Characterising respiratory aerosol emissions from speech and therapy activities using Wideband Integrated Bioaerosol Sensor (WIBS-NEO)

Introduction

Respiratory aerosols are a major vector for the transmission of respiratory diseases such as COVID-19. Phonation and speech are known sources of respirable aerosol in humans. Previous studies have shown that intensified vocal activities can produce aerosol concentrations exceeding those from conversational speech by more than a factor of 10, and those from quiet breathing by up to a factor of 30.^1,2 The number and mass concentrations of aerosols emitted during breathing, speaking, and singing, as well as their dependence on vocal loudness, are now relatively well characterised.^3,4 However, a clear gap remains in time-resolved and single-particle measurements of respiratory aerosol composition, and in the application of instruments widely used in atmospheric bioaerosol research^5,6 to clinical and voice-related settings. Addressing this gap is critical for improving our mechanistic understanding of respiratory aerosol generation and for informing safer clinical practice.

Method

The WIBS-NEO was deployed in a zero-background clinical setting, allowing aerosols to be directly attributed to specific vocalisations. 14 healthy participants performed a range of speech and voice activities, including humming (/m:/), sustained phonation (/a:/), fricatives (/ʒ/ pulses), projection (“Hey!”), and tongue trills, with breathing and speaking as reference measurements. The WIBS-NEO measured aerosol size (optical diameter, µm), shape (asymmetry factor, AF), number concentration (cm⁻³), and fluorescence intensity, while an Aerodynamic Particle Sizer (APS; TSI) was deployed concurrently to validate size distributions and concentrations.

Results

Several key findings emerge from this study. Figure 1 shows box-and-whisker plot of the total particle and fluorescent particle number concentrations measured by the WIBS-NEO across the different activities. Based on the mean values (rather than max/min), aerosol emissions increase in the following order: breathing (~0.1 particles/cm³), speaking, fricatives, projection, phonation, humming, and tongue trills (~0.5 particles/cm³), spanning about five orders of magnitudes across activities. Total particle number concentrations measured by the WIBS-NEO are comparable in magnitude to those obtained using the APS.

Fluorescent particles contribute approximately half of the total particle number concentration for most activities, indicating that 50% of emitted aerosols exhibit detectable fluorescence (60% for fricatives). This suggests that a substantial fraction of respiratory aerosols carry proteinaceous and other fluorescent compounds derived from human respiratory fluid.

Single-particle fluorescence analysis further shows that, among fluorescent particles, type A particles dominate (>90%), followed by type AB particles (~8%). This distribution indicates that respiratory aerosol fluorescence is primarily associated with fluorophores in the tryptophan- and albumin-dominated regions, with additional contributions from flavins (e.g., riboflavin). These findings are consistent with complementary bulk fluorescence spectroscopic measurements of human respiratory fluid samples.

Conclusion

This study demonstrates the suitability of the WIBS-NEO for characterising respiratory aerosols generated during human vocal activities in a clinical environment. Voice-related tasks produce elevated aerosol emissions relative to quiet breathing, with a substantial fraction exhibiting protein-associated fluorescence consistent with respiratory fluid. These fluorescent aerosols may serve as carriers of airborne pathogens and as potential markers for their detection. The WIBS-NEO’s ability to deliver time-resolved, single-particle fluorescence measurements supports its use for identifying higher-risk vocal tasks and informing evidence-based mitigation strategies in clinical practice.