Purchase this article with an account.
Sebastiaan Mathôt, Jean-Baptiste Melmi, Lotje Van der Linden, Stefan Van der Stigchel; The mind-writing pupil: near-perfect decoding of visual attention with pupillometry. Journal of Vision 2015;15(12):176. doi: 10.1167/15.12.176.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
When the eyes are exposed to light, the pupils constrict. This is the well known pupillary light response. What is less well known is that the pupillary light response is not a simple reflex to light, but is modulated by visual attention: When you covertly attend to a bright stimulus, your pupil constricts relative to when you attend to a dark stimulus. Here we describe a human-computer interface that is based on this principle, i.e. decoding the focus of covert visual attention with pupillometry. Participants fixated in the center of a display, and selected (i.e. covertly attended to) one of several stimuli presented in a circular arrangement. Each stimulus was presented on a background with alternating brightness. Small changes in pupil size reflected the brightness alternations of the selected stimulus' background, and this allowed us to determine which stimulus was selected with nearly perfect accuracy on a trial-by-trial basis. An extension of this technique, in which the stimulus array serves as a virtual keyboard, even allows for arbitrary text input. As a human-computer interface, this technique has several key advantages: It is intuitive, because there is a direct mapping between the task (attending to a stimulus) and the goal (selecting a stimulus); It allows for bi-directional communication, because the analysis can be performed on-line; It is non-invasive and can be done with currently available low-cost eye trackers; It is reliable, i.e. decoding accuracy is nearly perfect under good conditions. We discuss potential applications, such as communication with locked-in patients and ultra-secure password input.
Meeting abstract presented at VSS 2015
This PDF is available to Subscribers Only