Designing Effective Speech Interfaces

Speech Recognition

“My grandkids are going to look back at the keyboard and giggle.”

IBM product evangelist David Barnes, Home Office Computing (September 1999)

The term speech recognition refers to the technologies that enable computers or other electronic systems to identify the sound of a human voice, separate that sound from noise in the environment, and accept the messages from the voice as input for controlling the system.

Speech recognition has been a popular topic with technologists for decades. Although speech recognition has existed in research laboratories and as part of extremely sophisticated computer systems, only now has the technology become sufficiently advanced to be practical in typical household and business computer systems. The speech recognition software that is part of these typical systems can be divided into two categories: continuous and discrete.

Continuous versus Discrete Recognition

Continuous recognition allows a user to speak to a system in an everyday manner without using specific, learned commands. This technology is related to natural language, which will be discussed later. Although continuous recognition systems do exist, they are found mainly in niche markets, such as medical and defense systems, and in newer dictation systems. Continuous recognition systems are the technology promised by science fiction—several people interacting with a computer simultaneously and using normal language and speech patterns. Such systems go far beyond simply ...