Depth-Sensitive Optical Sensing for Non-Invasive Measurement of Human Muscle Activity.

Human muscle anatomy consists of multiple layers, each contributing to movement through complex patterns of activation. Conventional non-invasive sensing techniques, such as surface electromyography (sEMG) and mechanomyography (MMG), primarily capture aggregate muscle activity and provide limited depth-dependent information. As different movements may involve distinct combinations of superficial and deeper muscles, access to depth-dependent information could improve the discrimination of motion patterns that are difficult to distinguish using surface measurements alone. To address this limitation, we developed an optical sensor capable of depth-sensitive measurement using near-infrared light. The sensor comprises a light source and an array of photodetectors arranged at six source-detector distances (SDDs) ranging from 12 to 48 mm within a compact wearable module. Two experiments were conducted to evaluate the sensor. First, depth sensitivity was investigated using Monte Carlo simulations and phantom experiments, demonstrating distinct sensitivity profiles for different SDDs and providing preliminary evidence of depth-dependent sensing. Second, the sensor was attached to the forearm to measure signals during nine hand and wrist movements. Machine learning models were evaluated for motion classification, with Linear Discriminant Analysis (LDA) achieving the highest performance. Using all six SDD channels, an average classification accuracy of 87.5% was achieved across 10 subjects. An ablation study evaluating all 63 possible channel combinations further showed that classification performance improved systematically with the inclusion of multiple SDD channels, indicating that measurements obtained at different sensing depths provide complementary information for motion discrimination. These results demonstrate the feasibility of multi-SDD optical sensing for capturing depth-dependent physiological information and highlight its potential as a compact, non-invasive sensing approach for wearable human-machine interface applications.