Analysis of prosodic correlates of emotional speech data

Abstract

The study of expressive speech styles remains an important topic as to their parameters detection or prediction in speech processing. In this paper, we analyze prosodic correlates for six emotion styles (anger, disgust, joy, fear, surprise and sadness), using data uttered by two speakers. The analysis is focused on the way pronunciations and prosodic parameters are modified in emotional speech, compared to neutral style. The analysis concerns speech pronunciation modifications, presence of pauses in sentences, and local prosodic behavior, with an emphasis set on the analysis of the prosody over prosodic groups and breathing groups.