14 DAY TRIAL // Investigators with the Swiss Federal Institute of Technology (EPFL) in Lausanne, working with colleagues at PSA Peugeot Citroën, are currently developing a new type of sensor that will be able to assess the emotional states of car drivers in real time, potentially contributing to improving overall driving safety. The proof-of-concept study is still relatively far away from reality, the team says.
At this point, EPFL experts say, embedded cameras are the best way to move forward with this project. These instruments can be used to film drivers' faces, and then run the video feed through a series of specialized algorithms. The latter would assign each facial expression to one of seven universal emotions, including anger, sadness, disgust, suspicion, fear, joy and surprise.
This is not the first time that cameras able to identify human emotions are employed by technology companies. These have been in use for video game development, marketing, and medicine for many years, but the research group plans to apply the same principles in driving as well, to improve safety, and reduce the number of accidents on public roads.
The emotional state of a driver is the second-most important cause of accidents on the road, following fatigue. One of the feelings that lowers attentiveness and increases driver aggression is irritation, which many are unable to control. Most often, irritation is caused by other drivers that do not adhere to traffic laws and rules, making it difficult for others to drive properly.
The research group has until now developed only a small prototype of the new system, but early tests indicate that the innovation could have very promising applications. EPFL scientists say that it is very difficult to develop a method of accurately identifying emotions that is non-invasive, and small enough to be used unobtrusively in a space as small as a car.
The study was led by experts with the EPFL Signal Processing 5 Laboratory (LTS5), who specialize in detecting, monitoring and analyzing facial expressions. Scientists at PSA Peugeot Citroën suggested that the detector come in the shape of an infrared camera to be mounted behind the steering wheel.
At first, researchers led by Hua Gao and Anil Yüce decided to focus their efforts on getting the camera to understand just two facial expressions, anger and disgust. The algorithms first had to learn how to identify the two emotions, by being exposed to images of people displaying them. The next step was to make sure that they were able to identify these expressions on their own.
One of the main challenges facing the project, and one that is unlikely to go away any time soon, is the wide variability that exists in the way humans express their emotions. Regardless of how well adapted detection algorithms are, there will always be natural variability that will not be accounted for.
In the experiments, the EPFL group had the camera work in tandem with a fatigue detector, an instrument capable of measuring the percentage and frequency of eyelid closures in drivers.