Neural-based, Camera-Agnostic Robust Driver ID Recognizes Drivers with Mask On
Bridging the gap between technology and user is no easy feat. Smartphones have seemingly achieved it by reacting to users’ wishes – recognizing them in the blink of an eye either to let them unlock the system or to accept the request of a payment. Soon, it will be cars everywhere reacting to drivers, enabling never-before-seen in-cabin experiences. The key to unlocking that driving experience is already here: DTS AutoSense Robust Driver ID.
As the name indicates, a Driver ID can turn a routine trip out of the city into a unique experience by personalizing in-cabin perks to fit the person driving.
A Robust Driver ID, however, can deliver on that promise in almost all the conditions you can think of.
In no time, it became clear that one of those conditions would be the use of face masks while driving. As a preventive measure against COVID-19, authorities imposed the use of face masks around the clock, with ride-sharing services insisting upon it even after the restrictions were lifted.
Wearing the mask on and off was in no way a viable solution, so Robust Driver ID was trained to accurately enroll and recognize drivers with and without face masks on.
A single neural network was trained in various, custom scenarios, as face masks differ in shape, color pattern and texture. By combining real masks with artificially generated ones, and tweaking the software accordingly, we ensured our neural network covered all possible scenarios.
The decision to go with one single neural network instead of two, as is the standard in the industry, may seem surprising but, in fact, it brings about a series of benefits. By leveraging a single network, this solution improves processing speed, saves time, and puts less strain on the internal memory as data is saved locally.
Every registration and authentication request is carried off in real time and without putting the driver’s data at risk. Security is ensured by taking care of the data processing locally, not in cloud.
Besides being a real-time, neural, secure solution, Robust Driver ID also works with every camera sensor on the market, helping OEMs eliminate one more bump in the road towards custom in-cabin experiences.
Performance-wise, our solution behaves impeccably both with drivers whose features have been obstructed by masks as well as those not wearing face masks. We’ve seen similar FAR (False Acceptance Rate) and FRR (False Recognition Rate) in both cases. This means our solution recognizes the driver correctly, rejecting other persons who sit behind the wheel, with or without mask.
As part of DTS AutoSense, Robust Driver ID can become the key that unlocks countless in-cabin experiences, each different and tailored to the specific driver and his needs, in the safest, easiest, most seamless way imaginable.
Gabriel Costache has developed biometric solutions for the last 18 years, starting with speaker recognition for his master’s degree in 2003 and continuing with face and person recognition based on imaging data for his PhD degree in 2004. He joined XPERI in 2006 where he is currently a Snr Director leading the Biometric R&D group. His focus is to develop innovative biometric technologies for the end user’s convenience while at the same time ensuring user privacy.