Brief Status Summary for Automatic Speech Recognition at the start of the 80’s 800195
Automatic speech recognition has demonstrated high capability, first in the laboratory and now in field operations. This summary presents the characteristics of today’s research and commercial systems, and discusses issues and trade-offs affecting their performance. Independently conducted tests of Dialog System’s new speaker dependent, real time continuous speech recognition system demonstrated 99.5% word recognition accuracy in a task with a vocabulary of 10 digits plus 10 control words. The words were spoken by 5 speakers (3 male, 2 female) in a variety of regional American dialects against a background of heavy machine clatter and other noise.