Site: Electrotechnical Laboratory
Agency Of Industrial Science and Technology
Ministry Of International Trade And Industry
1-1-4 Umezono, Tsukuba
Ibaraki Prefecture 305, Japan
Date Visited: May 25, 1995
Report Author: T. Sheridan
ETL is one of the largest of Japan's national institutes, with 528 researchers, 120 of whom are in information processing. It was founded in 1891 as a testing lab. The annual R&D spending is ¥10 billion (~$100 million). The four major fields are information technology, energy, electronics, and standards and measurement. The Machine Understanding Division of ETL is divided into sections on machine inference, natural language, speech processing, and image understanding. The Intelligent Systems Division contains sections on computer vision, intelligent machine behavior, autonomous systems, interactive interface systems, and communicating intelligence.
A major new activity of ETL described by Dr. Nobuyuki Otsu (1993) is the Real World Computing (RWC) Program. Dr. Otsu directs 60 researchers on the RWC project for ETL. Based on massively parallel, optical computing, and neural system technologies, the goal of this project is to develop new and more flexible technology for recognition and understanding, problem solving, simulation and human interfaces, and autonomous and cooperative control. All of these areas are characterized by ill-defined problems, large amounts of data, and the requirements for flexibility, ability to cope with noise, ability to learn and adapt, and so forth. The RWC project was begun in 1992 and will run for 10 years on a budget of about $500 million. It involves 21 cooperating firms or institutions, including one in Germany, one in Sweden, one in the Netherlands, and one in Singapore. As examples of application, Dr. Otsu cited household robots and electronic secretaries and librarians. Dr. Toshikazu Kato described ETL work on image processing and graphical image matching (Hirata and Kato 1992; Kato 1995). An exemplary application is graphical trademark infringement. Data from experiments in subjective comparisons of logo images were learned by the computer and the computer could then decide if there is infringement (interestingly, males and females gave systematically different results). Dr. Fumiaki Tomita demonstrated research on vision sensing and corresponding touch shape display aimed, perhaps, at aiding persons with limited vision. The display was a coarse pin matrix, each pin driven by a stepper motor. Dr. Kazuyo Tanaka presented work on graphical human-machine interfaces augmented by speech recognition and synthesis (Katunobu et al. 1994; Itou et al. 1993; Hasegawa et al. 1995). Researchers in this project aspire to realistic facial expression and speech interaction, coupled with a mental model of the context or dialog situation (e.g., to identify the user).
Hasegawa, O., K. Itou, T. Kurita, S. Hayamizu, K. Tanaka, K. Yamamoto, and N. Otsu. 1995. Active agent-oriented multimodal interface system. IJCAI.
Hirata K. and T. Kato. 1992. Query by visual example. Lecture notes in computer science. In Advances in database technology, vol. 580, ed. A. Pirotte et al. Berlin: Springer-Verlag.
Itou, K., S. Hayamizu, K. Tanaka, and H. Tanaka. 1993. System design, data collection and evaluation of a speech dialog system. IEICI Trans. Information and Systems E76-D(1).
Itou, K., T. Akiba, O. Hasegawa, S. Hayamizu, and K. Tanaka. 1994. Collecting and analyzing nonverbal elements for maintenance of dialog using a wizard of oz simulation. 1994 International Conference on Spoken Language Processing, 18-22 September, Yokohama, Japan.
Kato, T. 1992. Database architecture for content image retrieval. Image Storage and Retrieval Systems. SPIE 1662.
Kato, T. 1995. Human media technology: Human-centered approach to information environment. Proc. Information Systems and Hypermedia, 23-24 March, University of Tokyo.
Otsu, N. 1993. Toward flexible intelligence: MITI's new program of Real World Computing. Proc. 13th IJCAI 93:1. France.