Detection of wrist movement using EEG signal for brain machine interface

Brain machine interfaces (BMIs) allow patients suffering from neuromuscular disorders to control the movement of robotic limb or wheelchair under their own guidance. So far only invasive technologies e.g. Electrocorticography (ECoG) or intracranial EEG (iEEG) have been widely acknowledged in the design of BMIs. In this paper Electroencephalography (EEG), a non-invasive technology, has been used. The paper deals with study of the features of EEG signals corresponding to two different movements of human hand, namely flexion and extension. The movements have been detected on the basis of the energy and entropy of the corresponding signals. A total of twelve features have been used. Using different combinations of these features a surprisingly high accuracy of 87% has been obtained. Moreover, the use of only discrete cosine transformation of energy and entropy has yielded even a higher average accuracy of 91.93%. With such results, this wrist movement detection algorithm is successfully implemented on a robotic arm.