Site: Matsushita Electric Industries, Central Laboratory
3-4 Hikaridai, Seika-cho
Souraku-gun, Kyoto 619-02, Japan
Tel.: 0774-98-2509
Fax: 0774-98-2587
URL: http://www.mei.co.jp/index.html

Date Visited: May 22, 1995

Report Author: T. Sheridan

ATTENDEES

JTEC:

S. Chipman
J. Foley E. Glinert
J. Hollan
R. Kraut T. Sheridan
T. Skelly

Hosts:

T. Mitsumata
General Manager, Central Research Laboratories
Y. Yamada
Leader, Central Research Laboratories
K. Matsui
Leader, Central Research Laboratories
S. Maruno
Leader, Central Research Laboratories
E. Nomiyama
Central Research Laboratories
T. Nakanishi
Senior Coordinator, Overseas Department
Y. Ohiwa
Coordinator, Overseas Department

BACKGROUND

Matsushita was founded in 1918 by Konosuke Matsushita. The company supports the development of products under many trade names such as Panasonic, National, Technics and Quasar. Matsushita's product lines are in audio, video, home appliances, information, communication, housing, building, and industrial products. At the time of the JTEC visit the company had 254,000 employees and annual sales of $64 billion. R&D expenditures were $3.71 billion, or 5.8% of sales.

There are 250 researchers in the Central Lab, which interacts with both the product development and the business (sales) divisions. Human interface is one of eight departments in this lab. Matsushita operates a speech technology lab in Santa Barbara, CA, and 9 other offices or labs in the United States.

RESEARCH AND DEVELOPMENT ACTIVITIES

Y. Yamada made a presentation on speech training technology, primarily on speech training for the deaf. By means of sensors on the subject's neck, nose, and tongue (63 electrodes!), and measurement of airflow in addition to microphone signal, subjects are able to receive feedback on whether they are saying things properly and how to improve. After subjects receive sufficient training, Matsushita's system is significantly better than conventional methods. Two hundred of the systems are now being used in schools for the deaf. Interactive displays were also demonstrated that encourage speakers to give commands to make a basket, or to make it rain, and so forth.

Yoshizumi discussed the company's digital hearing aid technology, which makes extensive use of DSP chips. One version is 59 x 63 x 26 mm and weighs 98 g. A new system has interesting features, such as suppression of impulsive sounds (noises), spectral shaping to enhance higher frequencies, means to separate consonant spectra and otherwise improve consonant perception. The clarity enhancement was demonstrated. K. Matsui presented and demonstrated the firm's work on voice synthesis and recognition technology for both Japanese and English. Matsushita is interested in inputting speech recognition into consumer products (a VCR with speech recognition adds only $30), car navigation, e-mail applications, public information services, and an electronic receptionist. S. Maruno made a presentation of research on neural network and fuzzy technologies. The company's approach utilizes a multifunction layered network with a quantizer neuron model. This has been implemented in a Matshushita-built neural chip. The company's researchers also are experimenting with neural networks interacting with symbolic processing. They are developing applications to such products as washing machines, air conditioners, and camcorders. The JTEC panel was finally introduced to a fascinating project to study the linear and rotary motors inherent in bacteria flagella, which rotate at 100,000 rpm.

REFERENCES

Matsushita. 1994. Proceedings, Second International Symposium on Speed and Hearing Sciences. Sponsored by Matsushita Electric Industries, Ltd., 24-25 September, in Osaka.

_____. N.d. Matsushita speech recognition technology. Internal report.

Mekata, T., Y. Yoshizumi, Y. Kato, E. Noguchi, and Y. Yamada. 1994. Development of a portable multi-function digital hearing aid. Proceedings, ICSLP 94, Yokohama.

Tsukoba, E. and J. Nakahashi. 1994. On the fuzzy vector quantization based hidden markov model. IEEE paper 0-7803-1775-0/94.

Zhao, X., and H. Wakita. N.d. Experiments with a speaker independent continuous speech recognition system on the time database.


Published: March 1996; WTEC Hyper-Librarian