Investigating the acoustic fidelity of remote recording methods

Our study tests the acoustic fidelity of remote recordings, using a large variety of stimuli and recording environments. With standard recording environments not available due to COVID-19, more studies investigate remote recordings for acoustic analyses [e.g., Guan and Li (2021); Freeman and De Decker (2021)]. High fidelity remote recordings also support crucial uses like reaching isolated populations and more speakers. A 188-word list was constructed from each English consonant followed by each vowel. Words recorded by a male and female speaker in a sound attenuated booth were input for test recordings. Stimuli were recorded on six devices across five operating systems, four tele-conferencing platforms, and three browsers, using internal and external microphones. Acoustic analysis investigates the impact of these recording configurations on features including pitch, relative intensity, vowel formant measures, spectral moments, spectral tilt, spectral rolloff, and Mel Frequency Cepstral Coefficients. Preliminary analyses found durational differences between original and test-recorded stimuli, posing challenges for automatic segmentation and alignment. Aperiodic noise was also introduced. We hypothesize further distortions in other measures. The findings from this study will allow us to identify acoustic measures which are robust across varied remote recording conditions and to highlight configurations least likely to introduce problematic artifacts.