抄録
Measuring laugh as smile or laughter can lead to understand the phenomenon of laugh, evaluate the quality and quantity of laugh, and can be used as a cue for human-robot interaction. We focus on laughter and develop laughter detection. In this paper, we analyzed laughter from various kinds of laughter's voices. The result of analysis showed that the fundamental frequency of laughter is higher than that in daily conversation. Moreover, we founded that the pattern of laughter which accounts for more than 90% of the total is a series of /ha/ or /hu/ like hahahaha. Intensity of these laughter has characteristics of periodic peak. The algorithm which used this feature acquired 82% precision rate and 73% recall rate. In addition, we developed laughing robot with a person and considered an application of the algorithm.
