are considered as“ bags of words ”by search engines. The primary goal of this project is to propose methods for construction of queryspecific indicative summaries that are specialized for a single information need, the user query. Their aim is to support end users in their decision making actions. According to the functional approach, information needs can be described using the natural language. The internal representation of documents preserves semantic relations between words. We designed methods responsible for finding fragments in the retrieved documents which are semantically related to the user query (user information needs). These fragments may be good candidates for indicative summaries. To design and implement our methods, we combined statistical and functional approaches. We implemented improved Web search service. The prototype of the search system was developed to support this service. Results were presented at several international conferences and published in the proceedings. Master students were involved in this project. Principal Investigator: Daming Wei Title: Biomedical Information Technology for Car Applications: A Brain-Computer Interface for Car Applications: a Basic Study for the Future Amount: Y=1,596,000 This year ’s research was focused on a basic study on a brain-computer interface (BCI) for car applications. In addition on EEG, we took Electroencephalogram (EEG) into consideration, and develop a sensing technique for this that is able to detect both EEG and ECG in one measurement by applying signal processing technology. We also conducted a simulation to realize“ driving by imagination ”, a dream at present time, and, feasibility in the in the future, by developing an ECG-base Brain-computer Interface. Main results: (1) M. Sakai and D.Wei, Separation of Electrocardiographic and Encephalographic Components based on Signal Averaging and Wavelet Shrinkage Techniques, Computers in Biology and Medicine, Vol. 39 (7) pp. 620-639, 2009. (2) Motoki Sakai and Daming Wei. Wavelet shrinkage applications of EEG-ECGbased human-computer interface. IEEE 8th International Conference on Computer and Information Technology (CIT 2008), 08-11 July 2008, Sydney, Australia, pp. 538-543. (3) Sakai, Information Processing for Human-Computer Interfaces Based on Electroencephalogram and Electrocardiogram, Ph.D. dissertation (Supervised by Daming Wei), 2009. (4) Shigeno, Driving Simulation by EEG-based BCI, GT thesis (Supervised by Daming Wei), 2009 Principal Investigator: Haruo Terasaka Title: Research on Micro-scale Regional Model (AIZU model) Amount: Y=1,142,000 A micro-scale regional model is indispensable component for the very high-resolution weather forecasting system for Aizu area. To develop this micro-scale regional model, we have downscaling of Meso-scale models such as RAMS, CReSS, MSSG, etc. We also have been attempting upscaling of CFD (Computational Fluid Dynamics) Models because of their sophisticated styling of numerical treatments. Principal Investigator: Kenichi Kuroda Title: A Desktop Super Computer System with Reconfigurable Devices Amount: Y=2,018,000 We focused on the SPH method for simulations of liquid dynamics and developed a acceleration system parallel processing on hardware and adopted FPGAs as a hardware accelerator (PROGRAPE system). We studied three contents for acceleration. First, we clarified the bottleneck in the simulation. From the profiling result, the momentum equation was the bottleneck in our simulation. Thus we decided to implement floating-point arithmetic units of the momentum equation on FPGAs. Second, floating-point arithmetic units that were specialized in the momentum equation were designed. Processor Generator for Reconfigu-rable systems (PGR) was used to design arithmetic pipelines. Third, we optimized precision of floating point arithmetic units. As a result, the sufficient precision of floating-point arithmetic units was 16 bit of mantissa and 8 bit of exponent. The performance evaluation showed the execution speed on the PROGRAPE system is 1.9 times faster than the software implementation. The host processes on PC should be accelerated for SPH Calculation in the FPGA Computing System. Using the Intel Threading Building Blocks, we parallelized the calculations of the density and the pressure force from wall boundary. The density calculation was 6.63 times faster than the default state and the pressure force calculation speed improved to 6.47 times at most. When the time integration steps varies, the speed increase is from 1.25 times to 2.83 times faster than default method. When the number of particles varies, the speed increase is from 1.64 times to 3.36 times faster than default method. Principal Investigator: Ben Abderazek Title: An In-Body Wireless Sensor Network System for Elderly Monitoring Amount: Y=675,000 Traditionally, personal medical monitoring systems have been used only to collect data. Data processing and analysis are performed off-line, making such devices impractical for continual monitoring and early detection of medical disorders. Systems with multiple sensors for physical rehabilitation often feature unwieldy wires between the sensors and the monitoring system. These wires may limit the patient’s activity and level of comfort and thus negatively influence the measured results. In addition, individual sensors often operate as stand-alone systems and usually do not offer flexibility and integration with third-party devices. Simply stated, the existing systems are rarely made affordable. Our project focuses of the research about efficient novel in-body wireless network system to enable medical staff monitor elderly mobility and health recovery remotely. The in-body network sensors monitors how a patient moves or picks up vital health signs and sends this information wirelessly to make an alert via a home hub or portable monitor. Our ultimate goal of this project is to research and prototype an in-body embedded system to assist old people and people with a disability maintain independence, and promote well-being by introducing smart in-body systems that allow medical professionals to initiate interventions in the home environment. We have already developed ECG processing algorithm for low power systems and also developed its hardware prototype. References: 1.Y. Haga, A. Ben Abdallah and K. Kuroda. Embedded MCSoC Architecture and Period-Peak Detection (PPD) Algorithm for ECG/EKG Processing, The 19th Intelligent System Symposium (FAN 2009), pp.298-303, Sep. 2009. 2.Abderazek Ben Abdallah, Arquimedes Canedo, Tsutomo Yoshinga and Masahiro Sowa.The QC-2 Parallel Queue Processor Architecture, Journal of Parallel and Distributed Computing, Vol. 68, No. 2, pp. 235-245, 2008. 3.Abderazek Ben Abdallah, Sotaro Kawata and Masahiro Sowa. Design and Architecture for an Embedded 32-bit QueueCore,?Journal of Embedded Computing, Special Issue in embedded single-chip multicore architectures, Vol. 2, No. 2 / 2006, pp. 191-205 Principal Investigator: Subhash Bhalla Title: Interfacing Mobile Devices with Spatial Web Service for Maps and General Applications Amount: Y=942,000 Often, a user uses a form based interface, in order to query a database. Increasingly, a vast amount of data is becoming available, to the user through web services and other means. There are many situations, where the form based approach to query is not suitable. Semi-skilled users in a technical profession have general purpose query requirements. While map data, that is available from Spatial web services is commonly used in many applications. Many new forms of data are emerging such as the Electronic Healthcare Records (EHRs). It is important to develope query capabilities to the level of database query languages, that are easy to use for semi-skilled persons. This effort will enable greater participation of skilled and semi-skilled professionals for general purpose query.

2010 THE UNIVERSITY OF AIZU ALL RIGHTS RESERVED