Acoustic analysis and synthesis of laughter

Abstract

Laughter is an important emotional expression in speech. The production is characterized simply by quasiperiodic glottal opening/closing gesture with relatively constant articulatory configuration. The acoustic consequence, however, is complicated and appears in the temporal characteristics of voiced-unvoiced category and pitch, the spectral envelope, and the segmental duration. In this study, acoustic parameters that characterize laughter were investigated from acoustical and perceptual points of view. Laughter /hahaha/ was selected as a typical laughter from conversational speech recording and the acoustic characteristics were compared with those of normally uttered speech /hahaha/. The acoustic analysis was performed in terms of pitch in average and its intrasegmental pattern, the spectral tilt and formant bandwidth, and the segmental duration of the vowel /a/ and the consonant /h/. Then, perceptually significant acoustic features were investigated by using means of replacement of the acoustic parameters of the normally uttered speech by those of the laughter in speech analysis-synthesis system. The results showed that an increased pitch in average and distinct intrasegmental pitch pattern as well as spectral envelope with broad bandwidths are significant features to characterize the laughter. The laughter synthesis by rule was also investigated by manipulating these acoustic parameters.