Face masks are an important tool for slowing the transmission of respiratory infections, but they can make it more difficult to understand speech, especially for listeners with hearing loss. Face masks muffle high-frequency sounds that are critical for speech understanding and block visual cues that are especially important for people with hearing loss. As schools and businesses resume in-person operation, many are requiring teachers, students, workers, and customers to wear face coverings. While there is no perfect solution, talkers can make listening easier by choosing the right mask and using amplification technologies.
We can now choose from a variety of commercial and homemade masks, including single-use medical masks and respirators, washable cloth masks in different shapes and fabrics, and masks and shields made with transparent plastic. Earlier this year, our team performed acoustic measurements to compare the effects of different masks on speech signals.1 We used both a custom-built speech simulator and a human talker to measure acoustic attenuation for a distant listener. We also used a turntable to measure the directional effects of the masks and tested wearable microphones in different positions on and near the face of the human talker. Here, we summarize the study's key findings and present new results from a study using a larger set of cloth masks.
Most prior studies on face masks and speech have focused on medical masks, such as disposable surgical masks and N95 respirators. Medical masks were found to have little impact on speech intelligibility in audio-only listening experiments.2,3 Surgical masks have especially mild acoustic effects.4,5 However, opaque masks also block visual cues, which can make it more difficult for people to understand speech.6 Transparent masks may help to improve intelligibility for some listeners with hearing loss.7 Clear face coverings such as face shields and cloth masks with windows have become popular with teachers,5, 8, 9 but clear plastic can attenuate sound more strongly than the materials used in opaque masks.
Washable cloth masks have been widely used during the pandemic and are being distributed by many schools and universities, but their acoustic effects have not been as widely studied as those of medical masks. We tested over a dozen commercial and homemade cloth masks to study the impact of material, weave, and the number of layers on acoustic performance (watch study video online: https://bit.ly/35gTU7s). The results show a wide variation in acoustic performance between fabrics, suggesting that teachers and others who spend time talking while wearing a mask should be careful when choosing a cloth mask.
ACOUSTIC ATTENUATION OF FACE MASKS
To measure the acoustic effects of different masks, we performed two experiments in an acoustically treated laboratory. First, we placed each mask on a custom-built head-shaped loudspeaker designed to simulate the directional radiation pattern of human speech. The loudspeaker produced frequency sweeps that were captured by a microphone placed about two meters away. The sweeps were used to compute acoustic transfer functions, which characterize the effects of the masks on sounds at different frequencies. Second, to measure performance in more realistic conditions, we captured speech from a live human talker while wearing each mask. Although the subject tried to read the passage consistently between recordings, there is inherently greater variation in the human experiment than in the loudspeaker experiment.
In both experiments, the masks had little effect on sound below 1 kHz, modest attenuation between 1 and 4 kHz, and strong attenuation above 4 kHz. The two experiments differed in the magnitude of the measured attenuation, but they produced a similar ranking of the masks. Figure 1 shows the average high-frequency attenuation of sixteen masks from the loudspeaker experiment alongside frequency responses for four masks. More detailed data are available in our paper.1 The polypropylene surgical mask and KN95 respirator had excellent acoustic performance, while the N95 respirator caused slightly more high-frequency attenuation. These results are consistent with previous studies on medical masks.4,5,8
The performance of cloth masks varies depending on the fabric and the number of layers. Loosely woven fabrics, such as plain and jersey, have the least effect on sound. Densely woven fabrics, such as denim and twill, block more sound. To understand the effects of fabric choice and number of layers on speech transmission, we recently tested four cotton flannel masks that are identical in shape but made of either two or four layers of light or heavy fabric. These masks are grouped on the right in Figure 1. Although masks with more layers block more sound, the four-layer mask of loosely woven flannel performed better than the two-layer mask of tightly woven flannel. Similarly, a three-layer cotton/spandex mask caused less attenuation than a two-layer cotton/spandex mask with a tighter weave. While the number of layers does affect sound transmission, the weave of the fabric appears to be the most important factor for acoustic performance.
Clear window masks and shields allow listeners to see lip movements and facial expressions. Unfortunately, they perform poorly acoustically, attenuating high frequencies by as much as 13 dB. The face shield caused especially strong distortion, even at low frequencies. It appears that face masks can provide either high-frequency sound cues or visual cues, but not both. Fortunately, our results on the directional effects of face masks suggest that wearable microphones can be used with nearly any mask, including those with clear windows.
DIRECTIONAL EFFECTS & AMPLIFICATION STRATEGIES
The long-distance measurements presented above only partly explain the acoustic effects of face masks. To understand how masks redirect speech sounds, we used a rotating loudspeaker to measure attenuation in all directions. The results suggest that masks most strongly attenuate sound directly in front of the talker but have less effect on the sides.1 The most extreme effect was from the face shield, which amplified sound behind the talker. Notably, all masks had a relatively weak effect above and below the mouth, suggesting that amplification technologies such as remote microphones could help to improve the audibility of talkers wearing face masks.
We used the human talker to test amplification strategies with lavalier microphones placed on and near the face. Figure 2 shows the effect of a cloth mask and window mask on the recorded speech spectrum at a distant microphone and at a lapel microphone worn by the talker. While high frequencies were attenuated for the distant microphone, the speech spectrum at the lapel microphone was mostly unaffected. Many lecture halls already feature wireless lapel microphones for their public address systems and many classroom assistive listening systems use remote microphones worn near the chest. Our experiments suggest that these microphones can be used successfully with face masks, including clear window masks. Indeed, a recent study of normal-hearing adults suggests that remote microphones can improve intelligibility with at least some clear masks.9
WHICH MASK IS BEST?
The choice of face mask depends on acoustic performance, visual transparency, and, of course, effectiveness against virus transmission. The best choices overall may be single-use polypropylene surgical masks and KN95 respirators, which provide excellent acoustic performance and are designed to filter small particles. Among reusable masks, there appears to be a tradeoff: Studies of droplet transmission suggest that loosely woven, breathable fabrics provide poor protection,10, 11 but our data show that breathable masks perform best acoustically. Multiple layers of light fabric may offer a good compromise between acoustic performance and virus protection. One mechanistic study of homemade cloth masks showed that three layers of t-shirt fabric provide as much protection as medical masks,10 and a similar mask in our study caused only 4 dB of high-frequency attenuation.
For listeners who rely on visual cues for communication, masks with clear windows may be the only viable option. However, these masks strongly attenuate high-frequency sound cues and may be challenging for listeners who rely more on sound cues. Talkers who use clear masks—or any masks—should consider using a sound reinforcement or assistive listening system to improve audibility and reduce vocal fatigue. Communicating during a pandemic is already stressful enough; we should not have to worry about being heard.
Acknowledgments: Several of the cloth masks used in this study were sewn by Ms. Catherine Somers from the Department of Electrical and Computer Engineering at the University of Illinois at Urbana-Champaign.
Funding: This research was supported by an appointment to the Intelligence Community Postdoctoral Research Fellowship Program at the University of Illinois at Urbana-Champaign, administered by Oak Ridge Institute for Science and Education through an interagency agreement between the U.S. Department of Energy and the Office of the Director of National Intelligence.