Temporal variability of fundamental frequency contours
Temporal variability of fundamental frequency contours

Summary/Abstract: Intonation is one of the means of performing a speech style. Thus, observing pitch variation in an utterance may be a clue to identifying speech style. We design a cumulative slope (CS) index based upon the amount of pitch variation in a measured F0 contour and the duration of that contour. The more pitch changes there are and the greater their frequency range is, the greater the CS index is. This is confirmed by an experiment we conduct: the CS index of utterances with expressive intonation is higher than that of utterances with neutral intonation, and for utterances with neutral intonation the CS index is higher than for utterances with monotonous or flat intonation. However, as there is a great variability between speakers, the CS index as defined currently, cannot be used to universally differentiate between the styles. Results obtained using automatic voice activity detection (VAD) are close to those obtained with manual VAD and thus the extraction of CS index can be reliably automatized.

• Details
• Contents