Skip to main content Skip to main navigation


A New Emotion Database: Considerations, Sources and Scope

Marc Schröder; E. Douglas-Cowie; R. Cowie
In: Proceedings of the ISCA Workshop on Speech and Emotion: A Conceptual Framework for Research. ISCA Workshop on Speech and Emotion: A Conceptual Framework for Research, Pages 39-44, Textflow, 2000.


Research on the expression of emotion is underpinned by databases. Reviewing available resources persuaded us of the need to develop one that prioritised ecological validity. The basic unit of the database is a clip, which is an audiovisual recording of an episode that appears to be reasonably selfcontained. Clips range from 10 -- 60 secs, and are captured as MPEG files. They were drawn from two main sources. People were recorded discussing emotive subjects either with each other, or with one of the research team. We also recorded extracts from television programs where members of the public interact in a way that at least appears essentially spontaneous. Associated with each clip are two additional types of file. An audio file (.wav format) contains speech alone, edited to remove sounds other than the main speaker. An interpretation file describes the emotional state that observers attribute to the main speaker, using the FEELTRACE system to provide a continuous record of the perceived ebb and flow of emotion. Clips have been extracted for 100 speakers, with at least two for each speaker (one relatively neutral and others showing marked emotions of different kinds).