The AVLaughterCycle database has been recorded in the framework of the eNTERFACE'09 AVLaughterCycle project. Its objective was to build an
audiovisual laughter database, combining audio and accurate facial motion tracking. It is the first laughter database combining these 2 signals.
To obtain a robust facial motion tracking, it was decided to use facial motion tracking techniques using markers placed on the subject's face.
In consequence, subjects knew they were being recorded. An induction method was used to elicit spontaneous laughter.
It consisted in a funny video lasting around 10 minutes. The subjects' reactions while watching the video were recorded.
At the end of the video, the subject was instructed to perform one acted laughter.
The database contains about 1000 laughs of variable shapes (length, acoustic content, etc.).
It also contains other sounds (speech, breath, ...) but in lower quantities.
For the moment, audio recordings as well as the first round of annotation for each file can be downloaded from this page. Video recordings can be obtained upon request. To get a DVD copy, please contact Jérôme Urbain
The data can only be used for reserch purposes. Data may not be redistributed nor publicly broadcasted.
The database can be useful for a broad range of research topics: audio/video laughter recognition, modeling or synthesis;
the synchronization between the 2 modalities; analyses of the differences between spontaneous and acted laughter (even if the number of acted
laughter is much lower than the numpber of spontaneous ones); studies about the succession of laughter during the video; mesuring
diffrences between individuas; etc.
More information can be obtained from the AVLaughterCycle Database README.
The database can be downloaded from this link.