An Automatic Classification Method for Involuntary and Two Types of Voluntary Blinks

Abstract

Input systems using eye blinks have been proposed. A main purpose of these systems is communication aid for the severely disabled. For employing eye blinks as a command input, these input systems need to detect voluntary (conscious) blinks. We developed a measurement method for variation in pixels of open-eye area from an image sequence. This measurement method enables us to extract blinking wave patterns. We previously proposed an automatic classification method between one type of voluntary blinks and involuntary (unconscious) blinks. If the types of classifiable voluntary blinks increase, we can assign an individual command to each type. Applying this blink type classification to a human-computer interface will improve the efficiency when inputting commands. In this paper, we introduce a new type of voluntary blinks to increase classifiable blink type. In addition, our classification method is extended for automatic classification between two types of voluntary blinks and involuntary blinks. This new classification method is realized by two blink type determinations based on wave pattern parameters that we employ. We clarify proper parameter combinations for the new classification method. Using these proper parameter combinations, we achieved approximately 95% classification rates for 10 subjects.