Voice Quality: Functions, Analysis and Synthesis
August 27-29, 2003
This paper explores amplitude-based measures of the glottal flow signal, as it has been suggested that they may provide a more robust method for analysing voice quality than the more typical time-based ones. Centred around an analysis of some samples from the VOQUAL’03 Speech database 2 of Japanese spontaneous speech, this paper (i) presents a discussion of whether a single global source parameter, such as the effective pulse declination time (which can be calculated from two amplitudes) or the T0-normalised equivalent, is likely to capture the tense-lax dimension of voice quality differentiation, across tokens exhibiting large f0 differences; (ii) presents formulations for an extended set of amplitude-based source parameters, to provide a richer coverage of voice source variation, and (iii) compares results for the amplitude vs. time-based measures. Results for (i) indicate that the non-normalised measure AQ is more effective in capturing tense–lax differences across a large f0 range than the normalised NAQ. Results for (iii) show a promising correspondence between the proposed amplitude-based measures and the time-based ones. A brief description is given of our planned research on Irish prosody, for which these measures will be important.
Bibliographic reference. Gobl, Christer / Ní Chasaide, Ailbhe (2003): "Amplitude-based source parameters for measuring voice quality ", In VOQUAL'03, 151-156.