Multiple-tiered facial recognition

1. A method, comprising:

receiving probe image data associated with a biometric authentication request on a mobile device;

employing a first classifier to generate a first probability metric of the probe image data being associated with the authorized user;

approving the biometric authentication request responsive to the first probability metric being greater than a first threshold;

denying the biometric authentication request responsive to the first probability metric being less than a second threshold;

responsive to the first probability metric being between the first and second thresholds, employing a second classifier to generate a second probability metric of the probe image data being associated with the authorized user; and

approving the biometric authentication request responsive to the second probability metric being greater than a third threshold.

2. The method of claim 1 , wherein the probe image data comprises a probe set of characteristic data.

3. The method of claim 2 , wherein the first classifier is trained using a first library of user sets of characteristic data associated with an authorized user of the mobile device and a second library of non-user characteristic sets not associated with the authorized user to generate the first probability metric.

4. The method of claim 2 , wherein the first classifier comprises a support vector machine.

5. The method of claim 1 , wherein the second classifier comprises a first pairwise component that compares the probe image data to each entry in a first library of user sets of characteristic data associated with an authorized user of the mobile device to generate a first component of the second probability metric, wherein the first pairwise component employs a first characteristic feature set.

6. The method of claim 5 , wherein the second classifier comprises a second pairwise component that compares the probe image data to each entry in the first library to generate a second component of the second probability metric, wherein the second pairwise component employs a second characteristic feature set different than the first characteristic feature set.

7. The method of claim 6 , wherein the second classifier comprises a third pairwise component that employs a convolutional neural network to compare the probe image data to an average set of image data generated by averaging entries in the first library to generate a third component of the second probability metric.

8. The method of claim 7 , wherein the second classifier employs a majority voting technique using the first, second, and third components of the second probability metric.

9. The method of claim 8 , wherein each of the first, second, and third components of the second probability metric has an individual value of the third threshold.

10. The method of claim 9 , further comprising generating the individual values of the third threshold using the first library and a second library of non-user characteristic sets not associated with the authorized user.

11. A device, comprising:

a camera to generate probe image data associated with a biometric authentication request; and

a processor coupled to the camera to employ a first classifier to generate a first probability metric of the probe image data being associated with the authorized user, approve the biometric authentication request responsive to the first probability metric being greater than a first threshold, deny the biometric authentication request responsive to the first probability metric being less than a second threshold, responsive to the first probability metric being between the first and second thresholds, employ a second classifier to generate a second probability metric of the probe image data being associated with the authorized user, and approve the biometric authentication request responsive to the second probability metric being greater than a third threshold.

12. The device of claim 11 , wherein the probe image data comprises a probe set of characteristic data.

13. The device of claim 12 , wherein the first classifier is trained using a first library of user sets of characteristic data associated with an authorized user of the mobile device and a second library of non-user characteristic sets not associated with the authorized user to generate the first probability metric.

14. The device of claim 12 , wherein the first classifier comprises a support vector machine.

15. The device of claim 11 , wherein the second classifier comprises a first pairwise component that compares the probe image data to each entry in a first library of user sets of characteristic data associated with an authorized user of the mobile device to generate a first component of the second probability metric, wherein the first pairwise component employs a first characteristic feature set.

16. The device of claim 15 , wherein the second classifier comprises a second pairwise component that compares the probe image data to each entry in the first library to generate a second component of the second probability metric, wherein the second pairwise component employs a second characteristic feature set different than the first characteristic feature set.

17. The device of claim 16 , wherein the second classifier comprises a third pairwise component that employs a convolutional neural network to compare the probe image data to an average set of image data generated by averaging entries in the first library to generate a third component of the second probability metric.

18. The device of claim 17 , wherein the second classifier employs a majority voting technique using the first, second, and third components of the second probability metric.

19. The device of claim 18 , wherein each of the first, second, and third components of the second probability metric has an individual value of the third threshold.

20. The device of claim 19 , wherein the processor is to generate the individual values of the third threshold using the first library and a second library of non-user characteristic sets not associated with the authorized user.