Motor Imagery EEG-based Control of Intelligent Wheelchair Using Deep Belief Network Coupled with OVO-CSP Algorithm

Aiming at the shortcomings of low recognition rate and the cumbersome feature extraction of traditional machine learning methods during the process of controlling wheelchair with electroencephalography(EEG) signals, this paper presents a scheme for controlling intelligent wheelchair with multi-class motor imagery (MI) EEG based on deep learning framework. The representative model of deep learning deep belief network (DBN) was used to classify the MI EEG. Firstly, the improved OVO-CSP algorithm is used to extract the features of multi-class EEG. Then, five kinds of MI EEG signals are trained and classified by DBN. Finally, the comparison with traditional machine classification methods such as SVM and BP Neural Network proves the effectiveness and validity of the proposed method. The results show that the deep belief network can better extract the essential characteristics of multi-class EEG and improve the classification accuracy.