The Speed-Vel Project: a Corpus of Acoustic and Aerodynamic Data to Measure Droplets Emission During Speech Interaction

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 1991–1999 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 1991 The Speed-Vel Project: a Corpus of Acoustic and Aerodynamic Data to Measure Droplets Emission During Speech Interaction Carbone1 F., Bouchet2 G., Ghio1 A., Legou1 T., André1 C., Kadri1 S., Lalain1 M., Petrone1 C., Procino3 F., Giovanni1,4 A. (1) Aix Marseille Univ, CNRS, LPL, Aix-en-Provence, France (2) Aix Marseille Univ, CNRS, IUSTI, Marseille, France (3) Università degli Studi di Napoli Federico II, Naples, Italy (4) Aix Marseille Univ, ENT-HNS Department, La Conception University Hospital, Marseille, France LPL, 5 Avenue Pasteur, 13100 Aix-en-Provence, France {francesca.carbone, gilles.bouchet, alain.ghio, thierry.legou}@univ-amu.fr Abstract Conversations (normal speech) or professional interactions (e.g., projected speech in the classroom) have been identified as situations with increased risk of exposure to SARS-CoV-2 due to the high production of droplets in the exhaled air. However, it is still unclear to what extent speech properties influence droplets emission during everyday life conversations. Here, we report the experimental protocol of three experiments aiming at measuring the velocity and the direction of the airflow, the number and size of droplets spread during speech interactions in French. We consider different phonetic conditions, potentially leading to a modulation of speech droplets production, such as voice intensity (normal vs. loud voice), articulation manner of phonemes (type of consonants and vowels) and prosody (i.e., the melody of the speech). Findings from these experiments will allow future simulation studies to predict the transport, dispersion and evaporation of droplets emitted under different speech conditions. Keywords: droplets emission, speech production, aerodynamic data 1. Introduction Research on droplets emission during speech production has only recently gained attention after the emergence of the pandemic disease caused by SARS-CoV-2. The few lines of studies dealing with this topic have suggested that voice intensity (normal vs. loud speech), articulation manner of phonemes (i.e., type of vowels and consonants) and voicing (voiced vs. voiceless phonemes) are significant factors influencing the emission rate of the droplets. Concerning intensity, a common result is that loudness is positively correlated with the emission rate of droplets (Asadi et al., 2019; Anfirud et al., 2020; Stadnytskyi et al. 2020). For instance, Asadi and co-workers (2019) have shown that the number of droplets produced during loud speech is higher than during "calm" speech or calm breathing from 1 to 50 droplets/s (which corresponds to a rate of 0.06 to 3 particles per cm3), regardless of the language used (English or Spanish). Voice loudness, however, seems to not determine an increase of the airflow initial velocity which is a key factor influencing droplets trajectory (Bourouiba, 2020; Giovanni et al., 2020). In a preliminary work, Giovanni and co-workers (2020) have analysed the initial velocity of the exhaled air during vocal exercises through a propylene glycol cloud produced by 2 e-cigarettes’ users. Their first results showed relatively slow flow rates and unlikely to entail an increased risk of direct propagation of particles emitted in a loud voice (50 …

Content truncated to 500 words