Acoustic effects of medical, cloth, and transparent face masks on speech signals
11 Acoustic effects of medical, cloth, and transparentface masks on speech signals
Ryan M. Corey, Uriah Jones, and Andrew C. SingerUniversity of Illinois at Urbana-Champaign
Abstract —Face masks muffle speech and make commu-nication more difficult, especially for people with hearingloss. This study examines the acoustic attenuation causedby different face masks, including medical, cloth, andtransparent masks, using a head-shaped loudspeaker anda live human talker. The results suggest that all masksattenuate frequencies above 1 kHz, that attenuation isgreatest in front of the talker, and that there is substantialvariation between mask types, especially cloth masks withdifferent materials and weaves. Transparent masks havepoor acoustic performance compared to both medicaland cloth masks. Most masks have little effect on lapelmicrophones, suggesting that existing sound reinforcementand assistive listening systems may be effective for verbalcommunication with masks.
I. I
NTRODUCTION
As the world works to control the novel coron-avirus 2019 (COVID-19) pandemic, face masks areexpected to prove critical to slowing the spread ofthe virus. However, it can be difficult to understandspeech when the talker is wearing a mask, espe-cially for listeners with hearing loss [1], [2]. Bystudying the acoustic effects of masks on speechsignals, we can determine which masks are bestfor speech transmission and evaluate technologiesto make communication easier.Most prior research on masked speech has fo-cused on medical equipment such as surgical masksand N95 respirators. A recent study on the acous-tics of medical masks showed that surgical masksand N95 respirators can attenuate higher-frequencysounds by between 3 and 12 dB [3]. Listeningtests using audio-only recordings made with medicalmasks have not shown significant effects on speechintelligibility [4]–[6].To conserve supplies of medical masks, healthauthorities have recommended cloth masks, whichcan be made from household materials or purchasedcommercially. Recent studies suggest that the ef-fectiveness of cloth masks at blocking respiratory
Fig. 1. Masks used in experiments and described in Table I droplets depends on the fabric material, weave,and thickness [7], [8]. Because both medical andcloth face masks obstruct visual cues that con-tribute to speech intelligibility [9], some hearingloss advocates recommend the use of transparentface coverings [2]. In listening tests with audiovisualrecordings of talkers wearing lapel microphones,masks with clear windows were shown to improveintelligibility for listeners with severe-to-profoundhearing loss compared to paper masks [10].To understand the effects of masks on speech, wemeasured the acoustic attenuation of a polypropy-lene surgical mask, N95 and KN95 respirators,six cloth masks made from different fabrics, twocloth masks with transparent windows, and a plasticshield, as shown in Figure 1. Measurements wereperformed using both a head-shaped loudspeakerand a live human talker. The experiments showthat different masks have different high-frequencyeffects and that they alter the directivity of speech.Finally, to examine the effects of masks on soundreinforcement and assistive listening systems, wetook measurements with microphones placed onthe lapel, cheek, forehead, and next to the mouth.These amplification technologies may prove criticalto verbal communication during the pandemic. a r X i v : . [ ee ss . A S ] A ug Fig. 2. Speech signals were produced by a human talker andloudspeaker model. Microphones were placed at listener distance andat several points on and near the face.
II. M
ETHODS
To simulate sound heard by a conversation part-ner, a side-address cardioid condenser microphonewas placed two meters from the talker position.To study the effect of masks on sound reinforce-ment and assistive listening systems, omnidirec-tional lavalier condenser microphones were placednext to the mouth (“headset” position), on the lapel,on the cheek, and on the forehead of the talker,as shown in Figure 2. The laboratory walls areacoustically treated with 8-inch melamine and 2-inch polyurethane foam wedges.Sound was produced by two sources. A custom-built head-shaped loudspeaker produced ten-secondlogarithmic frequency sweeps to measure acoustictransfer functions between the talker and listenerpositions. The plywood loudspeaker uses a 2-inchfull-range driver and has a directivity pattern that iscloser to that of a human talker compared to studiomonitors. To characterize the directional effects ofmasks, the loudspeaker was placed on a turntableand rotated in 15 degree increments while the “lis-tener” microphone remained fixed.For more realistic speech signals, 30-second read-speech recordings were made from a human talker,who attempted to use a consistent speech level foreach recording. Recordings of the human talkerwere repeated three times non-consecutively witheach mask. Human subject research was approvedby the University of Illinois Institutional ReviewBoard with protocol number 19503.For both the loudspeaker and human experiments,measurements were first taken with no face coveringto establish a baseline. The recordings were thenrepeated with the twelve face coverings listed inTable I and shown in Figure 1. III. R
