Eye Movement Detection Using k Nearest Neighbor Method

Abstract

Switch using bio-signal are researching for communication of a physically disabled person such as ALS. We have researched bio-signal device for daily life support of ALS patients using ocular potential. We suggested Doubled Constant to RMS (DCR) method that is dynamic calculate the threshold value using RMS to detect to intentional action. However, DCR method has problem that the threshold increase from signal fluctuation, and cannot be detected large electrooculography change by intentional action. We need research better method for detection to intentional action. In recent years, it has been found that classification by pattern recognition in intentional action gives high accuracy for EMG. We compared the DCR method and the k neighborhood method which is one of the pattern recognition. It is because, Bio-signal switch has need better detection method. In result, each method has different detect conditions. In future, DCR method and knn method are combined for improved accuracy, and consider online analysis on EOG.