Surface Electromyography Signal Processing for Upper-limb Hybrid Assist-as-need Exoskeleton intended for Stroke Patients

Abstract

Stroke, which is the second largest disability worldwide, may induce one side transient paralysis. The rehabilitation method is a key point for minimizing the time of recovery, for increasing the patient comfort or for decreasing the cost of stroke therapy for the society. This thesis aimed to process the raw surface electromyography (sEMG) signal coming from the biceps to extract useful information to trigger an actuator combined to a exoskeleton or to apply functional electrical stimulation (FES) for stroke rehabilitation. Thus, the FES/actuator triggering will be assist-as-needed.

The data acquisition was first performed using an sEMG prototype circuit. The sEMG signal was recorded at a given sample frequency (1 kHz) and was displayed in real-time. Several kinds of surface electrodes were tested and characterized.

Next, several muscle activity evaluations were analysed to find the most appropriated. The sEMG signal in presence in FES was recorded. It has been shown the alteration of the stimulation artifacts and the need of further analysis to remove them for providing consistent data.

Finally, interesting information such as intention, force and fatigue estimation were extracted from the sEMG signal in real-time. Intention is the time activation of a muscle and can be used to trigger FES or to control the actuator. Force estimation is the signal normalisation according to a reference providing an estimation of the force exerted by the muscle. This data provides a proportional indication for FES or the actuator. From the sEMG signal, localized muscular fatigue was assessed. This may optimize the rehabilitation session and may also track the patient progress. Some characteristics of sEMG signal during an elbow flexion/extension movement were studied.