Speech recognition via visual 3D image acquisition of speech movements
The recognition of speech mainly occurs via the audible sound. With the McGurk effect, however, it has been shown that everyone additionally integrates visual impressions into their language perception and increases the language comprehension. Professional lip readers are even able to recognize the language exclusively from facial movements. Automated speech recognition from acoustic signals is already commercially available as software and is integrated in many applications. For situations where the acoustic signal is disturbed or unavailable (e.g. in noisy environments, mutes), mechanical lip reading is required.
The I³-project "3D lip reader" carried out investigations on mechanical lip reading by means of fast 3D-measuring procedures.