Automatic Speech Recognition

In William J. Hardcastle, John Laver, and Fiona E. Gibbon, editors, Handbook of Phonetic Sciences, chapter 22 (Wiley Blackwell, 2010, ISBN: 978-1-4051-4590-9) — on pp. 804—838.Speech recognition—the transcription of an acoustic speech signal into a string of words—is a hard problem owing to several cumulative sources of variation. Specifying the units of speech, such as phonemes, words, or syllables, is not straightforward and, whatever units are chosen, identifying the boundaries between them is challenging. Furthermore the relation between the acoustic speech signal and a symbolic sequence of units is complex due to phenomena such as varying rates of speech, context-dependences such as coarticulation, and prosodic effects.