Processing signals from surface electrode arrays for noninvasive 3D mapping of muscle activity

This paper presents a new technique for reconstruction and imaging of volumetric neuromuscular activity using digital signal processing of multichannel surface potential recordings. We use magnetic resonance images to model anatomical structures with finite element analysis and to quantify material properties within the inhomogeneous volume. Time-frequency distributions of the multichannel EMG array data decompose the broadband source localization problem into a narrowband framework. Poisson's equation is then solved using finite element methods coupled with signal processing estimation techniques to localize neuromuscular activity. This new imaging tool can be used to enhance clinical diagnosis of neuromuscular disorders and to improve understanding of human locomotion for biomechanics and robotics.<>