アナウンサーが発声した「怒り」の韻律的特徴と聴覚的印象の主成分分析を用いた比較

概要

論文の詳細を見る
This paper describes PCA-based analysis of the prosodic features of "anger" expressions in Japanese speech. Aural impressions of speech were compared with the physical parameters of the prosodic features. We used 240 speech samples uttered by four announcers for a listening test. The speech samples consisted of "neutral" speech as well as speech with three types of emotions ("anger," "joy," and "sadness") of three degrees ("light," "middle", and "high"). Prosodic-feature parameters were speech rate and F0 parameters. Analysis results were summarized as follows: (1) The differences of features depending on the degrees of emotions which the speaker intended to express and those which the subjects identified as were found in the variations of eigenvalues of the second principal component for the maximum fundamental frequency and the magnitude of accent command, (2) the differences of features depending on the degrees of emotions which the subjects identified as were found in the variations of the same eigenvalues, conspicuously in those for the magnitude of accent command, and (3) the subjects tended to identify the emotions as either of two groups: "neutral" or "anger" regarding any degrees of anger as in one group.