A type of speech recognition that is used to recognise isolated words, as compared to continuous speech recognition. Although largely superseded, it is still used for some applications. For example, the algorithms are simple enough that some of the newer mobile phones can call a number that you say to them and this is done in hardware; or sometimes automated call response systems make use of similar recognition systems when they give you a set of options to choose from.

Are usually implemented using either dynamic time warping or hidden Markov models.