Hybrid EEG-EOG system for intelligent prosthesis control based on common spatial pattern algorithm

Abstract

A newly brain computer interaction (BCI) system which combined electrooculography (EOG) with electroencephalography (EEG) was designed and realized to make prosthesis control with multicontrol commands come true. Based on two different imaginary tasks, time-frequency spectrum and brain topographic mapping were used to analyze and validate event related synchronization/desychronization (ERS/ERD). Obviously frequency band of EEG is chosen based on common spatial pattern (CSP) algorithm, and then support vector machine (SVM) were implemented to classify data's characteristics. In order to improve the performance of BCI system, the EOG signals caused by eye movements and EEG signals of ERS/ERD were simultaneously provided as the input of hybrid BCI system. In the lab environment, the proposed system was tested on a multi-functional robotic prosthesis. The robot is light in weight and low in power consumption. Taking advantage of both EOG and EEG, recognition accuracy of entire system reaches 91.1%. The experiments results show that feature information of EEG signal which is evoked by motor imagery have obvious spatial specificity. This hybrid system can recognize different motions of the upper limb, whereas the task can be executed effectively and operator controls the prosthesis easily as their own arm.