Objective: Reliable acoustic classification of voice signals is critical for clinical assessment of voice disorders, yet traditional measures are limited by sensitivity to noise and dependence on signal periodicity. The nonlinear energy difference ratio (NEDR) and voice type component profile (VTCP) have shown strong accuracy in controlled conditions. To expand on earlier work limited to ideal recording environments, this study evaluated their robustness across phoneme type, vocal intensity, and recording environment to determine their clinical applicability.
Methods: Thirty-one normophonic adults produced sustained vowels (/a/, /i/, /u/) and fricatives (/s/, /z/) at five intensity levels (-10 to +10 dB relative to baseline). The vowel /a/ was additionally recorded in soundproof and noise-masked environments. Trimmed 0.75-second samples were analyzed in MATLAB to compute NEDR and VTCP values. Two-way repeated measures ANOVAs tested the effects of phoneme, intensity, and environment on each metric, with significance set at P < 0.05.
Results: NEDR remained stable across vowels and voiced fricatives, with significant deviations limited to very soft phonation (−10 dB, P < 0.05) and the voiceless fricative /s/ (P < 0.001). VTCP demonstrated greater sensitivity to intensity and environmental variation, particularly within Voice Types 3 and 4, which frequently misclassified normal phonation as disordered in soft or noisy conditions. Excluding /s/ reduced variability across both measures, confirming its disproportionate effect. Overall, NEDR exhibited stronger robustness than VTCP across all experimental factors.
Conclusions: NEDR reliably quantifies acoustic stability across variable phonation and environmental contexts, indicating its potential for clinical and remote assessment applications. VTCP’s higher sensitivity to loudness and phoneme type underscores the need for careful task control in clinical use. These findings support NEDR as a resilient, objective measure of vocal stability suitable for implementation beyond sound-treated settings.