Acoustic voice assessment reliability in varying reverberation times

Objective: Acoustic voice and speech assessment is a non-invasive and cost-effective tool for the clinic and the laboratory. However, the effect of room environment can impact the voice and speech metrics. This study investigates the threshold levels of simulated reverberation times above which acoustic measures become unreliable for voice quality assessment. Methods: Fifteen male and female subjects were recorded producing the sustained /a:/ vowel three times in a sound booth. Using Audacity software, the recordings were mixed with various simulated reverberation levels (T30). Acoustic measures related to voice quality assessment—such as shimmer, jitter, Cepstral Peak Prominence (CPP), and Harmonics-to-Noise Ratio (HNR)—were estimated for both original and reverberation-affected recordings. Statistical analyses were performed to determine significant differences in these measurements across the different reverberation levels. Results/Conclusions: The analysis demonstrated that CPP and jitter were more robust against high reverberation levels, showing high T30 threshold values, while shimmer and HNR were more vulnerable to reverberant rooms with low T30 threshold values. The outcomes recommend specific reverberation thresholds in clinics and recording environments for accurate acoustic measurements. Identifying these reverberation limits aids in optimizing recording environments and establishing standard room conditions for reliable acoustic voice assessments.