SYSTEMS AND METHODS FOR PERFORMING FINGERPRINT BASED USER AUTHENTICATION USING IMAGERY CAPTURED USING MOBILE DEVICES

• US 20180018501A1
• Filed: 09/14/2017
• Published: 01/18/2018
• Est. Priority Date: 02/06/2015
• Status: Active Grant

• Current Assignee: Veridium IP Limited (Harvard Investments LP)
• Original Assignee: Veridium IP Limited (Harvard Investments LP)
• Inventors: Mather, Jonathan Francis, Othman, Asem, Tyson, Richard, Simpson, Andrew
• Granted Patent: US 10,521,643 B2
• Time in Patent Office: Days
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06V 40/12 Fingerprints or palmprints
• CPC Class Codes: G06V 40/1341 Sensing with light passing through the finger
• CPC Class Codes: G06V 40/1347 Preprocessing; Feature extraction
• CPC Class Codes: G06V 40/1353 Extracting features related to minutiae or pores
• CPC Class Codes: G06V 40/1359 Extracting features related to ridge properties; Determining the fingerprint type, e.g. whorl or loop
• CPC Class Codes: G06V 40/1365 Matching; Classification
• CPC Class Codes: G06V 40/1371 Matching features related to minutiae or pores
• CPC Class Codes: G06V 40/1376 Matching features related to ridge properties or fingerprint texture
• CPC Class Codes: G06V 40/1382 Detecting the live character of the finger, i.e. distinguishing from a fake or cadaver finger
• CPC Class Codes: G06V 40/1388 using image processing
• CPC Class Codes: G06V 40/1394 using acquisition arrangements

Abstract

Technologies are presented herein in support of a system and method for performing fingerprint recognition. Embodiments of the present invention concern a system and method for capturing a user'"'"'s biometric features and generating an identifier characterizing the user'"'"'s biometric features using a mobile device such as a smartphone. The biometric identifier is generated using imagery captured of a plurality of fingers of a user for the purposes of authenticating/identifying the user according to the captured biometrics and determining the user'"'"'s liveness. The present disclosure also describes additional techniques for preventing erroneous authentication caused by spoofing. In some examples, the anti-spoofing techniques may include capturing one or more images of a user'"'"'s fingers and analyzing the captured images for indications of liveness.

Claims

1. A method for performing fingerprint recognition, the method comprising:

capturing, by a mobile device having a camera, a storage medium, instructions stored on the storage medium, and a processor configured by executing the instructions, images depicting a plurality of fingers of a subject;

detecting, with the processor using a finger detection algorithm, the plurality of fingers depicted in one or more of the images;

identifying, with the processor from one or more of the images according to a segmentation algorithm, a respective fingertip segment for each finger among the plurality of fingers;

extracting, with the processor for each finger, discriminatory features from the respective fingertip segment; and

generating a biometric identifier including the extracted discriminatory features;

storing the generated biometric identifier in the memory with the processor.

2. The method of claim 1, wherein the step of detecting the plurality of fingers comprises sequentially applying a plurality of finger detection algorithms.

3. The method of claim 2, wherein the order in which the plurality of finger detection algorithms is applied is determined by the processor based on one or more of:

measured environmental conditions, hardware capabilities of the mobile device, and one more expected characteristics of the subject.

4. The method of claim 1, wherein the step of detecting the plurality of fingers comprises:

selecting a first finger detection algorithm among a plurality of finger detection algorithms;

detecting, by the processor using the first finger detection algorithm, at least one candidate set of fingers depicted in the one or more images;

computing a quality metric representing a likelihood that the at least one candidate set of fingers represents the plurality of fingers; and

based on the quality not meeting a prescribed threshold, selecting another finger detection algorithm and repeating the preceding detecting and computing steps.

5. The method of claim 4, wherein the other finger detection algorithm is a classifier that is trained differently from the first finger detection algorithm.

