Paolo Rossi Castelli 9 March 2023 16 min read

AI takes care of dysphonia

A lightweight device, attached to the sternum and connected to an artificial intelligence system capable of self-learning, can warn about the risks of using your voice too much.

It’s called vocal fatigue, and it’s the fatigue that typically affects not only professional singers, but also teachers, call centre operators, politicians, sports coaches and anyone who uses their voice intensively in their activities. It usually doesn’t have any serious consequences, just a need to rest the vocal cords for a few hours. However, in some cases it can lead to permanent damage, or changes in tone and timbre, and the chronic inflammation associated with it can lead to the formation of polyps.

 To curb these problems, researchers from the Bienen School of Music at Northwestern University in Evanston (USA), who have been working for years on the voice (compromised, for example, by a stroke or an accident or, more recently, by Covid and Long Covid) have developed a technologically advanced device that can help you realise when it’s time to stop using your voice before it’s too late.

AI helps voice problems  

Thanks to AI and the involvement of professional singers and bioengineers, American scientists have developed a small device that can be applied to the sternum with a simple patch, which records vibrations associated with the voice and sends the information to a series of algorithms in a tablet. When the voice goes over the limit, the device alerts the wearer via a vibrating hospital-like wristband.

The system analyses data using real-time machine learning techniques and separates the 'vocal dosimetry' associated with speech and singing from background noise. The device also collects a wide range of information on cardiac and respiratory activities, and overall physical exertion.

The whole thing has its own app, which also helps set customised limits and programmes, because an individual may have different limits, according to their anatomy and voice use.

It takes years to 'train' the AI programme

As researchers reported in the scientific journal PNAS, creating a handy, versatile device like this took years of study and 'training' sessions for the AI programme. This was carried out by recording the voices of professional singers, which gave the AI a sort of complete archive of the possible vibrations produced by the human voice, which are different when speaking and singing.

The whole range of sounds was then calibrated to establish average daily limits that shouldn’t be exceeded, and everything was transferred to a programme that analyses the vibrations produced, processes them, and transmits the signals via Bluetooth to the bracelet, as well as to a programme on a tablet and the app.

The latter can also allow a therapist to monitor the situation or, in the case of professional singers, a teacher to set up customised study programmes to avoid damaging the voice.

Under normal conditions, vocal fatigue is solved with about twenty minutes of silence. Sometimes, however, more needs to be done, and those who use their voice for work should avoid the need for forced rest. Northwestern's device could become a valuable aid for this.

New call-to-action  New call-to-action  Follow us on LinkedIn


Paolo Rossi Castelli

Journalist since 1983, Paolo has been dealing with scientific divulgation for years, especially in the fields of medicine and biology. He is the creator of Sportello Cancro, the site created by on oncology in collaboration with the Umberto Veronesi Foundation. He collaborated with the pages of the Science of Corriere della Sera for several years. He is the founder and director of PRC-Comunicare la scienza.