Pragmatic temporal voice range profile as a tool in the research of speech styles
Antti Iivonen · 2001
Speaker’s pragmatic temporal voice range profile (TVRP) and F0 distribution histogram were used for comparing two radio speakers and two speaking styles in one speaker. It is assumed that the internal distribution of the temporal voice range, relative placement of the range on semitone scale and the concentration of F0 points in certain semitone classes of the F0 histograms are connected with the speech style and the factors bedhind the style. Placing a single utterance against the background of TVRP its features can be better understood within the prosodic repertoire of the speaker. 1. Voice range profile and pragmatic temporal voice range profile The voice range profiles are developed for voice research ([1]:153). They do not include the time domain. Speaker’s pragmatic temporal voice range profile (below TVRP) indicates the distribution of fundamental frequency contours on a time/frequency scale (Fig. 1). The distribution of the measured F0 points can also be displayed on a frequency histogram (with semitone classification from 19 to 58 ST; Fig. 2), but then the distribution on time scale will be lost. “Pragmatic” means that the F0 contours represent a selected group of speaker’s real utterances, read, imitated, acted or spontaneous. A larger corpus including different speech styles would be indicative of speaker’s total temporal voice range. Single utterances can be depicted againts the background of the TVRP (Fig. 3) [2]. 2. Phonetic or linguistic unit applied The shape of the temporal voice range profile depends on the utterance notion, because the zero point of TVRP represents the beginning of every utterance. Breathgroup or intonation unit (intonational phrase) are more natural phonetic alternatives. Sense-group, sentence, and clause would be more linguistic alternatives. In spontaneous conversation short elliptical utterances can occur among longer and complex turns. On computer screen, the pixel resolution of the screen sets practical limitations for the time domain. Here a time domain until about 5.5 sec was displayed (Fig. 1).