A team of students from AGH UST have examined how face masks can affect the extent to which speech is comprehensible . At the Technical Acoustics Laboratory (Department of Mechanics and Vibroacoustics) they studied the impact of various face coverings on speech transmission index (STI). The findings show that covering one’s face does not create favourable conditions for speech comprehension. Both the results of the measurements taken as well as the opinions revealed by a survey conducted as part of the study showed that, among other things, face shields negatively impact the quality of message.
The current COVID-19 pandemic has made it compulsory to cover one’s nose and mouth and keep social distancing between interlocutors. Both these factors have a significant impact on speech intelligibility – sound propagates differently when the speaker’s mouth is covered by a face mask or a shield. At the same time, increasing the distance between interlocutors further impedes proper understanding of a spoken message.
The research conducted by Michaela Murzyniec, Emilia Puchała and Kinga Sapieja - members of the Architectural Acoustics Student Research Society affiliated with the degree programme in acoustic engineering - studied the effects of various types of face coverings based on objective and subjective data. The researchers tested the most commonly used face coverings: disposable face mask, double-layer cotton face mask, double-layer folded cotton face mask, surgical face mask, single-layer neck gaiter, mini and full face shield.
The most important parameter used to describe speech intelligibility is the speech transmission index (STI). In order to determine STI values a series of measurements was conducted in an echo chamber, where echo was reduced through the use of sound-absorbing materials. 50 dBA pink noise* was used as background. In order to take measurements the research team applied a specialised microphone and a sound source – artificial mouth, which generated signal that was filtered in such a way that its characteristics matched those of human speech.
Adam Pilch, DSc. – Project Supervisor – said: “Having analysed the results obtained we conclude that increasing social distance from 1.0m to 1.5m resulted in speech intelligibility dropping by approximately 10% in the acoustic conditions of the study. Further 10% is lost when you put on a double-layer or surgical face mask or a mini shield. The greatest loss of intelligibility was observed with a full face shield, which reduced the speech intelligibility parameter by 25% compared to the reference value.”
The authors of the study also emphasise the impact of microphone location on the study results. Speech produced by a person wearing a face mask will be better understood by an interlocutor facing the speaker directly while with a face shield better speech comprehensibility was obtained when the receiver was positioned at a 45O angle to the axis of the loudspeaker.
In addition, the research team conducted a subjective survey-based study, which involved recording statements taken from Corpory – the Polish Speech Corpus developed by Professor Stefan Grocholewski. The recordings were conducted in an anechoic chamber with the speaker standing 1.5 m from the microphone. The recorded samples were later used in a survey designed to compare two types of face coverings: a disposable face mask and a face shield as well as clean production with no face covering. Background noise was used to best imitate real-life conditions. The survey was conducted on a group of 61 respondents (5 of whom were non-native speakers of Polish with working knowledge of this language).
The subjects surveyed were asked to pick the most intelligible statement of all three. A vast majority of respondents chose the one that was uttered without any face covering. A handful of respondents reported the statement uttered with a face mask while the one uttered with a face shield was never mentioned as the most intelligible.
* pink noise – stationary signal with constant power per octave, frequently used in acoustic measurements
All rights reserved © 2021 AGH University of Science and Technology