Eyelid contour detection and tracking for startle research related eye-blink measurements from high-speed video records

Abstract

Using the positions of the eyelids is an effective and contact-free way for the measurement of startle induced eye-blinks, which plays an important role in human psychophysiological research. To the best of our knowledge, no methods for an efficient detection and tracking of the exact eyelid contours in image sequences captured at high-speed exist that are conveniently usable by psychophysiological researchers.

In this publication a semi-automatic model-based eyelid contour detection and tracking algorithm for the analysis of high-speed video recordings from an eye tracker is presented. As a large number of images have been acquired prior to method development it was important that our technique is able to deal with images that are recorded without any special parametrisation of the eye tracker. The method entails pupil detection, specular reflection removal and makes use of dynamic model adaption.

In a proof-of-concept study we could achieve a correct detection rate of 90.6%. With this approach, we provide a feasible method to accurately assess eye-blinks from high-speed video recordings.

Introduction

Emotions are considered to be a readiness for action and a strong motivational force: emotional states activate the organism for a certain behaviour [1]. Furthermore, emotion is closely connected to other cognitive processes, to which environmental stimuli we attend to and how we process them [2]. For this reason, the study of emotion is of principal interest to the psychological science. However, the measurement of emotion is associated with certain methodological limitations [3]. While self-report remains useful for many experimental fields, it also entails some profound disadvantages. In many situations people are not totally aware of their affective states or unable to verbalise them. Furthermore, self-report might be biased by social desirability. These constraints led to the development of research tools that measure behavioural and physiological responses to emotional stimuli. Emotions can be inferred from vocal changes, facial expressions and body postures [3]. Reaction times and error rates in tasks that require fast behavioural responses to emotional stimuli allow conclusions about the individual's affective states and evaluations [4]. In addition, affective states can also be measured via physiological markers—an object of research for a sub-discipline named psychophysiology. A typical experimental setting usually requires the participant to view or listen to emotional information while changes in physiological parameters are recorded. Such a psychophysiological method is the startle eye-blink response, which is modulated by the perceptual processing of affective information [5].