Acoustic studies of voice variation

Abstract

The present study deals with variation in the human voice. The functions of the voice are addressed and the major practical and theoretical approaches to the voice introduced. A three-fold approach is taken involving phonetics, logopedics and phoniatrics, and singing pedagogy. Voice research is described from three perspectives: voice production, acoustics and perception. A central concept is the adduction-abduction continuum, to which various descriptive categories of voice (e.9. creaky and breathy) are related. The study aims at describing the spectral properties of dysphonic voices, of voices varying in vocal intensify, and of supported and covered singing voices. The study also aims at relating the spectral observations to voice production, especially to the adduction-abduction continuum. In addition, the instrumental method used, long-term average spectrum (LTAS) analysis, is evaluated. Four sets of materials were investigated by means of LTAS analysis. Two were spoken materials: dysphonic voices (n = 87) and voices varying in vocal intensity (n = 10); two were sung materials: supported and unsupported voices (n = 8) and coverered and open voices (n - 1). The results indicate that the method used can differentiate between various voices. Several significant differences were observed in dysphonic voices, e.g. voices of patients with paralysis of vocal folds compared to those with laryngeal cancer. The spectra of loud and soft voices differed systematically. There were spectral differences in supported vs. unsupported voices, even though there was much individual and gender-related variation. The LTA spectra of covered and open voices differed systematically. Female and male voices also differed in spectral slope. Even though direct comparison of voices from the four data sets was not possible, the groups which showed distinct spectral differences appeared to form two clusters, "powerless" and "powerful", characterized by steep and shallow spectral slopes, respectively. The study showed LTAS analysis, despite its limitations, to be a useful means of investigation into the human voice and its variation.