APPARATUS AND METHODS FOR BIOMETRIC VERIFICATION

• US 20200134145A1
• Filed: 10/30/2018
• Published: 04/30/2020
• Est. Priority Date: 10/30/2018
• Status: Active Grant

• Current Assignee: Qualcomm, Inc.
• Original Assignee: Qualcomm, Inc.
• Inventors: BAPST, Aleksander, PAGNOTTA, Christopher Alan, LAFLEUR, Ryan, PETERSON, Evan James
• Granted Patent: US 11,244,035 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 21/316 by observing the pattern of computer usage, e.g. typical user behaviour
• CPC Class Codes: G06F 21/32 using biometric data, e.g. fingerprints, iris scans or voiceprints
• CPC Class Codes: G06F 21/45 Structures or tools for the administration of authentication

Abstract

Example apparatus and methods may include: (1) a system and method for enrolling a single user'"'"'s gait pattern using only accelerometer magnitude data; (2) an efficient and low-power method for continuously generating biometric match scores against the enrolled model at run-time using two phases: detection and authentication; (3) an adaptive method for continuously updating the user model by comparing temporary models generated at regular intervals to the canonical user model; and (4) accelerometer samples captured on a wearable and streamed to a mobile device for further processing.

Claims

1. A method for biometric verification, the method comprising:

detecting, by an accelerometer, a plurality of samples;

determining, by a mobile device coupled to the accelerometer, a gait based on the plurality of samples;

determining if an authorized user of the mobile device is enrolled;

if the authorized user of the mobile device is determined to be not enrolled;

extracting a plurality of gait segments from the plurality of samples;

detecting a plurality of peaks in the gait segments;

extracting a plurality of inter-peak segments from the plurality of gait segments;

interpolating the plurality of inter-peak segments to a preset length; and

storing an enrollment model of the authorized user in a memory of the mobile device, wherein the enrollment model is based on the interpolated plurality of inter-peak segments.

2. The method of claim 1, further comprising generating a plurality of clusters based on the interpolated plurality of inter-peak segments;

determining a plurality of centroids for the plurality of clusters; and

storing the plurality of centroids as the enrollment model of the authorized user.

3. The method of claim 1, further comprising if the authorized user of the mobile device is determined enrolled:

extracting a plurality of gait segments from the plurality of samples;

detecting a plurality of peaks in the gait segments;

extracting a plurality of inter-peak segments from the plurality of gait segments;

interpolating the plurality of inter-peak segments to a preset length;

generating a plurality of clusters based on the interpolated plurality of inter-peak segments;

determining a plurality of centroids for the plurality of clusters;

generating a match score by comparing the plurality of centroids to an enrollment model of the authorized user; and

authenticating the authorized user based on the match score.

4. The method of claim 3, wherein the match score is determined based on an average minimum distance between each of the plurality of centroids and a plurality of user model centroids.

5. The method of claim 1, wherein the plurality of samples are accelerometer magnitudes.

6. The method of claim 1, wherein the accelerometer is incorporated into the mobile device and the memory is a ring buffer.

7. The method of claim 1, wherein the mobile device is incorporated into a device selected from the group consisting of a music player, a video player, an entertainment unit, a navigation device, a communications device, a mobile phone, a smartphone, a personal digital assistant, a tablet computer, a computer, a wearable device, and a laptop computer.

8. A non-transitory computer-readable medium comprising instructions that when executed by a processor cause the processor to perform a method comprising:

detecting, by an accelerometer, a plurality of samples;

determining, by a mobile device coupled to the processor, a gait based on the plurality of samples;

determining if an authorized user of the mobile device is enrolled;

if the authorized user of the mobile device is determined to be not enrolled;

extracting a plurality of gait segments from the plurality of samples;

detecting a plurality of peaks in the gait segments;

extracting a plurality of inter-peak segments from the plurality of gait segments;

interpolating the plurality of inter-peak segments to a preset length; and

storing the interpolated plurality of inter-peak segments in a memory of the mobile device as an enrollment model of the authorized user.

9. The non-transitory computer-readable medium of claim 8, wherein the method further comprises generating a plurality of clusters based on the interpolated plurality of inter-peak segments;

determining a plurality of centroids for the plurality of clusters; and

storing the plurality of centroids as the enrollment model of the authorized user.

10. The non-transitory computer-readable medium of claim 8, wherein the method further comprises if the authorized user of the mobile device is determined enrolled:

extracting a plurality of gait segments from the plurality of samples;

detecting a plurality of peaks in the gait segments;

extracting a plurality of inter-peak segments from the plurality of gait segments;

interpolating the plurality of inter-peak segments to a preset length;

generating a plurality of clusters based on the interpolated plurality of inter-peak segments;

determining a plurality of centroids for the plurality of clusters;

generating a match score by comparing the plurality of centroids to an enrollment model of the authorized user; and

authenticating the authorized user based on the match score.

11. The non-transitory computer-readable medium of claim 10, wherein the match score is determined based on an average minimum distance between each of the plurality of centroids and a plurality of user model centroids.

12. The non-transitory computer-readable medium of claim 8, wherein the plurality of samples are accelerometer magnitudes.

13. The non-transitory computer-readable medium of claim 8, wherein the accelerometer and the processor are incorporated into the mobile device and the memory is a ring buffer.

14. The non-transitory computer-readable medium of claim 8, wherein the non-transitory computer-readable medium is incorporated into a device selected from the group consisting of a music player, a video player, an entertainment unit, a navigation device, a communications device, a mobile phone, a smartphone, a personal digital assistant, a tablet computer, a computer, a wearable device, and a laptop computer.

