Estimation of Muscle Activity in One-Leg Stance from 3D Surface Deformation

Muscular activity during human motion is usually quantified by measuring the electrical potential during muscle activation using electromyography (EMG). However, apart from producing electrical activity, muscular contraction of many skeletal muscles also induces subtle deformation of the skin surface. In this paper, we present a method to estimate muscular activation from such 3D skin deformation. To this end, we introduce a capture system that reconstructs the 3D motion of the skin from multi-view video data and simultaneously measures true muscle activity with EMG sensors. Our data reveals strong correlations between the skin deformation and muscular activity during one-leg stances. We propose a pose normalization procedure and a novel model based on Supervised Principal Component Regression that automatically segments individual muscles and estimates their activation from 3D surface deformation. Our evaluation shows that the model generalizes to varying body shapes and that the estimated activation closely fits the measured EMG data.