Application of Pitch-Based Measures for Aperiodic Pediatric Dysphonic Voices

The role of pitch perception in the assessment of pediatric dysphonia remains underexplored. Aperiodicity associated with severe dysphonia poses a vexing problem for conventional measurement methods that rely on signal periodicity, rendering them unreliable for most dysphonic voices. We tested the hypotheses that pitch height and pitch strength systematically vary with type 1, 2, and 3 signals (varying in periodicity) associated with glottal and supraglottal vibratory sources and may be used to augment signal typing classification.

Ten naive listeners judged pitch height through a single-variable matching task and pitch strength through an anchored magnitude estimation task for 42 pediatric dysphonic sustained /a/ vowels (14 per signal type). Computational estimates of pitch height and pitch strength were obtained from the auditory sawtooth waveform inspired pitch estimator prime (Aud-SWIPE') algorithm. Analyses of variance were used to determine the differences in pitch height and pitch strength across the three signal types. The relationship between perceptual and computational estimates was examined using correlation coefficients and their significance.

Listeners were able to reliably judge pitch height even in type 3 signals, which lack clear periodicity. Aud-SWIPE' consistently generated reliable pitch height estimates, even for aperiodic type 3 signals. Pitch strength was significantly higher in type 1 signals than in type 2 or 3, both perceptually (P = 0.01) and computationally (P < 0.001).

Pitch height and pitch strength can be successfully estimated in pediatric dysphonic voices for all signal types. These measures can also be applied to natural speech, enabling assessment using real-world communication.