Outline
In day by day life, humans perform speech recognition
Automatic speech recognition = speech recognition by computer
Recognition of speech from visual information only = lipreading.
- Lipreading is performed frequently by hearing-impaired individuals,
- But is also performed to some extent by normal hearing individuals, in noisy environments, as a support in audio recognition of speech
It has been noticed that the audio modality of speech and the visual modality of speech are complementary ? two sounds which are undistinguishable in one modality are distinguishable in the other.
Combining audio and visual information for speech recognition improves the speech recognition performance.
Department of Informatics
Aristotle University of Thessaloniki