Feasibility of Wearable PPG for Simultaneous Hand Gesture and Force Level Classification

Simultaneous decoding of hand gestures and force levels has the potential to unlock a broad range of home-based application scenarios in post-stroke rehabilitation. However, current wearable approaches cannot struggle to balance wearability and functionality well. We thus propose and investigate the feasibility of wearable photoplethysmography (PPG) for simultaneous hand gesture and force level classification. A customized wristband embedded with green, red, and infrared light PPG sensors was developed. Fourteen subjects wore this wristband and performed four pinching gestures at 10%, 40%, and 70% maximum volunteer contraction (MVC) force levels. The results showed that the proposed method achieved satisfactory recognition accuracy, and the combination of three reflected light PPG signals was more accurate than any single signal or two-signal combinations. Force level classification accuracy within each hand gesture was less robust to the variety of hand gestures, and hand gesture classification accuracy at different force levels decreased as the force level increased. In addition, the proposed method still achieved reasonable recognition accuracy, with a relatively small amount of training data, potentially saving time in practical applications. These results demonstrate the feasibility of a wearable, low-cost PPG approach for simultaneous hand gesture and force level classification, which could facilitate more widespread hand mobility and strength training both inside and outside the clinic.