Pupil Size and Blink Estimation for Human-Machine Interaction
Siyuan Chen
NICTA SML SEMINARDATE: 2013-09-05
TIME: 11:15:00 - 12:15:00
LOCATION: NICTA - 7 London Circuit
CONTACT: JavaScript must be enabled to display this email address.
ABSTRACT:
Eye activities, such as pupil size, blink and eye movement, contain valuable information about human cognition and are often used as an indicator of cognitive load change. A real-time cognitive load measurement (CLM) system is highly expected since some harmful consequences due to overload can be avoided if current cognitive load level is known. Currently, researchers often use commercial eye trackers to obtain eye activity dynamics and the question is whether it is the best option for such a real-time CLM system, since they are designed for eye gaze direction estimation and the accuracy of pupil size and blink, as a by-product, is often unknown. Meanwhile, calibrations are often required when environment settings or users are changed, which is harmful and inconvenient for a continuous measurement. In this talk, I will focus on the solution to pupil size measurement and blink detection from near-field infrared-illuminated eye videos recorded by webcams, which is applicable to any eye images without tuning parameters. I will illustrate the measurement problems of pupil diameter and blink from IR webcams for CLM, and discuss the results of using off-the-shelf IR webcams to obtain eye activity compared with that using an eye tracker. Based on the results, we can envisage a mini IR camera embedded in a lightweight glass frame, like Google Glass, for a convenient CLM in the future.