6. The method of claim 2, wherein the step of detecting the plurality of fingers further comprises:

measuring, with the processor using one or more sensors, ambient lighting conditions;

identifying, with the processor, a first finger detection algorithm from among the plurality of finger detection algorithms based on the measured ambient lighting conditions; and

applying, with the processor, the first finger detection algorithm to one or more of the images to detect the plurality of fingers.

7. The method of claim 6, wherein the plurality of finger detection algorithms include cascade classifiers trained according to respective ambient lighting levels.

8. The method of claim 6, wherein ambient lighting conditions are measured from one or more of:

a light level measured by a lux meter on board the mobile device and estimated from exposure settings of the camera.

9. The method of claim 1, further comprising:

determining, with the processor from one or more of the images, that the plurality of fingers within the one or more images are positionally stable.

10. The method of claim 9, wherein determining that the fingers are positionally stable within one or more of the images comprises:

determining, with the processor according to an object tracking algorithm, a respective position of one or more of the fingers in a first image and a respective change in position of the one or more fingers throughout a sequence of subsequent images among the images;

calculating a speed of movement of the one or more fingers based on the respective change in position; and

verifying that the plurality of fingers are positionally stable based on the speed falling below a threshold velocity.

11. The method of claim 10, wherein the step of determining according to the object tracking algorithm comprises:

dynamically switching between a first object tracking algorithm and a second object tracking algorithm based on the determined position and change in position.

12. The method of claim 11, further comprising:

applying a first object detection algorithm to determine the position of the one or more fingers in the first image;

applying, based on the identified first position, a second object tracking algorithm to determine a respective position of the one or more fingers in a subsequent image in the sequence; and

re-applying the first object detection algorithm to determine the respective position of the one or more fingers in the subsequent image upon detecting a failure of the second object tracking algorithm.

13. The method of claim 12, wherein the failure is detected as a function of the change in position of the one or more fingers between images in the sequence exceeding a threshold.

14. The method of claim 12, wherein the first object detection algorithm is a classifier, and wherein the second object tracking algorithm is one or more of a template matching algorithm and an optical flow tracking algorithm.

15. The method of claim 1, further comprising:

computing an adaptive skin model based on one or more of the images; and

processing the one or more images based on the adaptive skin model.

16. The method of claim 15, wherein computing an adaptive skin model, comprises:

detecting a region within the one or more images including pixels having color values that correspond to an expected skin tone;

computing a color model for the region based on the pixels;

fitting, with the processor, a probability density function (PDF) to distributions of hue and saturation for the pixels within the detected region;

locating, with the processor applying the PDF to the one or more images, additional pixels in the image meeting the PDF;

computing, for each located pixel, a p-value representing a likelihood that the pixel represents skin;

refine the computed color model based on the identified pixels having a p-value that exceeds a threshold p-value.

17. The method of claim 16, further comprising:

iteratively repeated the locating, computing and refining step until the number of pixels that define the refined model reaches convergence.

18. The method of claim 16, wherein processing according to the adaptive skim model comprises one or more of the following steps:

identifying the respective fingertip segment for each finger among the plurality is performed by applying a classifier to the adaptive skin model; and

enhancing the one or more images, with the processor, prior to identifying the respective fingertip segment, wherein enhancing includes brightening regions of the one or more images having pixels determined to represent skin relative to regions having pixels determined to not represent skin based on the p-values for the pixels in the adaptive color model.

19. The method of claim 1, wherein extracting discriminatory features for each of the respective fingertip segments comprises:

extracting from each of the respective fingertip segments, minutiae of a respective finger according to a minutiae extracting algorithm; and

for each of the fingers, encoding respective feature vectors representing the extracted minutia in the biometric identifier.

20. The method of claim 19, further comprising:

comparing, with the processor, each of the encoded feature vectors to a previously stored feature vector associated with the user;

computing similarity scores based on the comparison;

verifying that similarity scores individually computed for a prescribed number of fingers among the plurality of fingers exceed a secondary matching threshold, and that a combination of the computed similarity scores exceeds a primary matching threshold; and

determining that the subject is the user based on the verification.