Systems and methods for detecting impairment based upon facial image
Various embodiments provide systems and methods for identifying impairment using measurement devices and trained models, and/or for indicating interference with impairment testing. Embodiments discussed herein provide systems, methods, and/or devices that enable remote impairment testing that does not require a human monitor to be present or physically near the individual being monitored. Such an ability is an improvement.
1. A system for detecting drug or alcohol impairment based upon facial image, the system comprising:
a camera configured to capture a static facial image of an individual;
one or more processors;
a non-transient computer readable medium coupled to the one or more processors, and having stored therein instructions which when executed by the one or more processors, causes the one or more processors to:
receive the static facial image of the individual from the camera;
apply a facial image impairment model to the static facial image to yield a probability that the individual is drug or alcohol impaired;
indicate a likelihood of impairment based at least in part on a determination that the probability exceeds a first threshold;
indicate no impairment when the probability is less than a second threshold;
apply an anomaly detection model to the static facial image to yield an anomaly output corresponding to the individual; and
wherein the likelihood of impairment is indicated when both the anomaly output indicates that the static facial image is an anomaly for the individual and the probability exceeds the first threshold.
2. The system of claim 1 , wherein the anomaly detection model is trained using at least ten instances of static facial images of the individual.
3. The system of claim 1 , wherein the non-transient computer readable medium further having stored therein instructions which when executed by the one or more processors, causes the one or more processors to:
cause a request to be sent to the individual to perform an additional impairment test.
4. The system of claim 3 , wherein the additional impairment test is a voice based impairment test.
5. The system of claim 1 , wherein the non-transient computer readable medium further having stored therein instructions which when executed by the one or more processors, causes the one or more processors to:
forward the static facial image to a user for classification when the probability is both less than the first threshold and greater than the second threshold.
6. The system of claim 1 , wherein the non-transient computer readable medium further having stored therein instructions which when executed by the one or more processors, causes the one or more processors to:
report the likelihood of impairment to a recipient device apart from the one or more processors.
7. The system of claim 1 , wherein the facial image impairment model is a machine learning model trained using at least one hundred static facial images.
8. The system of claim 7 , wherein the at least one hundred static facial images correspond to at least ten different individuals undergoing a facial image based impairment test.
9. The system of claim 1 , wherein the anomaly detection model is trained on a plurality of images in which the individual is not drug or alcohol impaired.
10. A method for detecting impairment based upon facial images, the method comprising:
receiving, by a processor, a static facial image of an individual from a camera;
applying, by the processor, a facial image impairment model to the static facial image to yield a probability that the individual is drug or alcohol impaired, wherein the facial image impairment model is a machine learning model trained using at least one hundred static facial images, and wherein the at least one hundred static facial images correspond to at least ten different individuals undergoing a facial image based impairment test;
indicating, by the processor, a likelihood of impairment based at least in part on a determination that the probability exceeds a first threshold;
indicating, by the processor, no impairment when the probability is less than a second threshold;
applying, by the processor, an anomaly detection model to the static facial image to yield an anomaly output corresponding to the individual; and
wherein the likelihood of impairment is indicated when both the anomaly output indicates that the facial image is an anomaly for the individual and the probability exceeds the first threshold.
11. The method of claim 10 , wherein the anomaly detection model is trained using at least ten instances of static facial images of the individual.
12. The method of claim 10 , the method further comprising:
causing, by the processor, a request to be sent to the individual to perform an additional impairment test.
13. The method of claim 12 , wherein the additional impairment test is a voice based impairment test.
14. The method of claim 10 , the method further comprising:
forwarding, by the processor, the static facial image to a user for classification when the probability is both less than the first threshold and greater than the second threshold.
15. The method of claim 10 , the method further comprising:
reporting the likelihood of impairment to a recipient device apart from the one or more processors.
16. The method of claim 10 , wherein the anomaly detection model is trained on a plurality of images in which the individual is not drug or alcohol impaired.
17. A non-transient computer readable medium having stored therein instructions, which when executed by a hardware processing system cause the hardware processing system to:
receive a static facial image of an individual from a camera;
apply a facial image impairment model to the static facial image to yield a probability that the individual is drug or alcohol impaired, wherein the facial image impairment model is a machine learning model trained using at least one hundred static facial images, and wherein the at least one hundred static facial images correspond to at least ten different individuals undergoing a facial image based impairment test;
indicate a likelihood of impairment based at least in part on a determination that the probability exceeds a first threshold;
indicate no impairment when the probability is less than a second threshold;
apply an anomaly detection model to the facial image to yield an anomaly output corresponding to the individual; and
wherein the likelihood of impairment is indicated when both the anomaly output indicates that the static facial image is an anomaly for the individual and the probability exceeds the first threshold.
18. The non-transient computer readable medium of claim 17 , wherein the anomaly detection model is trained using at least ten instances of static facial images of the individual.
19. The non-transient computer readable medium of claim 17 , wherein the non-transient computer readable medium further having stored therein instructions which when executed by the one or more processors, causes the one or more processors to:
cause a request to be sent to the individual to perform an additional impairment test, wherein the additional impairment test is a voice based impairment test.
20. The non-transient computer readable medium of claim 17 , wherein the anomaly detection model is trained on a plurality of images in which the individual is not drug or alcohol impaired.