ESULTS AND D ISCUSSION
A. Acoustic attenuation of face coverings
Figure 3 shows the effects of several masksmeasured at the listener position. The plots onthe left show the differences in acoustic transferfunctions measured with and without masks on thehead-shaped loudspeaker. The plots on the rightshow the corresponding results for the human talkeraveraged over three non-consecutive recordings; thehuman spectra varied by roughly 1 dB betweenrecordings. The attenuation values shown in Table Iare logarithmically weighted averages from 2 kHzto 16 kHz, that is, means of the points shown in theplots.Most masks had little effect below 1 kHz but theyattenuated higher frequencies by different amounts.The surgical mask (1) and KN95 respirator (2) hadpeak attenuation of around 4 dB, which is consistentwith the results reported by Goldin et al. [3] witha head-and-torso simulator. The N95 respirator (3)attenuated high frequencies by about 6 dB, whichis similar to the average attenuation reported byGoldin et al. [3].The cloth masks varied widely depending oncomposition and weave. The 100% cotton masks injersey (4) and plain (5) weaves had the best acousticperformance and were comparable to the surgicalmask. The cotton/spandex blends performed worse.Surprisingly, the 2-layer cotton/spandex mask (7)produced greater attenuation than the 3-layer cot-ton/spandex mask (6), perhaps because it has ahigher proportion of spandex and fit more snuglyon the face. Masks made from tightly woven denim(8) and bedsheets (9) performed worst acoustically.It appears that material and weave are the most im-portant variables determining the acoustic effects ofcloth face masks: More breathable fabrics transmitmore sound.Finally, the transparent masks (10–12) performedpoorly acoustically at high frequencies, blockingaround 8 dB for the human talker and 10–14 dBfor the loudspeaker. Although these masks are oftenrecommended to help listeners with hearing lossbecause they preserve visual cues, they also harmthe high-frequency sound cues that are crucial forspeech.
TABLE IM
ASK MEASUREMENTS AND K H Z ACOUSTIC ATTENUATION RESULTS
Material Layers Thickness(mm) Mass (g) Speaker atten. atlistener (dB) Human atten. atlistener (dB) Human atten. atlapel (dB)1 Polypropylene surgical 3 0.4 3 3.6 2.8 1.02 KN95 respirator (GB2626) 2 0.6 4 4.0 2.6 0.03 N95 respirator (3M 8210) 1 1.5 9 5.7 5.4 3.64 Cotton jersey 2 0.7 11 4.0 3.1 0.55 Cotton plain 2 0.5 11 4.0 4.3 1.46 Cotton/spandex jersey 3 1.5 16 6.1 5.2 2.37 Cotton/spandex jersey 2 0.9 17 8.2 6.1 2.08 Cotton plain & denim 2 1.1 21 9.4 10.0 3.29 Cotton percale bedsheet &polyester trim 2 1.0 14 12.6 9.5 3.110 Cloth & vinyl window 1 0.4 12 10.8 7.8 − .
11 Cloth & PVC window 1 0.3 7 12.5 8.0 0.412 Plastic shield 1 0.4 50 13.7 8.2 − .
250 500 1000 2000 4000 8000 16000 − −
250 500 1000 2000 4000 8000 16000 − −
250 500 1000 2000 4000 8000 16000 − − − −
250 500 1000 2000 4000 8000 16000 − − − − R e l a ti v e l e v e l ( d B ) Loudspeaker HumanFrequency (Hz) R e l a ti v e l e v e l ( d B ) Frequency (Hz)Surgical (1)KN95 (2)N95 (3)Window (10)Window (11)Shield (12) Surgical (1)KN95 (2)N95 (3)Window (10)Window (11)Shield (12)2L jersey (4)2L plain (5)3L blend (6)2L blend (7)2L denim (8)2L bedsheet (9) 2L jersey (4)2L plain (5)3L blend (6)2L blend (7)2L denim (8)2L bedsheet (9)
Fig. 3. Effect of different masks on sound levels measured at the listener position for a head-shaped loudspeaker (left) and human talker(right). − − − No maskCloth (6)Window (11)Shield (12)
Fig. 4. Spatial distribution of 2–16 kHz sound energy for a head-shaped loudspeaker with different masks, in dB relative to no maskat 0 degrees.
250 500 1000 2000 4000 8000 16000 − − Frequency (Hz) R e l a ti v e l e v e l ( d B ) ListenerLapelForeheadHeadsetCheek
Fig. 5. Effect of mask 11 on sound levels measured at differentmicrophones relative to the same measurements with no mask on ahuman talker.
B. Effect of face coverings on speech directivity
Figure 4 shows the relative high-frequency soundlevel as a function of angle for the head-shapedloudspeaker. The plot shows a logarithmicallyweighted average of relative sound level from 2 kHzto 16 kHz. For all masks tested, acoustic attenuationwas strongest in the front. Sound transmission tothe side of and behind the talker was less stronglyaffected by the masks, and the shield (12) amplifiedsound behind the talker. These results suggest thatmasks may deflect sound energy to the sides ratherthan absorbing it. Therefore, it may be possible touse microphones placed to the side of the mask forsound reinforcement.
