A wristband developed by researchers at North Carolina State University can detect gestures made on the wearer’s forearm, including taps, swipes and two-finger movements such as those used to enlarge a map. It can also respond to different levels of finger pressure.
Called EpiPad, the system sends out millimetre waves and analyses changes in the reflected signals when fingers deform the skin. Machine-learning software uses those changes to work out where the fingers are and how they move. The radar sensor assembly measures about 5 by 4 millimetres; no sensors or other equipment need to be attached to the forearm itself.
In a study involving 20 participants, the system correctly identified 92.8% of gestures after users had become familiar with it and completed individual calibration. That process required participants to make prescribed movements so EpiPad could adapt to their arms and inputs. It took an average of three minutes and 54 seconds. For new users without individual calibration, gesture recognition fell to 56.7%.
The system distinguished pressure levels correctly in 90.2% of cases. Its average error in locating a finger was 10.2 millimetres, however, potentially making it difficult to select small controls.
The prototype is not yet integrated into a smartwatch: its wristband has a protruding circuit board, and real-time tests used a cable to send sensor data to a laptop for processing. Loose clothing over the forearm can interfere with recognition.
The researchers see longer-term possibilities for smartwatches and other wearables, including smart glasses without an input surface of their own. The electronics are estimated to cost less than US$10, a materials estimate rather than a prospective selling price.