An evaluation method for risk of virus exposure by droplets with a focus on the sound environment Sohei Tsujimura 1 Graduate School of Science and Engineering, Ibaraki University 4-12-1 Nakanarusawa, Hitachi, Ibaraki 316-8511, Japan Kohyoh Kitani 2 Graduate School of Science and Engineering, Ibaraki University 4-12-1 Nakanarusawa, Hitachi, Ibaraki 316-8511, Japan Yanrong Li 3 Graduate School of Science and Engineering, Ibaraki University 4-12-1 Nakanarusawa, Hitachi, Ibaraki 316-8511, Japan

ABSTRACT The aim of this study was to propose a new method to evaluate the risk of viral infection by droplets from the aspect of the sound environment. To achieve this aim, we attempted to grasp quantitatively the relationship between speech levels and the amount of droplets when talking while eating at a restaurant. A psychoacoustic experiment was conducted with a speaker and a listener as subjects under a condition in which the pink noise level varied at 5-dB intervals from 55 dB to 70 dB. Based on the results, the effects of background noise level on speech level were examined and the relation- ship between speech level and the number of droplets was clarified. Although speech levels tended to increase with increases in the background noise level within the range of 60–70 dB, no signifi- cant difference was found in the number of droplets with particle sizes  125  m. 1. INTRODUCTION

In January 2020, the first person in Japan infected with the novel severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) was identified. Even as of March 2022, the resultant coronavirus dis- ease 2019 (COVID-19) pandemic continues to unfold. The main infection controls for COVID-19 are organized by infection route [1]. Hand washing and disinfection by alcohol have been shown to be effective for preventing contact transmission, while active indoor ventilation such as opening windows is recommended for airborne infection. For droplet infection, wearing a mask and social distancing are mentioned as infection control measures. As a countermeasure against COVID-19 droplet and airborne infection, it is important to reduce exposure to droplets and droplet nuclei. Re- garding droplet infection, it has been recommended to maintain social distancing of at least 2 m from others because the spread of droplets generated during utterance is about 1–2 m [2]. For air-

1 sohei.tsujimura.fifty@vc.ibaraki.ac.jp

2 kitkouad@gmail.com

3 yanrong.li.mech@vc.ibaraki.ac.jp

worm 2022

borne infection, the Ministry of Health, Labour and Welfare of Japan has officially announced a ventilation method for improving “closed spaces with poor ventilation” in commercial facilities, etc. This method recommends meeting a ventilation requirement of 30 m 3 fresh air volume/h per person and setting the air change rate to ≥2 times/h. However, these are not surefire infection pre- vention measures. In addition, “Avoiding poorly ventilated closed spaces”, “Reducing the density of people”, and “Avoiding short-distance conversations, vocalizations, and singing” have been pro- posed by an expert conference of the Ministry of Health, Labour and Welfare in Japan as infection controls to prevent situations conducive to outbreaks of disease clusters.

Focusing on restaurants that serve alcohol, the food service industry has been greatly affected by the application of closings and shorter business hours as necessitated by emergency COVID-19 measures and the declaration of a national state of emergency. Infection controls such as “Wear a mask except during meals”, “Do not speak loudly”, and “Ensure sufficient space between seats and install partitions” recommended at restaurants. However, wearing a mask other than during a meal is difficult for patrons, and conversations at short distances are likely to occur. It has been reported that the droplet particles emitted during utterance change with voice amplitude, and that the amount of droplet particles emitted tends to increase when the voice amplitude is large [3]. Regarding the relationship between conversation conditions and speech levels, it has been reported that speech levels change depending on the distance from the conversation partner and are affected by the background noise in the room [4]. Aligning these findings, if speech levels rise because of increased background noise, the amount of droplets will increase, and consequently, the risk of viral infection. However, the risk of viral infection has seldom been discussed from the aspect of the sound envi- ronment. Therefore, in the present study, to clarify quantitatively the relationship between noise levels and the amount of droplets in an utterance, the infection probability was calculated based on the amount of virus in droplets. In addition, the risk of droplet infection in situations with relatively higher background noise levels, such as restaurants that serve alcohol, was investigated. 2. PSYCHOACOUSTIC EXPERIMENT

In the present study, to grasp quantitatively the relationship between indoor noise levels in a restau- rant during a conversation and the amount of droplets, we conducted a psychoacoustic experiment under sound environment conditions with relatively high background noise levels. The probability of infection was then calculated based on the amount of virus in droplets, and the relationship be- tween background noise levels and the risk of droplet infection was evaluated.

