Blink recognition to detect the possibility of eye dystonia based on electrooculogram analysis

The present work proposes a system for detecting the possibility of eye dystonia from the analysis of eye movements. Eye movement data is recorded using an Electrooculogram signal acquisition system developed in the laboratory. Combinations of Hjorth parameters and Autoregressive (AR) parameters with Power Spectral Density (PSD) are used as signal features. Blinks are classified from other types of eye movements using Support Vector Machine (SVM) classifier with different kernel functions. We obtain a maximum average accuracy of 93.40% over all classes and participants using RBF-SVM classifier with a feature space of AR parameters of order 5 and PSD taken together. The system is designed to count the number of blinks in a particular interval of time thereby detecting the risk of eye dystonia, a neural disorder that causes a person to blink excessively.