Differential specificity of acoustic measures to listener perception of voice quality

Abstract

The purpose of this project was to differentially examine the specificity of two acoustic measures, relative fundamental frequency (RFF) and the cepstral/spectral index of dysphonia (CSID), to listener perceptions of voice quality across four dimensions: breathiness, roughness, strain/vocal effort, and overall severity. An auditory perceptual experiment was conducted to estimate listener perception of said dimensions. The Pearson's correlation coefficient between RFF, CSID, and the perceptual ratings of voice quality was calculated in order to comment on the relationship between calculations of RFF and CSID and the current "gold standard" of listener perception. The hypothesis for this project was that measures of RFF would have a strong negative correlation with listener perception of strain/vocal effort, and that measures of CSID would have a strong positive correlation with listener perception of overall severity and breathiness. An unexpected result with a significant impact was found to be that listeners' ratings of the four voice qualities were highly correlated with one another. Unfortunately, the poorly differentiated perceptual ratings significantly impact the validity of this project in addition to hindering any reliability of its results. Thus overall, the correlations between measures of RFF, CSID, and distinct qualities of listener perception are rendered uninterpretable. Methodological considerations and future directions are henceforth reported.