SPEECH RECOGNITION

`` Every time I fired a linguist, our speech recognizer improved''

Fred Jelinek, IBM

At the centre of every speech recognition system is a speech recognition engine, which translates raw audio input into text. A ``speech aware'' system then takes the output from the speech engine and acts upon it. Since most speech aware applications deal only with the output from the speech recognition engine, it can be considered a black box. The inner workings of a speech recognition engine are beyond the scope of this paper.

However, since the types of speech recognition engine and the modes in which they operate are issues a designer will need to concern himself with, even when using an ``off-the-shelf'' product, they are discussed below.