2.1. Experimental Procedure The study participants were 20 males in their 20s who had normal hearing that did not interfere with their daily lives. One speaker and one listener were assigned to each group, and 20 participant groups were created. Participant groups in this experiment were 10 pairs of males and 10 pairs of male speakers and female listeners, for a total of 20 pairs. The speakers and listeners were not ac- quainted with each other. The interpersonal distance was set to 2 m, and pink noise was used as background noise to establish four sound environment conditions at 5-dB intervals between 55–70 dB. The experimental conditions are shown in Table 1. The operating noise of the laser irradiation device (KANOMAX CW532-3MGF) used for the measurement of droplets was about 60 dB under the condition in which no pink noise was added, so the background noise level condition during the actual experiment is the value shown in parentheses in Table 1. The background noise level was adjusted at the speaker’s position. For the speech level, the A-weighted sound pressure level ( L Aeq,1min ) was calculated by excluding the background noise energy at the same point from the voice data recorded at 1 m in front of the speaker and 1.2 m in height. As the experiment was con- ducted for eight participants under the condition involving only the operating sound of the laser ir- radiation device without pink noise, only the result of this condition is shown as a reference value. The participants were classified into speakers and listeners, each of whom was seated in a chair and asked to imagine a scene at a restaurant involving a normal conversation. The speaker was asked to adjust the height of the table so that he could speak while assuming a natural posture. When speak

ing, the participants were instructed to do so without regard for the camera, to move the head as lit- tle as possible, to look at the listener’s face, and to remain conscious of making the other member of the pair understand the content of the conversation, which involved 21 types of sentences of about 450 characters (7–8 characters/s) created based on the contents of a book that explained various sit- uations of daily life using knowledge from psychological research. The speaker was instructed to comprehend fully the content of the text before speaking to the listener. After the speaking was completed, the speaker was asked to evaluate the loudness of his own voice using a 7-point rating scale, and the listener was asked to evaluate the impression of ease of hearing in the same way. The 7-point rating scale used in the experiment is shown in Figure 1.

Table 1: Experimental conditions.

Background noise Noise level

Number of participants in

L Aeq,1min

the experiment

Operating sound of the laser irradiation device (60.3 dB) 8

55 dB (62.1 dB)

60 dB (63.7 dB)

Pink noise (+operating sound of the laser irradiation device)

20

65 dB (66.6 dB)

70 dB (70.7 dB)

Close to the impression of A

Close to the impression of B

Extremely

Fairly

Slightly

Fairly

Extremely

Slightly

Neutral

1

2

3

4

7

6

5

A

B

(Speaker) Quiet

Loud (Speaker)

Easy to hear (Listener)

(Listener) Difficult to hear

Figure 1: The 7-point rating scale used in the present experiment.

2.2. Measurement Method for the Amount of Droplets To measure the amount of droplets, a laser beam was emitted at a position 0.6 m in front of the speaker using a laser irradiation device. The droplets were photographed within a range of 0.8 m  0.6 m (resolution: 640  480 pixels; 1 pixel: 125  m) centered on the speaker’s mouth using a high- speed camera (KATO KOKEN k3). It has been reported that the decrease in the diameter of drop- lets due to evaporation and the change in the size of droplet particles due to condensation in exhaled breath have almost no effect at a distance of about 0.6 m from the mouth [5]. The shooting time of the high-speed camera was set to about 35 s at a frame rate of 200 frames/s from the point where the speaker spoke 45 characters or more (approximately 6–7 s after the start of speech) of a sen- tence of about 450 characters.

3. RESULTS AND DISCUSSION

3.1. Relationship Between Indoor Noise Levels and Speech Levels The relationship between noise levels during utterance and speech levels is shown in Figure 2. As a result of polynomial approximation with speech level as the objective variable and noise level as the explanatory variable, Equation 1 was obtained:

𝑦= −0.024𝑥 2 + 3.81𝑥−79.4 (1)

where y is the speech level and x is the noise level. The coefficient of determination of this approx- imate expression is 0.995. From this equation, in the range of background noise from 60 to 65 dB, the utterance level increased by about 0.80 dB when the noise level increased by 1 dB, but in the background noise range of 65– 70 dB, the speech level tended to increase by about 0.54 dB as the background noise increased by 1 dB.

According to these results, although the speech level tended to increase with noise levels during increased utterances, the speech level only rose to about 72 dB. A previous study reported that when the noise level increases by 1 dB in the background noise range of 60–70 dB, the speech level rises by about 0.60 dB [4], which was similar to the results in the present study.

95% CI (max)

70

Speech level ( L Aeq,1min ) [dB]

Mean

95% CI (min)

67

64

61

58

r 2 =0.995

55

60

70

65

75

55

Noise level ( L Aeq,1min ) [dB]

Figure 2: Relationship between noise levels during utterance and speech levels.

