Psychometric functions for rough voice quality.

Abstract

Dysphonic voices arising from laryngeal lesions are generally believed to vary across three perceptual dimensions—breathiness, roughness, and strain. This experiment sought to evaluate the psychometric functions for roughness and to determine the acoustic cues for its perception. Ten normal-hearing listeners, all native speakers of American English, were recruited for this experiment. The perception of roughness in dysphonic voices was evaluated for 34 voices from the Kay Elemetrics Disordered Voice Database (Kay-Pentax, Inc., Lincoln, NJ). Listeners compared the roughness of these test stimuli against that for a sawtooth wave with a 40-Hz squared-cosine amplitude modulation (AM). The modulation depth of the sawtooth wave was varied from low to high in a paired comparison task. Psychometric functions for roughness were derived by plotting the listener response for each stimulus pair against the modulation depth of AM sawtooth wave. The data for each stimulus were fitted with a logistic function, and various parameters (threshold, slope, and saturation point) were computed. These parameters were then compared to various acoustic measures obtained from each voice to determine candidate acoustic cues for roughness.