Method and system of distinguishing between a glance event and an eye closure event

A subject monitoring system to distinguish between a downward glance event and an eye closure event, the system comprising:

• a camera to capture a number of images of at least a portion of a subject's face including at least one eye of the subject over a period of time, each image being captured at a corresponding time within the period of time;
• and
• a processor operatively associated with the camera to:
• process the images to determine a closure state of the at least one eye from a plurality of predefined closure states;
• detect potential eye closure events as a sequence of images in which the closure state of the at least one eye is determined to be in an at least partially closed state;
• upon detection of a potential eye closure event, further processing the sequence of images corresponding to the detected potential eye closure event to determine a corresponding head pitch angle of the subject across the sequence of images;
• and
• classify the detected potential eye closure event as either a downward glance event or an actual eye closure event,
• wherein a downward glance event occurs when a head pitch angle of the subject is detected to change by an angle greater than a predetermined head pitch angle and the at least one eye is determined to be in an at least partially closed state across the sequence of images corresponding to the detected potential eye closure event, and wherein an actual eye closure event occurs when the head pitch angle of the subject is detected to change by an angle less than the predetermined head pitch angle and the at least one eye is determined to be in an at least partially closed state across the sequence of images corresponding to the detected potential eye closure event.

A method of distinguishing between a downward glance event and an eye closure event in a subject monitoring system, the method comprising:

• a) capturing a number of images of at least a portion of a subject's face including at least one eye of the subject over a period of time, each image being captured at a corresponding time within the period of time;
• b) processing the images to determine:
• (i) a head pose of the subject;
• and
• (ii) a closure state of the at least one eye;
• c) detecting potential eye closure events as a sequence of images in which the at least one eye is determined to be in an at least partially closed state;
• d) upon detection of a potential eye closure event, further processing the sequence of images corresponding to the detected potential eye closure event to calculate a correlation between the head pose and the closure state of the at least one eye across the sequence of images;
• and
• e) based on the calculated correlation, classifying the detected potential eye closure event as either a downward glance event in which the at least one eye is not closed or an actual eye closure event in which the at least one eye is closed,
• wherein a downward glance event occurs when a head pitch angle of the subject is detected to change by an angle greater than a predetermined head pitch angle and the at least one eye is determined to be in an at least partially closed state across the sequence of images corresponding to the detected potential eye closure event, and wherein an actual eye closure event occurs when the head pitch angle of the subject is detected to change by an angle less than the predetermined head pitch angle and the at least one eye is determined to be in an at least partially closed state across the sequence of images corresponding to the detected potential eye closure event.