Accuracy of traditional and formant acoustic measurements in the evaluation of vocal quality

Abstract

Purpose: Investigate the accuracy of isolated and combined acoustic measurements in the discrimination of voice deviation intensity (GD) and predominant voice quality (PVQ) in patients with dysphonia. Methods: A total of 302 female patients with voice complaints participated in the study. The sustained /ε/ vowel was used to extract the following acoustic measures: mean and standard deviation (SD) of fundamental frequency (F0), jitter, shimmer, glottal to noise excitation (GNE) ratio and the mean of the first three formants (F1, F2, and F3). Auditory-perceptual evaluation of GD and PVQ was conducted by three speech-language pathologists who were voice specialists. Results: In isolation, only GNE provided satisfactory performance when discriminating between GD and PVQ. Improvement in the classification of GD and PVQ was observed when the acoustic measures were combined. Mean F0, F2, and GNE (healthy × mild-to-moderate deviation), the SDs of F0, F1, and F3 (mild-to-moderate × moderate deviation), and mean jitter and GNE (moderate × intense deviation) were the best combinations for discriminating GD. The best combinations for discriminating PVQ were mean F0, shimmer, and GNE (healthy × rough), F3 and GNE (healthy × breathy), mean F0, F3, and GNE (rough × tense), and mean F0, F1, and GNE (breathy × tense). Conclusion: In isolation, GNE proved to be the only acoustic parameter capable of discriminating between GG and PVQ. There was a gain in classification performance for discrimination of both GD and PVQ when traditional and formant acoustic measurements were combined.