15. An apparatus for biometric verification, the apparatus comprising:

an accelerometer configured to detect a plurality of samples;

a processor coupled to the accelerometer, the processor configured to;

determine a gait based on the plurality of samples;

determine if an authorized user of the apparatus is enrolled;

if the authorized user is determined to be not enrolled;

extract a plurality of gait segments from the plurality of samples;

detect a plurality of peaks in the gait segments;

extract a plurality of inter-peak segments from the plurality of gait segments;

interpolate the plurality of inter-peak segments to a preset length; and

store the interpolated plurality of inter-peak segments in a memory as an enrollment model of the authorized user.

16. The apparatus of claim 15, wherein the processor is further configured to generate a plurality of clusters based on the interpolated plurality of inter-peak segments;

determine a plurality of centroids for the plurality of clusters; and

store the plurality of centroids as the enrollment model of the authorized user.

17. The apparatus of claim 15, wherein the processor is further configured to, if the authorized user of a mobile device is determined enrolled:

extract a plurality of gait segments from the plurality of samples;

detect a plurality of peaks in the gait segments;

extract a plurality of inter-peak segments from the plurality of gait segments;

interpolate the plurality of inter-peak segments to a preset length;

generate a plurality of clusters based on the interpolated plurality of inter-peak segments;

determine a plurality of centroids for the plurality of clusters;

generate a match score by comparing the plurality of centroids to an enrollment model of the authorized user; and

authenticate the authorized user based on the match score.

18. The apparatus of claim 17, wherein the match score is determined based on an average minimum distance between each of the plurality of centroids and a plurality of user model centroids.

19. The apparatus of claim 15, wherein the plurality of samples are accelerometer magnitudes.

20. The apparatus of claim 15, wherein the accelerometer, the processor, and the memory are incorporated into a mobile device and the memory is a ring buffer.

21. The apparatus of claim 15, wherein a mobile device is incorporated into a device selected from the group consisting of a music player, a video player, an entertainment unit, a navigation device, a communications device, a mobile phone, a smartphone, a personal digital assistant, a tablet computer, a computer, a wearable device, and a laptop computer.

22. A method for biometric verification, the method comprising:

detecting, by an accelerometer, a plurality of samples;

determining, by a mobile device, if the plurality of samples indicates a gait; and

when the gait is determined;

determining if an authorized user of the mobile device is enrolled;

when the authorized user of the mobile device is not enrolled;

collecting an initial plurality of gait samples;

determining a peak for the initial plurality of gait samples;

extracting inter-peak segments for the initial plurality of gait samples;

interpolating the inter-peak segments of the initial plurality of gait samples to a preset length;

generating an initial user model based on the interpolated inter-peak segments of the initial plurality of gait samples; and

authenticating the user of the mobile device;

when the authorized user of the mobile device is enrolled;

selecting a current user model;

collecting a first plurality of gait samples;

determining a peak for the first plurality of gait samples;

extracting inter-peak segments for the first plurality of gait samples;

interpolating the inter-peak segments of the first plurality of gait samples to a preset length;

determining if the interpolated inter-peak segments of the first plurality of gait samples matches the initial user model;

authenticating the user of the mobile device when the interpolated inter-peak segments of the first plurality of gait samples match the initial user model; and

after authenticating the user of the mobile device;

collecting a current plurality of gait samples;

determining a peak for the current plurality of gait samples;

extracting inter-peak segments for the current plurality of gait samples;

interpolating the inter-peak segments of the current plurality of gait samples to a preset length;

determining if the interpolated inter-peak segments of the current plurality of gait samples matches the current user model;

generating an updated user model based on the interpolated inter-peak segments of the current plurality of gait samples when the interpolated inter-peak segments of the current plurality of gait samples do not match the current user model.

23. The method of claim 22, wherein the determining if the inter-peak segments match the current user model comprises:

detecting a plurality of cluster patterns in the plurality of gait samples;

determining one of a plurality of centroids for each of the plurality of cluster patterns;

comparing the plurality of centroids with a plurality of user model centroids to determine a match score; and

determining the inter-peak segments match the user model when the match score exceeds a threshold.

24. The method of claim 23, wherein the match score is determined based on an average minimum distance between each of the plurality of centroids and each of the plurality of user model centroids.

25. The method of claim 22, further comprising, when the interpolated inter-peak segments of the current plurality of gait samples do not match the current user model:

generating a fused user model based on the updated user model and the current user model;

notifying the user of the mobile device of the fused user model;

requesting permission from the user of the mobile device to update the current user model;

updating the current user model with the fused user model when the requested permission is received.

26. The method of claim 25, wherein the fused user model comprises a cluster of the updated user model that is within a pre-defined distance of a cluster in the current user model and a cluster of the updated user model that is outside the pre-defined distance.

27. The method of claim 26, further comprising:

determining if a total number of clusters in the current user model exceeds a threshold; and

removing an oldest cluster from the current user model when the total number of clusters in the current user model exceeds the threshold.

28. The method of claim 22, wherein the memory is a ring buffer incorporated into the mobile device.

29. The method of claim 22, wherein the mobile device is incorporated into a device selected from the group consisting of a music player, a video player, an entertainment unit, a navigation device, a communications device, a mobile phone, a smartphone, a personal digital assistant, a tablet computer, a computer, a wearable device, and a laptop computer.