Virtual reality is becoming more of a thing, now that we have high quality headsets and the computing power to generate attractive environments. Many VR systems use controllers held in mid air, or camera-based systems that track limbs and hands for interaction. However, productivity scenarios often require prolonged interactions over a long period of time, which typically necessitates working at surfaces that allow the body to rest intermittently. To help facilitate this, a group of researchers at ETH Zurich developed TapID, including a preprint paper (PDF) that will be presented at IEEE VR 2021 later this month.

TapID consists of a wristband that carries two motion sensors, with one worn on each wrist. This allows TapID to detect taps from each of the user’s fingers individually, thanks to a machine learning algorithm that analyses the unique vibrations through your skeletal system. This is demonstrated as being useful for VR environments, where the user can type into a virtual keyboard, or interact with virtual objects on a surface, using their fingers as they would in the real world. This is a sensor fusion with the features of modern VR headsets that include hand tracking. The TapID wristbands deliver granularity and detection of small motions that is not nearly as accurate through headset-mounted senors and camera-based detection.

Test hardware includes 4 accelerometers. Two on flexible PCBs are the sensing hardware used by the system, the other two on the rigid PCB are used as a baseline during testing but do not contribute to the tap detection.

We’re not entirely convinced of the utility of sitting down in a virtual environment to type at a fake keyboard when monitors and real keyboards are more tactile and cheaper. However, having a device that can accurately determine individual finger interactions is sure to have applications in VR. And whether or not the demonstrated use cases are viable, the technology does indeed work.

It’s exciting to see the wrist-band form factor. It brings to mind the possibility of improving tap interactions in smart watches for non-VR uses. We envision chorded keyboard type gestures that detect which fingers are tapping but don’t need positional accuracy.

Those experimenting in VR interfaces may find it useful to reverse engineer what’s already out on the market, as we’ve featured before. Or, you can simply build your own! Video after the break.