Online Voluntary Eye Blink Detection using Electrooculogram

This paper describes the voluntary eye blink detection method using electrooculogram (EOG). There are still challenge problems to put brain-computer interface (BCI) systems to real-life applications. In general BCI systems, there is a possibility of incorrect and unintentional input because input is automatically selected even if the requirements are accidentally met. This study aims to propose the voluntary eye blink detection method and apply it to the trigger switch of BCI systems. In the proposed method, normal blink, double blink, and wink can be detected from vertical and horizontal EOG signals. We employed the positive peak of vertical and horizontal amplitude and maximum cross correlation coefficient between vertical EOG and template signal of double blink in feature extraction. Eye blinks were classified by support vector machine. As the result of simulations, an average accuracy of 97.28% was obtained using our method. In addition, the best accuracy for voluntary eye blinks was obtained for wink with accuracy of 97.69 %. This paper proof wink is suitable for trigger switch of BCI system, and online method for voluntary eye blink detection.