Speech features of sleepiness

Our voice is a tool that, with the right systems, can detect a wide range of pathologies and provide precise information on the speaker's state of health. Easily implemented in real-life conditions, affordable and robust to noisy conditions, voice is a highly attractive tool for tomorrow's digital medicine.

One of the most promising applications for the use of voice is home monitoring of patients suffering from neuropsychiatric disorders. Thanks to a smartphone application, it is now possible to collect data on a regular basis, in real-life conditions, and dynamically adapt the information collected.

Among the most widespread neuropsychiatric disorders in the general population, sleep disorders often result in excessive daytime sleepiness (EDS), which is both a major cause of road accidents and a risk factor for cardiovascular and psychiatric pathologies.

While it's normal to be sleepy at certain times of the day (before bedtime, for example), sleepiness becomes pathological when it becomes chronic and interferes with people's daily lives.

The mechanisms underlying the EDS complaint are often explored by medical specialists using an iterative sleep latency test (TILE), which requires two nights and a full day's hospitalization, costing the social security system €1,500 and mobilizing qualified personnel.

Replacing this medical test with a voice-based application therefore offers both financial and human benefits, while opening up new prospects for diagnosis and treatment.

This article presents the collaboration between researchers from the fields of artificial intelligence and neuropsychiatry to develop voice biomarkers for EDS, with the ultimate aim of implementing them in a virtual doctor.