Eye tracking device, eye tracking method, and computer-readable medium

An eye tracking method, comprising the following steps:

• sending radiation to the eye of a user by one or more glint sources such that said radiation is reflected off the cornea of said eye and is received by said event-based optical sensor in a form of one or more individual glints;
• receiving a signal stream of events produced by an event-based optical sensor due to radiation reflected off the eye of the user being received by said event-based optical sensor, each event corresponding to detection of a temporal change in the received radiation at one or more pixels of said optical sensor;
• and
• generating based on said signal stream of events:
• gaze information of said eye at a first instant of time, extracted by an extraction module that comprises a machine learning module and/or an artificial neural network,
• first glint information of said one or multiple glints at said first instant of time, and
• second glint information of said one or multiple glints at a second instant of time that is later than the first instant of time,
• wherein said predicted gaze information of said eye is generated at said second instant of time based on said second glint information;
• and model parameters, which are live-trained using glint data and pupil data prior to the next extraction of gaze information.

A computer-readable medium comprising instructions which, when executed by a computer or microcontroller, cause the computer or microcontroller to carry out the following steps:

• sending radiation to the eye of a user by one or more glint sources such that said radiation is reflected off the cornea of said eye and is received by said event-based optical sensor in form of one or more individual glints;
• receiving a signal stream of events produced by an event-based optical sensor due to radiation reflected off the eye of the user being received by said event-based optical sensor, each event corresponding to detection of a temporal change in the received radiation at one or more pixels of said optical sensor;
• and
• generating based on said signal stream of events:
• gaze information of said eye at a first instant of time,
• extracted by an extraction module that comprises a machine learning module and/or an artificial neural network,
• first glint information of said one or multiple glints at said first instant of time, and
• second glint information of said one or multiple glints at a second instant of time that is later than the first instant of time,
• wherein said predicted gaze information of said eye is generated at said second instant of time based on said second glint information and model parameters, which are live-trained using glint data and pupil data prior to the next extraction of gaze information.