September 1999, Issue 110

Taking Orders:
A Speech Recognition Module

RECOGNITION STYLE

Speech-recognition systems deal with another factor concerning the style of speech they can recognize. There are three distinct styles of speech recognition—isolated, connected, and continuous.

Isolated speech-recognition systems only handle words that are spoken separately (i.e., the user must pause between each word spoken). Ideally, the word is isolated by a moment of silence before and after it is spoken. This is the most common speech-recognition system available today.

Connected systems are a halfway point between isolated word and continuous speech recognition that permit users to speak multiple words. The VoiceDirect module recognizes verbalizations up to 3.2 s long. However, there shouldn’t be a pause or period of silence longer than 0.5 s during the verbalization.

Continuous speech is the natural conversational speech we use everyday. It’s extremely difficult for a recognizer to sift through speech because the words tend to merge together (a string like "Hi, how are you doing" sounds more like "Hiowryudoin"). Continuous speech-recognition systems are on the market and are under continual development.