Towards Objective Measures for Comparing Speaking Styles

This paper reports results from the quantitative comparison of readings of isolated sentences as well as spontaneous narratives by four female speakers, two each from different dialects. A multi-dimensional analysis of prosodic parameters was performed employing Fujisaki model based parameterizations of F0 contours, as well as a model of the perceptual local speech rate in order to capture the durational characteristics. Analysis shows that inter-speaker differences are vast with respect to speaking style regarding intonation as well as timing. In read utterances subjects generally accent more frequently and assign less prominence.