Perceptual sensitivity to a model of the source spectrum

Abstract

A psychoacoustic model of the source spectrum has been proposed in which four spectral slope parameters describe perception of overall voice quality: H1-H2 (the difference in amplitude between the first and second harmonics), H2-H4, H4-2000 Hz (i.e., the harmonic nearest 2000 Hz), and 2000–5000 Hz. The goals of this study are to evaluate perceptual sensitivity in the mid-to-high frequency range of the model and determine how sensitivity to one parameter varies as a function of another. To determine listener sensitivity to slope changes for each parameter, just-noticeable differences were obtained for series of stimuli based on synthetic copies of one male and one female voice. Twenty listeners completed an adaptive up-down paradigm. To provide a baseline of listener sensitivity to each spectral slope parameter, the synthetic voices were manipulated so that spectral slope varied by 0.5 dB increments for each parameter while other parameters remained constant. We then assessed how listener sensitivity to a given harmonic slope parameter changes when the others covary. These results will help assess the validity of the model and determine what sources of cross-voice variability in spectral configuration are perceptible.