speech recognition

In computing, any technique by which a computer can understand ordinary speech. Spoken words are divided into ‘frames’, each lasting about one-thirtieth of a second, which are converted to a wave form. These are then compared with a series of stored frames to determine the most likely word. Research into speech recognition started in 1938, but the technology did not become sufficiently developed for commercial applications until the late 1980s.

Discrete word recognition required a short pause between words. This was replaced in the early years of the 21st century by continuous speech recognition, which allows the use of normal but carefully articulated speech.