For drivers, there are times when it feels as if that fast-approaching emergency vehicle, with its siren blaring, has come from nowhere. On a packed multi-lane street or at a busy junction, its sudden appearance can confuse a driver into making the wrong move when trying to clear the way, making matters worse for the response vehicle as it tries to reach its destination.
In such cases, an earlier warning about the vehicle’s approach, including where it’s coming from, would allow the driver to make better and safer decisions.
The scenario is one of several that inspired researchers at the Fraunhofer Institute for Digital Media Technology (IDMT) in Oldenburg, Germany, to create an A.I.-based solution that allows cars to listen, and respond to, their surroundings.
The technology uses a microphone placed inside a modified roof fin. This links to a control unit containing software that powers the system. The software uses a library of different noises and computer algorithms that enable it to recognize the acoustic signature of specific sounds while discarding irrelevant background noise. The team used a wide range of archived noises to train the algorithms that power the platform.
Fraunhofer IDMT also designed its own so-called “beamforming algorithms” that allow the technology to dynamically locate moving sound sources such as a siren on an approaching emergency vehicle.
With today’s vehicles built to keep external noise out, and booming sound systems that ensure you hear nothing but the audio you’re listening to, the technology would offer an efficient way to alert drivers to important events nearby, including anything from approaching emergency vehicles to children playing in the street.
The researchers point out that the system could be made to be so sensitive that it would even be able to pick up the sound of a nail in a tire tapping on the road surface, enabling it to alert the driver to the potential danger. Other unusual noises from the vehicle could also be recognized, allowing the system to keep tabs on the car’s condition.
The researchers point out that the technology would also be useful for autonomous vehicles.
“Despite the huge potential of such applications, no autonomous vehicle has yet been equipped with a system capable of perceiving external noises,” Danilo Hollosi, head of the Acoustic Event Recognition group at Fraunhofer IDMT, said in a release.
The equipment is still under development but the team hopes to have something ready for commercialization within the next five years.
- Here’s what Bosch hopes to learn from deploying autonomous cars in San Jose
- Bosch’s A.I.-powered tech could prevent accidents by staring at you
- Good at StarCraft? DARPA wants to train military robots with your brain waves
- What Lyft and Aptiv learned from 100,000 self-driving car trips
- Electrical tape on speed limit signs tricks Tesla vehicles into violations