Mind Robotic Rehabilitation Based on Motor Imagery Brain Computer Interface

In this paper, a human robot interface called mind robotic rehabilitation is developed for regular training of neurological rehabilitation for stroke patients. The mind robotic rehabilitation is developed based on non-invasive motor imagery brain computer interface (BCI) technology. The use of a spatial filtering algorithm, common spatial pattern (CSP), is proposed for extracting features that maximize the discrimination of two different brain states, left hand movement imagination and right hand movement imagination, during motor imagery of the subject. Furthermore, we find that a feature fusion of feature vectors from both CSP and autoregressive (AR) spectral analysis can obviously improve the performance of the BCI. Quadratic discriminant analysis (QDA) is applied to the combined feature vectors and classifies the vectors into left or right motor imagery category. For evaluation of the proposed BCI, we compare the performance of the proposed method against methods using single feature extraction algorithm, i.e. CSP only or AR spectral analysis only, under an equivalent experiment environment and using the same classifier to estimate the classification accuracy. It is found that feature fusion significantly improves BCI performance.