Site: Hitachi Research Laboratory
1-1, Omika-cho 7-chome
Hitachi-shi
Ibaraki-ken, 319-12, Japan
URL: http://www.hitachi.co.jp/

Date Visited: May 19, 1995

Report Author: J. Foley

ATTENDEES

JTEC:

J. Foley

Hosts:

Dr. Junzo Kawakami
Deputy General Manager; Manager, Planning Office
Shinya Tanifuji
Manager, Second Department of Systems Research
Koyo Katsura
Senior Chief Researcher, First Department, Systems Research
Masayuki Tani
Senior Chief Researcher, First Department, Systems Research
Dr. Ken'ichi Anjyo
Senior Researcher, Second Dept., Systems Research
Yoshio Iwase
Senior Researcher, International Coordination Office
Toshifumi Arai
Researcher, First Department, Systems Research

BACKGROUND

Hitachi is one of the three largest industrial corporations in Japan, along with Matsushita and Toyota. The company has 7 research labs:

In fiscal year 1994 (1 April 1994 to 31 March 1995), Hitachi reported annual sales of 3.7 trillion (~$37 billion) and profits of 56.4 billion (~$564 million). The company invested 10.2% of annual sales in R&D during that period. Hitachi has a total of 77,185 employees; HRL employs 843 researchers.

RESEARCH AND DEVELOPMENT ACTIVITIES

Graphics Hardware and User Interfaces

Koyo Katsura provided an overview of the activities of the First Systems Department in LSI for graphics hardware, graphics applications, and user interfaces. One commercially available game system uses two Hitachi SH2 RISC chips and a special purpose graphics chip developed in this group. The RISC chip is a 30-MIPS, 32-bit chip. The graphics chip is able to render texture mapped polygons for realistic scene generation.

A no-parallax LCD design has been developed to facilitate pen-based computing. The front plate of the LCD is essentially a big fiber-optic bundle that brings light to the surface of the display. The technology is currently too expensive for use.

An automotive navigation application was developed in the lab.

In the user interface area, Toshifumi Arai demonstrated InteractiveDESK (Arai et al. 1995), which extends Wellner and Newman's DigitalDesk, using the same notion of an overhead camera pointing down on a desk area to sense physical objects (in this case indicated by differently-shaped and colored labels on the objects), plus a bottom-projected large screen embedded in the desk with a pen-sensing overlay. A conventional display facing the user is mounted in a vertical panel at the back of the desk. Bringing real-world objects, such as a file folder with camera-sensible labels on it, causes a directory listing for that file folder to appear on the screen. Moving the keyboard into the middle of the desk causes information to be displayed on the conventional display, as the desk-embedded display is partially obscured. One portion of the tablet can be used for inputting kanji symbols, which are automatically recognized. This is a further move toward integrating computer and physical desktops. Multimedia Interfaces. Two user interfaces demonstrated in the Second Systems Department use multimedia. The first, called Object-Oriented Video, is intended for process monitoring and control applications. Object-Oriented video provides an operator interface that is a combination of traditional process flow diagrams (of the type used with power plants, chemical processes, and manufacturing plants) and actual video views of process equipment (boilers, rolling mills, furnaces, and control panels) (Tani et al. 1992). Live video images from cameras in a plant are shown to the operator, providing a more direct relationship to the process. Hyperlink "hot spots" on the videos provide more detailed video views or overlayed graphics giving more information about the current status (temperature, flow rates, etc.) of the physical object. Camera positions are fixed in the plant, allowing the hot spots to be defined by their screen coordinates. Control panels shown in the videos, consisting of slider dials, switches, and the like, can be directly manipulated in the video view to control the process: this requires that sliders and switches be controlled by servos linked to the computer system, which in practice is unlikely. User studies have not yet been conducted to determine whether the video images assist operators in their monitoring and control tasks.

The second system, Courtyard (Tani et al. 1994), integrates a large-screen display and several individual workstation displays for process control purposes. On the large display is an overall process flow diagram. Operators are able to call up on their individual displays more detail about objects on the large display. The same object might be shown differently in the detailed views, based on the information requirements of tasks assigned to each operator. Smooth techniques were developed for cursor movement between the large screen and individual displays.

