Abstract Zero-field optically pumped magnetometers (OPMs) represent a promising modality in biomagnetism research due to their ultra-sensitive capabilities contained within compact sensor packages that enable contactless biomagnetic measurements. In this study, we utilized a zero-field OPM to simultaneously measure three-axis components of biomagnetic fields produced by muscle activity in the human forearm caused through independent finger movements, with the goal of recognizing multiple gestures using only a single magnetic sensor. We recorded muscle activity caused by the opening and closing of each finger 60 times, triggered by an auditory tone, and trained a neural network to classify the different finger movements. We utilized the triaxial OPM and trained neural network in a live-testing mode to replicate the participant’s hand gestures onto a robotic-hand proxy, with 1~ s delay, firstly for three fingers (little, middle, and thumb) and subsequently for all five fingers. Using a single OPM sensor, we captured, recognized, and replicated finger gestures from one participant in a controlled laboratory setting. In the three-gesture live test, all three finger movements were identified and replicated with ≥ 98% accuracy. In the five-gesture live test, four out of five fingers were identified and replicated with ≥ 90% accuracy, with the little finger being the most challenging to distinguish with a reduced 45% accuracy. These results demonstrate that a single zero-field OPM sensor in a triaxial measurement scheme can be effectively used for contactless recognition and replication of multiple finger movements, providing a proof-of-concept toward future wearable OPM-based magnetomyography systems.
Dawson et al. (Wed,) studied this question.