Cars that can recognize when drivers are angry or irritated, and warn when emotional states might make them dangerous on the road, are in development in France, using dashboard cameras to track facial expressions associated with roadrage. The technology – which could, researchers suggest, be paired with lip reading AIs that could pick up on times when you cuss out the driver who cuts in front of you – initially reacts to expressions of anger or disgust.
The tracking research at the EPFL’s Signal Processing 5 Laboratory (LTS5) is being carried out in association with PSA Peugeot Citroën, which provided a car for real-world testing.
An infrared camera behind the steering wheel examines the movements of the face, and compares the driver’s expression to what the system has been “taught” about anger and disgust expressions. Exactly what would happen next hasn’t been defined: the car could perhaps simply warn the driver to take a break or flag up that their emotions might make them a danger, or, in more extreme circumstances, autonomous systems could take over some of the driving responsibilities.
However, it’s not perfect quite yet. Part of the problem is in the accuracy of the recognition system, which can be confused if the driver doesn’t express their feelings in a way that conforms to what the computer has previously learned.
The French team plans to double up on the cameras used, with two HD video feeds of the driver running through real-time GPU accelerated processing. There’ll also be face-detection and support for nighttime use, with IR illumination. Meanwhile, a self-teaching database of expressions that could learn different ways of portraying frustration could also be implemented.
There are also plans to use the cameras for other purposes, such as lip-tracking. That could better educate the safety system, watching out for keywords or expressions demonstrating anger, but also be used to improve speech-recognition and hands-free calls by isolating the driver’s speech and cutting out background noise.
It’s not the first use of camera-vision based emotion tracking we’ve seen. Last year, the Fraunhofer Institute demonstrated its own system, designed for use in store windows among other things, which could identify hundreds of different people in frame and estimate their mood, age, and other details.