Human-in-the-Loop Capacitive Microphone Sensors-Based Muscle Sensing System for Predictive and Adaptive Exoskeleton Assistance

Abstract

Mobility impairments among older adults and individuals with neuromuscular weakness motivate the need for timely and adaptive exoskeleton assistance. This paper presents a human-in-the-loop muscle sensing and control system based on capacitive microphone sensors (CMS) that capture subtle mechanical muscle vibrations preceding observable motion. CMS signals were shown to occur 20–30 ms earlier than IMU-based kinematics, enabling anticipatory intent detection and feedforward assistive control. A five-sensor CMS array positioned over major thigh muscles is combined with a two-stage control strategy that integrates threshold-based pre-assist triggering and machine-learning-based torque refinement. Experiments across walking, stair ascent, sitting, and standing achieved over 90% classification accuracy under both non-fatigued and fatigued conditions with low latency. Robustness evaluations demonstrate stable CMS performance under realistic wearable perturbations, including perspiration and attachment variation. Extended experimental sessions (1-2 h) and preliminary feedback from five participants indicate comfortable wear and natural interaction. These results highlight the potential of CMS-based anticipatory sensing for practical wearable exoskeleton deployment in daily scenarios.