ISCA Archive SpeechProsody 2022
ISCA Archive SpeechProsody 2022

Computational modeling of intonation patterns in Arabic emotional speech

Aitor Arronte Alvarez, Elsayed Issa, Mohammed Alshakhori

The expression of emotion in speech communication has been frequently studied from the analysis of F0 contours. Global features such as the mean level and range of F0 as well as the slope of the contour, have been related to the degree of activation of an emotion. However, the existence of specific patterns associated with basic human emotions has not been empirically demonstrated, and studies generally find no conclusive answer to this question. In this paper we present a computational study of emotional speech in Arabic. A computational method for obtaining tonal contours based on F0 peak-to-peak and valley-to-valley distances is presented that is able to capture tonal rhythm (macro-rhythm). We introduce a model for extracting frequent tonal patterns to characterize emotions based on the typology of the patterns. Comparative analysis between sets of neutral and emotional utterances shows that distinctive patterns exist for anger, sadness, happiness, and surprise, that are completely absent in neutral speech. Findings highlight the effectiveness of the comparative computational methodology presented to discover macro-rhythmic emotion-specific patterns.

doi: 10.21437/SpeechProsody.2022-125

Cite as: Alvarez, A.A., Issa, E., Alshakhori, M. (2022) Computational modeling of intonation patterns in Arabic emotional speech. Proc. Speech Prosody 2022, 615-619, doi: 10.21437/SpeechProsody.2022-125

  author={Aitor Arronte Alvarez and Elsayed Issa and Mohammed Alshakhori},
  title={{Computational modeling of intonation patterns in Arabic emotional speech}},
  booktitle={Proc. Speech Prosody 2022},