SLEDS SLEDS (SLEepiness Detection in Speech) aims at identifying
phonetic-linguistic changes
of speech induced by sleepiness, and at developing an
according automatic procedure for measuring sleepiness.
Cognitive-physiological effects on the speech production
process can lead to a reduced fluency (self corrections,
interruptions, Pauses), flattened intonation, imprecise
accentuation, slurry articulation, and a stronger
breathiness and nasality. On the linguistic side, simplified
syntactic structures and reduced vocabulary are expected.
For validating these hypotheses, a sleep deprivated speech
corpus is recorded. Using the recent advances in automatic
emotion recognition in speech, a comprehensive feature set
will be extracted, which includes through the use of
automatic speech recognition, for the first time also
linguistic information. The resulting measurement method can
be used in communication-intensice tasks (such as pilot/air-
traffic interaction) for the continuous monitoring of the
sleepiness state. | Project manager: Prof. Dr.-Ing. Elmar Nöth
Project participants: Dipl.-Inf. Florian Hönig, Dr. phil. Anton Batliner
Keywords: ergonimics; human-machine-interfaces; speech processing; sleepiness research; acoustic signal processing
Duration: 1.1.2012 - 31.12.2013
Sponsored by: Deutsche Forschungsgemeinschaft
Mitwirkende Institutionen: Lehrstuhl für Experimentelle Wirtschaftspsychologie
Contact: Hönig, Florian E-Mail: florian.hoenig@fau.de
|