Eye tracking method and system

Inventors

De Villers-Sidani, Etienne • DROUIN-PICARO, Paul Alexandre

Assignees

Innodem Neurosciences

Abstract

A method for determining a series of gaze positions of at least one eye over time is provided. The method comprises capturing a video of a user's face simultaneously with displaying a stimulus video on a screen and extracting at least one color component for each one of a plurality of images obtained from the video of the user's face. Based on the at least one color component for each one of the plurality of images, the series of gaze positions of the user's face over the time of the video is determined. A system for determining a series of gaze positions of at least one eye over time is also provided.

Core Innovation

The invention determines a series of gaze positions of at least one eye of a user over time without dedicated tracking hardware, by receiving a camera video of a user's face while stimuli are displayed on a screen. The method determines coordinates of stimuli to be displayed and receives the video of the user's face from the camera captured during displaying of the stimuli. From the resulting video, it extracts at least one color component for each image obtained from the video.

Based on the extracted color component(s), the invention determines the series of gaze positions over time of the video of the user's face. In system terms, a camera is configured to capture a video of a user's face during displaying of a stimulus video on a screen, and a processing unit extracts at least one color component for each image and determines a sequence of gaze positions based on the color component(s).

The disclosure further includes embodiments in which internal representations derived from the color components for each image are used with feature fusion to determine gaze coordinates. It also includes embodiments in which illumination/illuminant values and facial landmarks are processed via respective neural-network streams and combined with gaze estimates, and calibration models are used to transform gaze estimates into screen-referenced gaze coordinates.

Claims Coverage

The independent claims cover three main approaches: a computer-implemented method using stimulus coordinate determination, face-video capture, color-component extraction, and gaze-position inference; a corresponding gaze-determination system with the same core processing pipeline; and a method variant that emphasizes simultaneous stimulus-video presentation and gaze determination over the video time using color components. Across the independent claims, the coverage centers on extracting color component(s) from multiple images and determining gaze positions over time based on those color component(s).

Across the independent claims, the core inventive scope is the use of color component(s) extracted from images in a user's face video to determine a time sequence of gaze positions, with the system and method variants differing primarily in how the overall architecture is framed and in the parallel presentation of stimulus video and the gaze-determination time axis. Dependent refinements include internal representation derivation from color components and calibration/database-based refinements tied to stimulus coordinates and subsequent gaze determination.

Stated Advantages

Documented Applications

No documented applications found