Japan's RWC project offers glimpses into IT's future

A glimpse of the technology of the future, including products incorporating "intelligent" hardware and software, as well advances in parallel and distributed computing, was on display here at the Real World Computing (RWC) Project's final exhibit and symposium.

Launched by Japan's Ministry of Economy, Trade and Industry (METI) in 1992, the 10-year RWC Project's mission was to help make advances in information processing technology for the 21st century. As this year is its 10th and final year, the final exhibition and symposium, which ended late last week, offered product demonstrations and updates on research results.

A total of 54 research laboratories from 16 Japanese companies, one government institution and four foreign companies took part in the project, and all participated in the exhibition, said Noriko Minakoshi, a spokeswoman from the RWC.

Project research was divided into two categories: what the project organizers called "real world intelligence technology," or software and hardware equipped with artificial intelligence; and parallel and distributed computing technology. Nineteen percent of the research is currently ready to be launched commercially. RWC officials predict that, within two years, 42 percent of the research done as part of the project will be commercialized and in five years, that number will more than double, to 89 percent.

Among the technology developed in the intelligent technology category, aspects of the Cross Mediator software developed by Advanced Industrial Science and Technology (AIST) a government agency, have been adopted in products from by Fuji Xerox Co. Ltd.

Cross Mediator, which incorporates voice recognition technology, is a database search engine system for digital video and audio data. A user can locate the exact scene in a video, for example, by saying aloud a keyword into the voice recognition interface. Users can also locate songs in the database by humming tunes. The system is designed primarily to find data stored on hard drives; however, by using a still image as a "keyword," users can command the system to locate related images on the Internet.

Technology in the parallel computing category that is already on the market includes the PC clustering software called "SCore," which lets PCs work at a supercomputer-processing level by networking them. It was developed by one of the AIST's laboratories. NEC Corp. and Fujitsu Ltd. have adopted the technology in some of their products, an RWC official said.

Some of the potential research that may be commercialized in the future include, in the real world intelligence technology category:

-- Matsushita Electric Industrial Co. Ltd.'s "Digital Smart Pixel Array," which is a high-speed image parallel-processing sensor system. The technology allows a video camera to have an image processing time of 1 millisecond per frame, 30 times faster than a current CCD (charge coupled device). This means that a camera could capture, at one centimeter per frame, an object moving at speeds up to 30 kilometer per hour. Such as system would help, for instance, video cameras detect faults in chips in fast-moving manufacturing processes.

-- A high-speed face recognition system from Singapore's Kent Ridge Digital Lab. The system registers five still images of a person's face with a three-lens camera. Within a minute of processing, the system can recognize a person's face and match it to the correct name. The system has been tested within the company, which hopes to commercialize it in 2002.

In the parallel and distributed computing category, the RCW exhibit showed:

-- Sanyo Electric Co. Ltd.'s trademark searching system, based on pattern-matching software. Current technology requires inputting data about the color and shape of a pattern, and only allows for recognition of patterns when a camera is directly in front of them. The new Sanyo system needs to input only the contour data of a pattern, but in detail, registering the data in one-pixel-by-one pixel segments.

This level of detail allows the system to recognize logos and trademarks even when a camera is scanning three-dimensional moving images, or viewing them from distorted angles. The system can recognize, with 80 percent accuracy, images viewed at 50 degree angles, said the company. The system could be used, for example, by vendors to see, during a football game program, how often their logos and trademarks are exposed on television. The system is costly at present and the company is not marketing it.

However, it also could potentially be used in car navigation systems, according to Sanyo.