Development and Validation of a Single-Variable Comparison Stimulus for Matching Strained Voice Quality Using a Psychoacoustic Framework

Purpose: Acoustic and perceptual quantification of vocal strain has been a vexing problem for years. To increase measurement rigor, a suitable single-variable matching stimulus for strain was developed and validated, based on the matching stimulus used previously for breathy and rough voice qualities. Method: A set of 21 comparison stimuli for a single-variable matching task (SVMT) was synthesized based on a speech-shaped sawtooth waveform mixed with speech-shaped noise. Variable bandpass filter gain in mid-to-high frequencies achieved a wide range of computed sharpness (in constant sharpness steps) and served as the independent variable for the SVMT. Ten natural /[open back unrounded vowel]/ stimuli with a wide range of the primary voice quality of strain and a minimum of breathiness or roughness were selected and assessed using the SVMT. Natural voice samples and synthetic comparison stimuli were also assessed using a perceptual magnitude estimation (ME) task. Results: ME data validated the correspondence of the set of comparison stimuli to varying perceived strain. Perceived strain magnitudes of the comparison stimuli increased significantly and linearly with computed sharpness (r[superscript 2] = 0.99). A linear regression revealed that strain matching values were significantly predicted by computed sharpness (r[superscript 2] = 0.96) and perceived strain magnitudes (r[superscript 2] = 0.95) of the natural voice stimuli. Conclusion: The perception of vocal strain is strongly associated with computed sharpness and is captured accurately and precisely using an SVMT, in which the independent variable is the bandpass filter gain (in steps of equal sharpness) applied to the comparison stimuli.