3.2. Relationship Between Indoor Noise Levels and Psychological Evaluations To examine the effect of noise levels on the speaker’s phonation type, the relationship between noise levels and the impression of the loudness of vocalization was investigated. As shown in Fig- ure 3, the impression of the loudness of vocalization tended to rise as the noise level increased. The results of a simple linear regression analysis using the loudness evaluation value of the speaker’s vocalization as the objective variable and the noise level as the explanatory variable, Equation 2 was obtained:

𝑦= −0.23𝑥−10.3 (2)

According to Equation 2, the noise level evaluated higher than “4. Neutral” in terms of the loudness of vocalization is about 62.2 dB. In fact, when the noise level is 62.2 dB or higher, the speaker’s vocalization is self-evaluated as loud, which suggests that the speaker is conscious of speaking loudly.

From the above, it was considered that the speech level is likely to rise because the speaker con- sciously uses a loud voice when the noise level in the room is about 62–63 dB or higher. In this study, the speaker appeared to be conscious of spontaneously raising his voice at all four levels of

the sound environment conditions (62.1, 63.7, 66.6, and 70.7 dB when the operating noise of the laser irradiation device is added) in which pink noise (55–70 dB) was used as background noise.

7

95% CI (max)

Mean

Loudness of vocalization

6

95% CI (min)

Pink noise condition

5

Only the operating sound of the laser irradiation device (Reference)

4

3

2

1

60

62

64

72

58

66

68

70

Noise level ( L Aeq,1min ) [dB]

Figure 3: Relationship between noise levels and the impression of loudness of vocalization.

3.3. Relationship Between Indoor Noise Levels and Amount of Droplets In the measurement of the amount of droplets, a binarization method at the brightness values of 0– 255 of the images was adopted [6]. The relationship between the droplets reflected in the image and the brightness value of the background was examined, and the threshold of the brightness value was set to 130 based on the results. In the captured image, particles with a brightness value exceeding 130 were regarded as droplets and the number was measured. The particle size of the droplets was calculated from the number of pixels, where 1 pixel of the image was 125  m. In other words, the measurement targets of this experiment were droplets with a particle size  125  m.

Figure 3 shows the relationship between noises level and the number of droplets. The plots in the figure represent the mean values, and the error bars the 95% confidence intervals (CIs). Accord- ing to the results, no significant difference was observed in the number of droplets in the back- ground noise range from about 62 dB (experimental condition of pink noise 55 dB) to about 71 dB (experimental condition of pink noise 70 dB), and the number of droplet particles was about 8–9. A previous study [3] reported that the amount of droplets increases as the utterance level rises, which

20

95% CI (max)

Mean

95% CI (min)

16

Number of droplets

Pink noise condition

12

Only the operating sound of the laser irradiation device (Reference)

8

4

0

60

62

64

72

58

66

68

70

Noise level ( L Aeq,1min ) [dB]

Figure 4: Relationship between noise levels and the number of droplets.

is the result of the speaker’s consciousness of utterance changing from a small voice to a loud voice. Furthermore, the cultural background of the speaker also differed from that in the present study (Japanese). By contrast, within the noise level range in this study, the amount of droplets did not change significantly depending on the noise level. The reason for this could be that the speaker was conscious of using a loud voice in the noise level range used in this experiment. According to the results of Figure 4, the speaker consciously used a loud voice when the background noise was  62 dB. In contrast, the noise level of Japanese restaurants has been reported to be about 60–70 dB [7], so it was possible to show the tendency of the amount of droplets in this noise level range.

3.4. Calculation Method for Infection Probability In this study, we assumed that the droplets were perfectly spherical particles. When the virus parti- cles contained in 1 mL of the droplets are represented by  copies/m, the number of virus particles N contained in the droplets having a diameter of d  m are calculated using Equation (3):

3

𝑁= 4

3 𝜋 ൬ 𝑑

× 10 −12 × 𝜌 (3)

2 ൰

Although the  value varies greatly depending on the infected person, 10 7 copies/mL was used in this study with reference to the results of many polymerase chain reaction tests on virus-infected persons [8]. The infection probability P after inhalation of the number of virus particles N is ex- pressed by Equation (4) according to the Poisson distribution:

𝑃= 1 −𝑒𝑥𝑝 ൬ −𝑁

൰ (4)

𝑁 0

where N 0 is the number of virus particles that lead to infection. It has been reported that N 0 is in the range of 322–2012 copies [8]. These values were estimated based on the infection probability in cases where COVID-19 clusters actually occur (two cases from a Chinese tour bus and one case each from a Korean aerobics dance class, a large call center, and a choir in an American church). In addition, in the case of SARS-CoV-2, there is a report that N 0 is estimated to be hundreds to thou- sands of particles. Based on these findings, we analyzed N 0 as 300–2000 copies. These values were also used in the risk assessment of droplets and aerosol infection during events using the supercom- puter “FUGAKU” in a project of the Cabinet Secretariat [9]. In the present study, the risk of droplet infection was calculated considering the background noise level and conversation time using Equa- tions (3) and (4).

