Speech Recognition

Speech Recognition, in computer science, the ability of a computer to understand the spoken word for the purpose of receiving commands and data input from the speaker. Some systems have been developed that can recognize limited vocabularies as spoken by specific individuals, but developing a system that deals with a variety of speech patterns and accents, as well as with the various ways in which a request or a statement can be made, has so far proved a daunting task for systems designers.

Designers of speech recognition systems have pursued three approaches. One enables computers to recognize a small number of words spoken by a broad range of people. This method has proven to be fairly accurate. The second method enables the recognition of a larger number of words spoken by one individual. This method is also fairly reliable provided the speaker's speech patterns are consistent. The third method enables computers to learn a speaker's speech patterns and to apply the patterns over a broad range of words. Although this last method provides the most promise for full speech-to-text conversion, it is the least accurate of the three. See also Speech Synthesis.