The literature offers at least two methods to annotators for characterizing the pitch range of a prosodic phrase. One method is in terms of the distance between the F0 maximum of the phrase (HiF0) and the speaker's utterance-final pitch (LoF0). The other method is in terms of the distance between pitch peaks and pitch valleys in the prosodic phrase. In this paper we address two questions. The first question concerns the reliability of the different methods. Five experienced phoneticians applied both methods on a set of forty utterances taken from read aloud text. We found that reliability was higher for HiF0 than for distances between pitch peaks and valleys. The second question is whether variation that is not captured by the first approach does actually occur in pitch contours. The results suggests that the HiF0 approach captures all the variation relevant to measuring pitch range that occurs in our small corpus. We conclude that the HiF0 method is methodologically more adequate, and at the same time sufficiently powerful to represent pitch range variation adequately for read aloud text.
|Title of host publication
|Proceedings of the 7th European Conference on Speech Communication and Technology
|P. Dalsgaard, B. Lindberg, H. Benner, Z.-H. Tan
|Number of pages
|Published - 2001