An artificial intelligence could tell if you are sleep deprived based on the rhythm and tone of your voice.
People may be sleep deprived without knowing it, which could cause negative physiological changes to their bodies, says at the Paris Sciences and Letters University, France. The AI could be an easy and inexpensive way for medics to objectively gauge if someone is missing out on sleep from their voice, he says.
“This is important because sleep deprivation is estimated to cost billions to the economy and health systems,” says Pressnitzer. “If you’re sleep deprived, you’re more likely to have accidents and chronic health issues.”
Advertisement
Pressnitzer and his colleagues recorded the voices of 22 women aged between 30 and 50 as they read the same sections of a French novel. This occurred after they were allowed to sleep as normal, for around 7 to 8 hours over a night, and after two consecutive nights where their sleep was restricted to a maximum of 3 hours per night. Transgender people weren’t included in the study.
The researchers then used 75 per cent of the recordings to train an AI on the links between sleep deprivation and the tone and rhythm of the women’s speech.
Next, they tested the trained AI on the remaining 25 per cent of the recording data. It could tell when the speakers were sleep deprived with 77 per cent accuracy, on average.
“When you’re sleep deprived, you get an increase in inflammation and we think inflammation of the throat and nose could cause the change in timbre [or tone],” says Pressnitzer. “On the other hand, changes in the rhythm of how people speak after sleep deprivation are likely due to cognitive fatigue.”
However, factors such as a respiratory infection, drinking alcohol and shouting excessively can also inflame the vocal tract, he says. “We need to do further studies to see if this approach can work when considering other factors that change the voice.”
Further work with a larger number of participants is also needed to establish how well the findings apply to men and when people are speaking in other languages, says Pressnitzer.
This study suggests that voice analysis could be developed as a “stethoscope” for measuring sleep deprivation, says at Loughborough University in the UK.
bioRxiv
