The team led by Shahriar Nirjon at the University of Virginia in Charlottesville embedded a microphone into a set of headphones that listens to the throb of arteries in your ear.
That data, as well as activity levels gathered using an accelerometer, is sent over the internet to a recommendation engine which chooses the next song based on the user's current and desired heart rate.
It's not just for pumping people up, but the system can also be programmed to play music that calms a person by bringing their heart rate down.
As it is used, the system learns what types of music work for influencing a user's heart rate, based on the changes it sees after it recommends songs.
Environmental factors are also taken into account in the learning algorithm, meaning that songs that calm a person down when they are sitting still might be different to what works when they're just finished a run.
Their work will be presented next month at the SenSys conference in Toronto, Canada.
Source: ANI
