Method and apparatus for detection of drowsiness and quantitative control of biological processes
First Claim
1. A method for monitoring a biological process, said method comprising:
- receiving image data of an object having one or more features, said image data corresponding to frames comprising a plurality of pixels;
dividing each frame into one or more regions based on the one or more features and each region into one or more subregions;
filtering pixels in each subregion according to a pixel intensity range to provide a filtered output of pixels;
defining for each frame a first area within each subregion, said first area defined by one or more predetermined shape equations according to a shape of a feature to be monitored in the subregion;
evaluating for each frame the number of filtered output of pixels in the first area for each subregion to transform the image data to a scalar feature signal for each subregion; and
transforming each scalar feature signal to produce a behavior indicating output signal for each subregion, said step of transforming comprising setting a value of the behavior indicating output signal to a value of a scalar feature signal at a first time if a value of a scalar feature signal at the first time is equal to or greater than the value of the behavior indicating output signal at a time preceding the first time and reducing the value of the behavior indicating output signal by a fraction if the value of the scalar feature signal at the first time is less than the value of the behavior indicating output signal at a time preceding the first time.
0 Assignments
0 Petitions
Accused Products
Abstract
The present invention is directed to a real-time automated video cognizer that functions as a facial video processor for detecting drowsiness in operators of motorized vehicles, including the use of a video cognizer to provide pattern recognition and control signal generation during monitoring of macroscopic or microscopic biological processes. More specifically, the present invention accepts input from a video monitoring system that continuously captures the operator'"'"'s facial images; employing three sequential means of processing the digitized video information to extract the position and configuration of drowsy-relevant facial features and numerically processes this information to yield a quantitative estimate of drowsiness probability in each epoch of monitoring. The means of the present invention are noninvasive, do not restrict driver movement or performance, provide increased measurement reliability for actual driver behavior, and include the capability of generating or triggering suitable alarms when drowsiness occurs.
42 Citations
34 Claims
-
1. A method for monitoring a biological process, said method comprising:
-
receiving image data of an object having one or more features, said image data corresponding to frames comprising a plurality of pixels; dividing each frame into one or more regions based on the one or more features and each region into one or more subregions; filtering pixels in each subregion according to a pixel intensity range to provide a filtered output of pixels; defining for each frame a first area within each subregion, said first area defined by one or more predetermined shape equations according to a shape of a feature to be monitored in the subregion; evaluating for each frame the number of filtered output of pixels in the first area for each subregion to transform the image data to a scalar feature signal for each subregion; and transforming each scalar feature signal to produce a behavior indicating output signal for each subregion, said step of transforming comprising setting a value of the behavior indicating output signal to a value of a scalar feature signal at a first time if a value of a scalar feature signal at the first time is equal to or greater than the value of the behavior indicating output signal at a time preceding the first time and reducing the value of the behavior indicating output signal by a fraction if the value of the scalar feature signal at the first time is less than the value of the behavior indicating output signal at a time preceding the first time. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17)
-
-
18. An apparatus for monitoring a biological process, said apparatus comprising a processor programmed to perform a method, said method comprising:
-
receiving image data of an object having one or more features, said image data corresponding to frames comprising a plurality of pixels; dividing each frame into one or more regions based on the one or more features and each region into one or more subregions; filtering pixels in each subregion according to a pixel intensity range to provide a filtered output of pixels; defining for each frame a first area within each subregion, said first area defined by one or more predetermined shape equations according to a shape of a feature to be monitored in the subregion; evaluating for each frame the number of filtered output of pixels in the first area for each subregion to transform the image data to a scalar feature signal for each subregion; and transforming each scalar feature signal to produce a behavior indicating output signal for each subregion, said step of transforming comprising setting a value of the behavior indicating output signal to a value of a scalar feature signal at a first time if a value of a scalar feature signal at the first time is equal to or greater than the value of the behavior indicating output signal at a time preceding the first time and reducing the value of the behavior indicating output signal by a fraction if the value of the scalar feature signal at the first time is less than the value of the behavior indicating output signal at a time preceding the first time. - View Dependent Claims (19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34)
-
Specification