Analysis of prosodic features for end-of-utterance prediction in spontaneous Japanese

In this study, we analyzed prosodic features of accentual phrases and investigated their temporal changes to obtain cues for detecting boundaries at where turn-taking could occur in spontaneous conversations. The acoustic parameters used as prosodic features were the fundamental frequency, sound pressure level, and duration of accentual phrases in long utterance units. The results showed that the fundamental frequency shift between the first and second accentual phrases could be useful for detecting the number of accentual phrases in the long utterance unit. In addition, the results suggested that a rapid decrease in sound pressure and an extended duration of the accentual phrase constitute a cue for detecting the end of the utterance. That is, the acoustic predictor of the utterance length appeared at the beginning of the utterance, and the predictor of the utterance boundary appeared shortly before the end of the utterance.