C. Effect of microphone placement
Masks attenuate high-frequency sound for distantlisteners, but they have different effects on micro-phones on and near the face. Figure 5 shows theacoustic effects of the PVC window mask (11)
250 500 1000 2000 4000 8000 16000 − − Frequency (Hz) R e l a ti v e l e v e l ( d B ) N95 (3)Blend (7)Bedsheet (9)Window (11)Shield (12)
Fig. 6. Effect of several masks on sound levels at the lapelmicrophone on a human talker, relative to the same measurementswith no mask. on different microphones on a human talker. Thelistener and headset microphones experience similarhigh-frequency attenuation. The cheek microphonetaped under the mask recorded higher sound levels,but with spectral distortion. The lapel and foreheadmicrophones showed small and mostly uniformattenuation over the range of speech frequencies.Similar results were obtained for masks 1–10, al-though the performance of the cheek microphonevaried depending on the shape of the mask. Theshield (12) strongly distorted speech spectra for allmicrophones.Figure 6 compares the performance of severalmasks using a lapel microphone. Only the shieldhas a substantial effect on the speech spectra cap-tured by the microphone. Sound capture and rein-forcement systems used in classrooms and lecturehalls often rely on lapel microphones, and remotemicrophones that transmit to hearing aids are alsooften worn on the chest. These systems shouldwork with most masks with little modification. It isworth noting that lapel microphones were used forthe audiovisual recordings of [10], which showedintelligibility benefits with clear masks.IV. C
ONCLUSIONS
The experimental results presented here confirmthat face masks attenuate high-frequency sound infront of the talker, with the strongest attenuationabove 4 kHz. Ubiquitous polypropylene surgicalmasks offer the best acoustic performance amongall masks tested. If those masks are not avail-able, loosely woven 100% cotton masks performwell acoustically. Tightly woven cotton and cot-ton/spandex blends should be avoided if speech transmission is a concern. It is important to notethat this study did not consider the efficacy of masksat blocking respiratory droplets; it is possible thatloosely woven fabrics that perform well acousticallyare less effective against the virus and vice versa.Shields and masks with windows perform muchworse acoustically than opaque cloth masks. For-tunately, window masks do not strongly affect thelapel microphones used in sound reinforcement andassistive listening systems. To preserve visual cueswithout destroying high-frequency sound cues, talk-ers can wear clear window masks and lapel micro-phones. Although face masks make verbal commu-nication more difficult, amplification technologiescan help people with and without hearing loss tocommunicate more effectively during the pandemic.A
CKNOLWEDGMENTS
Mask 5 was sewn by Ms. Catherine Somersand mask 9 was sewn by Mr. Austin Lewis. Thisresearch was supported by an appointment to theIntelligence Community Postdoctoral Research Fel-lowship Program at the University of Illinois atUrbana-Champaign, administered by Oak Ridge In-stitute for Science and Education through an inter-agency agreement between the U.S. Department ofEnergy and the Office of the Director of NationalIntelligence. R
EFERENCES [1] J. Chodosh, B. E. Weinstein, and J. Blustein, “Face masks canbe devastating for people with hearing loss,”
BMJ
Hearing Review , vol. 27,no. 5, pp. 8–9, May 2020.[4] L. L. Mendel, J. A. Gardino, and S. R. Atcherson, “Speechunderstanding using surgical masks: A problem in health care?”
Journal of the American Academy of Audiology , vol. 19, no. 9,pp. 686–695, 2008.[5] F. Thomas, C. Allen, W. Butts, C. Rhoades, C. Brandon, andD. L. Handrahan, “Does wearing a surgical facemask or N95-respirator impair radio communication?”
Air Medical Journal ,vol. 30, no. 2, pp. 97–102, 2011.[6] A. J. Palmiero, D. Symons, J. W. Morgan III, and R. E. Shaffer,“Speech intelligibility assessment of protective facemasks andair-purifying respirators,”
Journal of Occupational and Envi-ronmental Hygiene , vol. 13, no. 12, pp. 960–968, 2016. [7] O. Aydin, M. A. B. Emon, and M. T. A. Saif, “Performanceof fabrics for home-made masks against spread of respiratoryinfection through droplets: A quantitative mechanistic study,” medRxiv , 2020.[8] A. Konda, A. Prakash, G. A. Moss, M. Schmoldt, G. D. Grant,and S. Guha, “Aerosol filtration efficiency of common fabricsused in respiratory cloth masks,”
ACS Nano , vol. 14, no. 5, pp.6339–6347, 2020.[9] C. Llamas, P. Harrison, D. Donnelly, and D. Watt, “Effectsof different types of face coverings on speech acoustics andintelligibility,”
York Papers in Linguistics Series 2 , no. 9, pp.80–104, 2008.[10] S. R. Atcherson, L. L. Mendel, W. J. Baltimore, C. Patro,S. Lee, M. Pousson, and M. J. Spann, “The effect of conven-tional and transparent surgical masks on speech understandingin individuals with and without hearing loss,”