Fig 1.
(A) N95 respirator, (B) surgical mask, (C) pleated cloth mask, (D) fitted cloth mask.
Fig 2.
Average spectra of speech sounds produced while wearing each type of mask.
The y-axis indicates the logarithmic power spectral density of the sound. Compared with the no-mask condition, face masks generally attenuated higher frequencies. They also had a greater overall effect for Talker 2, compared with Talker 1.
Fig 3.
Difference in band energy between the no-mask condition and each of the four face mask conditions for sound frequencies in octave-scale bands centered at 125, 250, 500, 1000, 2000, 4000, and 8000 Hz.
The acoustic effects of the masks depended on both mask type and frequency. Generally, there were small differences compared to the no-mask condition at lower frequencies and a 5–25 dB decrease in intensity at higher frequencies.
Fig 4.
Mean percentage of words correctly recognized in the sentences as a function of mask type, signal-to-noise ratio (SNR), and talker.
Horizontal dashed lines represent the mean of the no-mask condition in each panel. Overall, listeners were much more accurate at the high SNR (+13 dB) than at the low SNR (+3 dB), and they were more accurate for Talker 1 than for Talker 2. At the high SNR, only the fitted cloth mask led to poorer performance compared with the no-mask condition. At the low SNR, both cloth masks and the N95 respirator led to lower accuracy. The pleated cloth mask also caused lower accuracy for Talker 2. Errors bars represent 95% confidence intervals.