A Calibration Method for Blink Type Classification

Abstract

SUMMARYHuman–computer interaction systems that use eye blinks have been reported. One purpose of these systems is as a communication aid for persons with severe physical disabilities such as amyotrophic lateral sclerosis (ALS). Some of these systems classify the user's voluntary (conscious) blinks. If the voluntary blinks can be distinguished automatically, the usability of the input interface will be improved. We have proposed an eye blink extraction method that utilizes split interlaced images of the eye. These split images are odd‐ and even‐field images and are generated from interlaced images. The proposed method yields a time resolution that is double that of the 1080i Hi‐Vision format. We refer to this approach as a “frame splitting method”. We realized a method for automatic eye blink extraction using the frame splitting method. Based on this extraction method, we also proposed a blink type classification method in which a threshold is set within the duration of an eye blink. Our classification method requires a calibration process before measurement. In this paper, we clarify the required minimum number of eye blinks for calibration. We also discuss the performance of blink type classification using this calibration.