Human identification using Linear Multiclass SVM and Eye Movement biometrics

The paper presents a system to accurately differentiate between unique individuals by utilizing the various eye-movement biometric features. Eye Movements are highly resistant to forgery as the generation of eye movements occur due to the involvement of complex neurological interactions and extra ocular muscle properties. We have employed Linear Multiclass SVM model to classify the numerous eye movement features. These features were obtained by making a person fixate on a visual stimuli. The testing was performed using this model and a classification accuracy up to 91% to 100% is obtained on the dataset used. The results are a clear indication that eye-based biometric identification has the potential to become a leading behavioral technique in the future. Moreover, its fusion with different biometric processes such as EEG, Face Recognition etc., can also increase its classification accuracy.