Assessing how face masks affect speech intelligibility

Abstract

To limit transmission of Covid-19, the Centers for Disease Control recommends covering the mouth and nose with a mask while interacting in public. Although face masks protect public health by containing respiratory aerosols and droplets, they can hinder speech perception by filtering the acoustic signal. Here, we assess how masks affect speech intelligibility in audio-only conditions. Four young adult native speakers of American English, naïve to the purpose of the experiment, produced short low-probability sentences in four conditions: dual-layer cloth, disposable surgical, N95, and no mask. Multitalker babble was applied to the stimuli. Listeners recruited through Amazon Mechanical Turk rated the intelligibility of the sentences, with speaker and mask condition counterbalanced. Along with obtaining overall intelligibility ratings as a percentage of words recognized, we will also compute intelligibility for consonants that are known to be differentiated by high-frequency spectral cues, and assess the long-term average spectra for the different mask types. We hope that the information from this study will help us understand the impact of masks on speech perception. Going forward, we will carry out parallel experiments with older individuals and those with hearing loss.