A control system of lower limb exoskeleton robots based on motor imagery

Abstract

In this paper, we have developed an asynchronous brain-computer interface (BCI)-based lower limb exoskeleton control system based on motor imagery (MI). By decoding electroencephalography (EEG) signals in real-time, users were able to walk forward, sit down, and stand up while wearing the exoskeleton. EEG feature vectors associated with the motor imagery were extracted from the filtered EEG signals with common spatial patterns (CSP) method. And support vector machine (SVM) was employed to address an EEG-based three - class motor imagery classification task. Overall, four healthy subjects participated in the experiment to evaluate performance. To achieve a better classification result, parameters of CSP and SVM were trained in the offline experiment. In the subsequent online experiment, the results exhibited accuracies of 85.33%, 84%, 72.67% and 81.33%. It indicates that subjects can complete the task fluently and the control system can decode EEG signals with a high recognition rate. Further, the combination in decision table obviously reduces the probability of wrong actions.