The third system, User-Centered Video, is intended for applications in which multiple video windows are shown to a user but where bandwidth to the terminal is limited so that not all the videos are shown at high resolution and high refresh rates (Yamaashi et al. 1995). The question is how to allocate the available bandwidth to the various video data streams. This system allocates bandwidth based on an estimate of the user's interest in each of the videos. The "current" video window as indicated with the mouse and cursor is assumed to be of greatest interest; the user's interest in another video is assumed to be a negative exponential function of the distance from the current video to the other video. Also, the user can designate a region of a video to be shown with higher fidelity than the rest of the video. The user-centered video approach is appealing and appears viable. The next step would be to perform usability studies to determine the effectiveness of the heuristics used to measure a user's interest.

Modeling for Computer Graphics

Dr. Ken'ichi Anjyo showed a video illustrating several projects that seek to create realistic graphics images. In Takeuchi et al. 1992, flocking behaviors of different sorts are created by associating repulsive and attractive fields of different strengths with moving and fixed objects. Thus a car driving into a group of people causes the people to move away from the car, perhaps coming closer to one another than would happen were the car not present: the car repels people away from it with a greater force than people repel one another.

Human hair being blown by wind or going back and forth as one runs is modeled in great detail (Anjyo et al. 1992). Each strand of hair is considered to be up to 20 short segments, with a bending stiffness at each joint and a mass for each segment. Beautiful images are created with 50,000 strands of hair on a high-end workstation in less than a minute per frame.

Walking with emotion (Unuma and Takeuchi 1991) starts with digitized videos of humans walking with different strides (standard or basic walk, happy, purposeful, sulking). The basic joint movements for the standard walk are separated from the motions associated with the emotional walk. Then, new types of walks and interpolations between walks are created by using weighted combinations of the emotional walks added onto the basic walk. A further extension of this work allows real-time performance considering step constraints as well as emotional behaviors.

A stochastic fractal technique for modeling natural objects, such as clouds, terrains, and sea waves (Anjyo 1991), creates a large family of realistic-looking objects using a common mathematical framework. Parameters of the model have relatively intuitive meanings (directions of movements, silhouettes of shapes), allowing shapes to be created by those unfamiliar with fractals.

SUMMARY

The graphics hardware work primarily focuses on low-cost, moderate performance graphics for home entertainment systems. There is in general no comparable U.S. activity.

The InteractiveDESK work is in the same genre as earlier Xerox work and is a nice extension.

The process-control work (Object-Oriented Video, Courtyard, User-Centered Video) is on a par with and similar to work in Europe.

The computer graphics modeling work is very nice, and is competitive with work being done in the United States and Europe. Some of the techniques could be used right now in the context of games; other techniques await hardware performance gains.

All of the work demonstrated, with the possible exception of the nonparallax LCD display, has been reported in the literature in papers published between 1991 and 1995.

The work has a distinct engineering orientation, with no reported user experimentation to determine acceptability.

REFERENCES

Anjyo, K. 1991. Semi-globalization of stochastic spectral synthesis. The Visual Computer 7:1-12. Tokyo:Springer-Verlag.

Anjyo, K., Y. Usami, and T. Kurihara. 1992. A simple method for extracting the natural beauty of hair. In Proceedings SIGGRAPH '92 Conference, published as Computer Graphics 26(2):111-120.

Arai, T., K. Machii, S. Kuzunuki, and H. Shojima. 1995. InteractiveDESK: A computer-augmented desk which responds to operations on real objects. Interactive poster in CHI '95 Conference Companion, 141-2.

Takeuchi, R., M. Unuma, and K. Amakawa. 1992. Path planning and its application to human animation system. Creating and animating the virtual world, ed. N. M. Thalmann and D. Thalmann, 163 - 175. Tokyo:Springer-Verlag.

Tani, M., K. Yamaashi, K. Tanikoshi, M. Futakawa, and S. Tanifuji. 1992. Object-oriented video: Interaction with real-world objects through live video. Proceedings CHI '92, 593-98, 711-12.

Tani, M., M. Horita, K. Yamaashi, K. Tanikoshi, and M. Futakawa. 1994. Courtyard: Integrating shared overview on a large screen and per-user detail on individual screens. Proceedings CHI '94 XXX:44-49.

Unuma, M. and R. Takeuchi. 1991. Generation of human motion with emotion. Proceedings, Computer Animation 1991, 77-88.

Yamaashi, K, Y. Kawamata, M. Tani, and H. Matsumoto. 1995. User-centered video: Transmitting video images based on the user's interest. Proceedings CHI '95, 325-30, 591-92.


Published: March 1996; WTEC Hyper-Librarian