Modeling of Mixed Artificially and Voluntary Induced Muscle Contractions for Controlled Functional Electrical Stimulation of Shoulder Abduction. Issue 34 (2019)

• Record Type: Journal Article
• Authors: Schauer, T.; Vrontos, A.
• Is Part Of: IFAC-PapersOnLine. Volume 51:Issue 34(2019)
• Page Start: 284
• Page End: 289
• Publication Date: 2019
• Subjects: System Identification -- Functional Electrical Stimulation -- Electromyography -- Neuro-prosthetics -- Rehabilitation -- Stroke -- Feedback Control; Automatic control -- Periodicals; 629.805
• DOI: 10.1016/j.ifacol.2019.01.038
• Languages: English
• ISSNs: 2405-8963

Abstract

Abstract: This contribution investigates the modeling of shoulder abduction movements caused by simultaneous voluntary and artificially evoked muscle contractions. The latter are generated by feedback controlled functional electrical stimulation. To maintain a desired level of muscle activation (recruitment) by electrical stimulation also in presence of fatigue, we assess the recruitment in real-time from the M-wave, i.e., the electromyography (EMG) response caused by each applied stimulation pulse, and regulate it by a simple integral controller that adjusts the stimulation intensity. The reference of the FES-induced muscle recruitment serves as one input to the model. The voluntary muscle activity is also determined from EMG recordings by filtering and represents the second model input. A simple linear transfer function model of 2nd order captures the shoulder abduction angle in response to the estimated volitional and FES-induced muscle activity, while the two inputs are summed up after weighting. Least squares have been used to determine the model parameters from recorded input-output data obtained at a neurologically intact subject. The model describes the shoulder abduction movements well for the investigated angular range of 100 ◦ also under presence of muscular fatigue due to the applied feedback control of the FES-induced muscle activity. A root mean square error of 8 ◦ has been observed between the simulated and measured angle for validation data covering 5 minutes. The model can be used in future to design FES support systems for stroke patients with weak residual muscle activity. Due to the model's simplicity, online identification via recursive least squares and adaptive control schemes are feasible.