A Feasibility Study of Employing EOG Signal in Combination with EEG Based BCI System for Improved Control of a Wheelchair

摘要

For a fully paralysed person, EEG (Electroencephalogram) based Brain Computer Interface (BCI) has a great promise for controlling electromechanical equipment such as a wheelchair. Again EOG (Electrooculography) based Human Machine Interface system also provides a possibility. Individually, none of these methods is capable of giving a fully error free reliable and safe control, but an appropriate combination may provide a better reliability, which is the aim of the present work. Here we intend to use EEG data to classify two classes, corresponding to left and right hand movement, and EOG data to classify two classes corresponding to left and right sided eyeball movement. We will use these classifications independently first and then combine these with different weightage to find if a better and reliable control is possible. For this purpose offline classification of motor imaginary EEG data of a subject was carried out extracting features using Common Spatial Pattern (CSP) and classifying using Linear Discriminative Analysis. The independent EEG motor imaginary data classification resulted in 89.8% of accuracy in 10 fold one leave out cross validation. The EOG eyeball movement produces distinctive signals of opposite polarities and is classified using a simple discriminant type classification resulting in 100% accuracy. However, using EOG solely is not acceptable as there always will be unintentional eye movement giving false commands. Combining both EEG and EOG with different weightage to the two classifications produced varied degrees of improvement. For 50% weightage to both resulted in 100% accuracy, without any error, and this may be accepted as a practical solution because the chances of unintentional false commands will be very rare. Therefore, a combination of EOG and BCI may lead to a greater reliability in terms of avoidance of undesired control signals.