The acoustic properties of phonologically equal vowels show a large range of variability, so that automatic classification methods have achieved limited success in identifying vowels. The only secure basis for obtaining absolute vowel quality is to rely on judgements of a jury of phoneticians being well experienced in Daniel Jones's familiar Cardinal Vowel diagram. Because of the exclusiveness of this procedure a desire to replace the expert jury by automatic methods existed for a long time (Fant, 1978 ). In this paper an algorithm (WVT) is proposed for calculating absolute vowel quality from the speech signal. Since the results of this algorithm are equivalent to those of the jury and meaningful in phonetic terms, it can replace the jury. The algorithm can be applied to complex utterances and can display which vowel targets are really reached in the Cardinal Vowel diagram and how the trajectories run. So this tool can be used for research on diphthongs, undershoot/overshoot, coarticulation, assimilation and reduction, as some preliminary investigations show.
Bibliographic reference. Pfitzinger, Hartmut R. (1995): "Dynamic vowel quality: a new determination formalism based on perceptual experiments", In EUROSPEECH-1995, 417-420.