Acoustic-prosodic recognition of emotion in speech

Analysis of emotion in speech is manifested by the analysis of the vocal behavior of the nonverbal aspect of the speech. The basic assumption is that there is a set of objectively measurable voice parameters called prosodic aspects of speech, which can be assessed through computerized acoustical analysis. In this paper, we report results on recognizing emotional states (happy, sad, angry) from a corpus of short duration utterances using 18 acoustic-prosodic features. Four experiments were being conducted to perform correlations of the different prosodic-acoustic features and were evaluated against three different classifiers using a 10-fold cross-validation technique to estimate how accurately the classifier performs.