3.5. Droplet Infection Risk Due to Differences in Noise Levels Considering Conversation Time Figure 5 shows the results of the risk of droplet infection considering noises level and conversation time. From the result, it was found that the noise level does not significantly affect the infection risk in the noise level range of 60–70 dB, and that the risk of droplet infection tends to increase with in- creases in the conversation time. In this noise level range, even under the conditions where the risk is the lowest ( N 0 is 2000 copies), there is a risk of infection due to droplets of about 49.1%–55.0% in a 10-min conversation. Furthermore, it has been shown that the risk of infection is 86.8%–90.9% when the conversation time is 30 min, and 98.8% or higher when the conversation time exceeds 1 h. Under the condition that the risk of infection is highly evaluated ( N 0 is 300 copies), it was consid- ered that a risk of droplet infection of 98.9% or higher occurs when the conversation time exceeds 10 min.

These results indicate a high risk of viral infection due to droplets when the noise level is  60 dB in a restaurant, where conversations are likely to occur without a mask. The infection probabili- ties shown in Figure 5 are the results for droplets with a particle size of 125  m exhaled during conversation. In actual conversations, even larger size particle droplets (saliva) may occur. Some droplets with a particle size <125  m are also generated by conversation. Since the droplets with a

particle size <125  m were not measured in this experiment, it was presumed that the risk of viral infection is actually higher than suggested by our results.

35 s 10 min 30 min 60 min 90 min 120 min

Droplet infection probability [%]

Droplet infection probability [%]

100

100

80

80

60

60

40

40

20

20

0

0

60

62

64

72

66

68

70

60

62

64

72

66

68

70

Noise level ( L Aeq,1min ) [dB]

Noise level ( L Aeq,1min ) [dB]

b) N 0 is 300 copies

a) N 0 is 2000 copies

Figure 5: Risk of droplet infection considering noise levels and conversation time. 4. CONCLUSION

In this paper, we assumed a conversation in a restaurant that served alcohol with a relatively high indoor noise level to clarify the relationship between noise and speech levels. The probability of infection was calculated based on the amount of virus in the droplets generated during utterances, and the risk of droplet infection due to differences in noise levels was evaluated. As a result, the following findings were obtained: • When the indoor noise level is  62 dB, the speaker consciously tends to use a loud voice and the speech level tends to rise. • As the amount of droplets is substantial when the noise level is  60 dB, the risk of droplet in- fection is extremely high when talking without a mask. 5. REFERENCES

1. Minister of Health, Labour and Welfare. Basic policies for novel coronavirus disease control by

the government of Japan (summary), Japan, 1–12 (2020) 2. Kohanski, M. A., Lo, L. J. & Waring, M. S. Review of indoor aerosol generation, transport, and

control in the context of COVID-19, International Forum of Allergy & Rhinology , 10(10) , 1173–1179 (2020) 3. Asadi, S., Wexler, A. S., Cappa, C. D., Barreda, S., Bouvier, N. M. & Ristenpart, W. D. Aerosol

emission and superemission during human speech increase with voice loudness, Scientific Re- ports , 9(2348) , 1–10 (2019) 4. Tsujimura, S. Construction of indicators for sound environment on the risk of virus exposure by

airborne droplets –Effect of differences in interpersonal distance and background noise level on speech level–, Journal of the INCE of Japan , 46(1) , 36–43 (2022) 5. Chao, C. Y. H., Wan, M. P., Morawska, L., Johnson, G. R., Ristovski, Z. D., Hargreaves, M.,

Mengersen, K., Corbett, S., Li, Y., Xie, X. & Katoshevski, D. Characterization of expiration air jets and droplet size distributions immediately at the mouth opening, Journal of Aerosol Sci- ence , 40(2) , 122–133 (2009) 6. Yamamoto, F. & Uemura, T. Fundamentals and applications of PIV, Japanese Journal of Mul-

tiphase Flow , 6(1) , 65–79 (1992)

7. Sueoka, S., Uchida, H., Kikuchi, H., Kamoshida, H., Kadoya, M. & Tanaka, S. An investigation

of the new index of the noise level in Japan, Journal of Environmental Laboratories Associa- tion , 34(4) , 22–29 (2009) 8. Prentiss, M., Chu, A. & Berggren, K. K. Superspreading events without superspreaders: using

high attack rate events to estimate N 0 for airborne transmission of COVID-19. Preprints from medRxiv , Cold Spring Harbor Laboratory , 1–28 (2020) 9. Tsubokura, M. Risk assessment of droplet and aerosol transmission during events, COVID-19

AI & Simulation Project Powered by Cabinet Secretariat , Research Results Slides , 1–11 (2021)