A proposal of a model to extract Japanese voluntary speech rate control

To extract elements of prosodic features which relate to speakers' intentional control is required for speech information processing. Speech rate variation should be a "caution signal" to call listeners' attention strongly. To express and detect such "caution signals", we have proposed a new speech rate model. This model introduces two kinds of force to control the speech rate. One is a driving force which causes a global tendency of speech rate variation, and the other is damming force with speaker's voluntary control. The proposed model have been applied to several spontaneous conversational speech which we have newly recorded. A global tendency of speech rate has been expressed appropriately. In several cases, the intentional rate variations have been expressed appropriately.