Emotions – the present and future of speech technology


One of the challenges of human-computer interaction is how to make a computer recognize emotions from human speech and react adequately. The article provides a survey of the role of speech technology in affective computing, explains the concept of emotion in that field, introduces different approaches to emotion classification (categorical and dimensional, combined), and describes different emotion corpora (acted, elicited, spontaneous) as well as the methods of collecting emotional speech for each kind of corpus.

The paper highlights recent and ongoing work at the Institute of the Estonian Language in affective computing, in particular the creation and content of the Estonian Emotional Speech Corpus. A brief survey is given of the available research on the acoustics and perception of Estonian emotions as well as of the current trends in emotional speech synthesis and in emotion recognition from speech.