|
|
Face, Speech, and Acoustics |
|
| Introduction · Program · Local Information · Registration · Contact |
Emotions are often distinguished into categories which may be "anger", "joy", "sadness", etc. Using emotion dimensions provides more or less abstract scales which span an emotion space, wherein even the classical emotion categories can find their place. The main part of the talk will deal with a perception experiment where subjects have to arrange stimuli on a scale which corresponds to one of these dimensions. The 500ms long stimuli where extracted from a corpus of read speech where emotional expression are included. The scale used in the task is labeled from "frown" to "smile", in words of emotion dimensions this means roughly from "negativ" to "positiv". First results show high correlation between the subjects. For acoustical analysis a method for calculating voice source parameters will be suggested. The LF-model has been implemented, so that these parameters can be obtained from the inverse filtered signal using an iterative estimation. Problems of this current work in progress will be raised up and establish a discussion for new acoustical measurements in emotions research.