Classification of Eye Movements Using Electrooculography and Neural Networks

Electrooculography is a technique for measuring the cornea-retinal potential produced by eye movements. This paper proposes algorithms for classifying eleven eye movements acquired through electrooculography using dynamic neural networks. Signal processing techniques and time delay neural network are used to process the raw signals to identify the eye movements. Simple feature extraction algorithms are proposed using the Parseval and Plancherel theorems. The performances of the classifiers are compared with a feed forward network, which is encouraging with an average classification accuracy of 91.40% and 90.89% for time delay neural network using the Parseval and Plancherel features.