Measuring Affective and Social Signals in Vocal Interaction
Khiet P. Truong · 2010
In this paper, I will discuss how and what type of measurements of vocal interactional behavior can be used to recognize affective and social signals. Three studies will be presented that deal with 1) the collection and recognition of spontaneous vocal and facial expressions in a gaming context, 2) the detection of laughter in meetings, and 3) the relation between dominance and overlapping speech in multiparty conversations. On the basis of these studies, (dis)advantages and issues in speech processing for affective and social computing will be evaluated. Acoustic features, but also simple speech or no-speech information were employed in these studies. In addition, fundamental issues such as 'ground truth labeling' and collection of spontaneous data are also discussed.