Investigation of Prosodic Speech Features for Emotion and Stress Recognition

The way we say something is about as important as what we actually say. Intonation, speech rhythm, intensity and accentuation conduce to structuring and emphasis on all linguistic levels. Beyond, the meaning of what is said may be either underlined or even reversed (irony). Thus, prosody has an important role in transmitting information through speech, as it allows a conclusion to be drawn about the speaker’s emotional state and his or her attitude regarding the statement given.

This doctoral thesis is concerned with (psycho)acoustic correlates of prosodic features (including fundamental frequency, speech rhythm, loudness, and accentuation) and their eligibility to identify emotions, workload, and stress. The aim is to obtain quantified results which facilitate automated analysis.

Picking up a diploma thesis on robust speech features for stress classification (which lays the foundations for further research), this PhD thesis focuses on quantification of existing – qualitative – findings on emotional perception of prosodic features as well as the exploration of new ways to describe speech rhythm and accentuation patterns; the latter regarding pitch, intensity, and syllable length.