In this project, we present a novel gesture input technique for off-the-shelf smartwatches. Our technique is designed to detect 10 hand gestures on the hand wearing a smartwatch. The technique is enabled by analyzing features of the Photoplethysmography (PPG) signals that optical heart-rate sensors capture. In a study with 20 participants, we show that our system achieves 90.55% accuracy and 90.73% recall. Without requiring bespoke hardware, our system can be readily used on existing smartwatches.
This work appears in UbiComp 2018.