Perceptual differences among models of the voice source: Further evidence

Models of the voice source differ in how they fit natural voices, but it is still unclear which differences in fit are perceptually salient. This study describes ongoing analyses of differences in the fit of six voice source models to 40 natural voices, and how these differences relate to perceptual similarities among stimuli. Listeners completed a visual sort-and-rate task to compare versions of each voice created with the different source models, and the results were analyzed using multi-dimensional scaling (MDS). Perceptual spaces were interpreted in terms of variations in model fit in both the time and spectral domains. The discussion will focus on the perceptual importance of matches to both time-domain and spectral features of the voice. [Work supported by NIH/NIDCD grant DC01797 and NSF grant IIS-1018863.]