Processing and classification of electrooculogram sign for human machine interface

In this study, the source signal was generated from the Electrooculogram (EOG) signal to be used with the Human Machine Interface (IMA). EOG is a technique used to measure the movements of the muscles that move the human eye. EOG measurements were taken from a male individual. The received marks were preprocessed using Matlab and their attributes were made to be extracted. Features extracted from the EOG signals and labeled according to the system to be controlled. These labels were created horizontally on the right and left, on the vertical, up, down and winking. Features have been subjected to the classification process using Artificial Neural Networks (ANN). By using feature reduction algorithms, features that have little effect on the result were extracted and the classification success was tested with the obtained features. With the feature selection algorithm, the number of features selected according to the beginning was reduced by 60%, and among 22 features, Minimum, Average, Maximum, Median, Harmonic Mean, Skewness, T25 and T50 properties were obtained. The classification performances of these properties varied between 91% and 100%. When compared with the 91%, 95% and 98.7% classification performances obtained in similar studies in the literature, it is thought that the features obtained in this study can also be used in IMA. Keywords: Feature reduction, Artificial Neural Networks (ANN), Human Machine Interface (IMA), Electrooculogram (EOG)