|
Continuous-speech recognition is significantly more difficult
than isolated-word recognition. Words often blur together
with no pauses between them, and the characteristics of a
word ending depend in part on the word that follows. Moreover,
silences may be difficult to distinguish from non-voiced phonemes
(such as sibilants) in the presence of noise.
Tanner Labs has been working towards extending our speech
recognition technology to recognize continuous speech in real
time.
|
