Linearisation of electrically stimulated muscles by feedback control of the muscular recruitment measured by evoked EMG

A novel feedback control method for neuro-prosthetic systems is presented which linearises the static input non-linearity of muscles that are artificially activated by Functional Electrical Stimulation (FES). The proposed method controls the activation state (sum of recruited motor units by FES) of the paralysed muscle. This muscle recruitment state is measured by the FES evoked electromyogram (eEMG). Compared to standard control approaches, no cumbersome off-line or online identification of the complete nonlinear muscle dynamics, often described by Hammerstein or Hill models, and no model inversion are required. The developed approach robustly handles uncertainties and time-variances of the highly nonlinear muscular recruitment behaviour. When building motion control systems for neuro-prostheses, the developed feedback controller should be used at an inner loop receiving command signals (desired muscle activation) from a top level joint-angle control loop. The feasibility of the proposed control scheme has been experimentally demonstrated for the control of the elbow-joint angle in healthy subjects.