Pitch Behavior Detection for Automatic Prominence Recognition

Giovanni Abete, Francesco Cutugno, Bogdan Ludusan and Antonio Origlia, University of Naples

In this paper a non-supervised approach for automatic syllable prominence recognition is presented. Previous research indicates syllable nuclei energy and duration as the main cues to detect prominence. Fundamental frequency has also been investigated in the past but considered secondary or irrelevant for this task. The proposed system uses the energy and the duration of the nucleus while taking into account also the pitch behavior. The algorithm was tested by comparing its results with the annotations of two human experts and we found that this approach has a 5.6\% accuracy increase with respect to the system not using the pitch behavior.