ISCA Archive Interspeech 2024
ISCA Archive Interspeech 2024

Acoustical analysis of the initial phones in speech-laugh

Ryo Setoguchi, Yoshiko Arimoto

To elucidate the mechanism underlying the occurrence of speech-laugh, the acoustical difference between an initial phone of speech-laugh and phones during nonlaughing speech was examined by conducting acoustic analyses using two features representing vocal tract and voice source characteristics. First, a two-way analysis of variance (ANOVA) for the first and second formant frequencies (F1 and F2) was performed based on the factors of speech type (speech-laugh vs. speech) and five vowels (/a/, /e/, /i/, /o/, /u/). Second, using vowels and consonants, generalized linear mixed models (GLMMs) were developed with speech type as the objective variable and 12th-order mel-cepstral coefficients as explanatory variables. The ANOVA results revealed that the F1 values of the vowels /a/ and /o/ were greater at the beginning of the speech-laugh than during speech. The GLMM analysis showed that lower-order coefficients contributed to differentiating speech-laugh and speech.