Temporal decomposition of speech : compactness measures compared

Abstract

In articulatory phonetics speech production is described as a sequence of articulatory gestures. Each gesture specifies the movements of the positions of the tongue, the lips and other articulators characteristic of one phoneme. Adjacent gestures overlap one another, resulting in the characteristic transitions between phonemes that can be observed in almost any parametric representation of the acoustic speech signal. In some cases a target may not be reached before movement towards the next begins. It has long been assumed that such targets cannot be determined from the acoustic signal alone, detailed knowledge of the production of all component phonemes being required before the speech signal can be 'decoded".