A multimodal analysis of different types of laughter expression in conversational dialogues

Laughter is an important form of nonverbal communication and is manifested in various forms. In the present work, we analyzed an audio-visual conversation dataset and categorized laughter into several types according to its function. Four predominant categories were focused for multimodal analysis: mirthful, boosting, smoothing, and softening. We investigated the relationship between laughter types and expression modalities, including voice quality, facial expression, gaze, and body gestures. The results revealed that mirthful and boosting laughter produced with positive emotions or attitudes tend to be higher, longer and tenser voice quality, and accompanied by larger changes in facial expressions and body movements. We also conducted a careful analysis of gaze behaviors during laughter events, suggesting that boosting laughter tended to keep eye contact with both dialogue partners more than mirthful laughter.