Can we reliably predict sleepiness using voice recordings alone?

Embargoed until: Publicly released: 2024-02-07 14:23

International

Image by Adina Voicu from Pixabay

International researchers think they may have found a way to detect sleep deprivation through voice recordings. The team studied 22 healthy women aged 30-50, who were sleep-deprived in a lab setting – with no more than three hours of sleep – and asked them to read chapters of the same book for about 10 minutes while their voices were recorded. Participants were also asked about how sleepy they felt – which varied greatly between participants from night to night despite everyone having the same amount of sleep. The team used machine learning to analyse the voice recordings and found sleep deprivation could be detected using two different effects: changes in rhythm of speech, and changes in voice quality.

Attachments

Note: Not all attachments are visible to the general public. Research URLs will go live after the embargo ends.

Research PLOS, Web page The URL will go live after the embargo lifts.

Journal/
conference: PLOS Computational Biology

Research:Paper

Organisation/s: PSL University, France

Funder: Author ET was supported by grants ANR16-CONV-0002 (ILCB), ANR-11-LABX-0036 (BLRI) and the Excellence Initiative of Aix-Marseille University (A*MIDEX) (ET). Author DP was supported by grants ANR-22-CE28-0023-01, ANR19-CE28-0019-01, and ANR-17-EURE-0017. Author TA was supported by a Human Frontier Science Program Long-Term Fellowship (LT000362/2018-L). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Media Contact/s

Contact details are only visible to registered journalists.