Motion tracking of iris features to detect small eye movements

Abstract


 
 
 
 The inability of current video-based eye trackers to reliably detect very small eye movements has led to confusion about the prevalence or even the existence of monocular microsaccades (small, rapid eye movements that occur in only one eye at a time). As current methods often rely on precisely localizing the pupil and/or corneal reflection on successive frames, current microsaccade-detection algorithms often suffer from signal artifacts and a low signal-to-noise ratio. We describe a new video-based eye tracking methodology which can reliably detect small eye movements over 0.2 degrees (12 arcmin) with very high confidence. Our method tracks the motion of iris features to estimate velocity rather than position, yielding a better record of microsaccades. We provide a more robust, detailed record of miniature eye movements by relying on more stable, higher-order features (such as local features of iris texture) instead of lower-order features (such as pupil center and corneal reflection), which are sensitive to noise and drift. 
 
 
 

The Security-Utility Trade-off for Iris Authentication and Eye Animation for Social Virtual Avatars

“…Image data is encapsulated within a processing unit, reducing the chance that a malicious user can gain access. However, this also restricts applications that may utilize the iris for improved gaze estimation [14], realistic rendering of the user's eye [26], and iris authentication in cases where it is desired, such as logging into the Microsoft Hololens 2.…”

EllSeg: An Ellipse Segmentation Framework for Robust Gaze Tracking

“…Gaze estimation algorithms which solely rely on hand-crafted features are particularly susceptible to stray reflections (unanticipated patterns on eye imagery) and occlusion of descriptive gaze regions (such as eyelid covering the pupil or iris). Recent appearance-based methods based on Convolutional Neural Networks (CNNs) are better able to extract reasonably reliable gaze features despite the presence of reflections [3] or occlusions [31]. Additionally, for head-mounted eye-tracking systems, the degradation of gaze estimate accuracy over time due to slippage [25] can be minimized by estimating the 3D eyeball center of rotation [40] (loosely referred at as an "eyeball fit").…”

“…These elliptical fits are derived from identified pupil and iris segments. Efforts by Chaudhary et al [3] and Wu et al [47] demonstrate that CNNs can precisely segment eye images into its constituent parts, i.e., the pupil, iris, sclera and background skin regions.…”

“…EllSeg enables prediction of the pupil and iris as full elliptical structures despite the presence of occlusions. (2) To establish the utility of our methodology, we rigorously test our proposed 3-class ellipse segmentation framework using three network architectures, a modified Dense Fully Connected Network [20] (referred as DenseElNet), RITnet [3] and DeepVOG [48]. Performance is benchmarked with well defined train and test splits on multiple datasets, including some which are limited to labelled pupil centers only.…”

RITnet: Real-time Semantic Segmentation of the Eye for Gaze Tracking

“…In an effort to engage the machine learning and eyetracking communities in the field of eye-tracking for headmounted displays (HMD), Facebook Reality Labs issued the Open Eye Dataset (OpenEDS) Semantic Segmentation challenge which addresses part of the gaze estimation pipeline: identifying different regions of interest (e.g., pupil, iris, sclera, skin) in close-up images of the eye. Such tion of region-specific features (e.g., iridial feature tracking [2])and mathematical models which summarize the region structures (e.g., iris ellipse [17,1,13], or pupil ellipse [7]) used to derive a measure of gaze orientation.…”