Systems and methods for performing fingerprint based user authentication using imagery captured using mobile devices
Inventors
Othman, Asem • Tyson, Richard • Tavanai, Aryana • Xue, Yiqun • Simpson, Andrew
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-11263432-B2
Publication Date
2022-03-01
Expiration Date
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.
Core Innovation
The invention provides a method for performing fingerprint recognition using a mobile device with a camera that captures images depicting a plurality of fingers of a subject at a distance from the mobile device, such that the plurality of fingers do not contact the mobile device during capturing. The method identifies, in at least one captured image, respective fingertip segments for each finger among the plurality of fingers using one or more trained classifiers selected from the group consisting of a HOG classifier, an LBP classifier, and a Haar classifier.
After fingertip segmentation, the invention extracts features from the identified respective fingertip segments for one or more fingers. The method then generates a biometric identifier including the extracted features and stores the generated biometric identifier in memory. The workflow supports using extracted fingerprint features derived from distance-captured finger imagery to form a biometric identifier for fingerprint recognition.
Dependent claim content further refines the invention by extracting fingerprint minutia from a fingertip segment using a convolutional neural network (CNN), calculating quality scores for detected minutia points to identify and filter chains of minutia points based on measured versus expected characteristics for falsely detected minutia points, and matching a generated biometric identifier to a previously stored biometric identifier using a CNN. Additional refinements include determining liveness using either a three-dimensional representation derived from aligned images captured at different positions or a CNN trained on live-finger versus spoofed-finger image classes.
Claims Coverage
The independent claim contains one core method with three main inventive functional elements, supported by dependent claims that refine minutia extraction, matching, and liveness determination.
Distance-based multi-finger capture on a mobile device
Capturing, by a mobile device having a camera, images depicting a plurality of fingers of a subject without the plurality of fingers contacting the mobile device, such that the plurality of fingers are at a distance from the camera of the mobile device during the capturing of the images.
Fingertip segment identification using trained classifiers
Identifying in at least one of the captured images, with the processor using one or more trained classifiers, a respective fingertip segment for each finger among the plurality of fingers, wherein the trained classifiers are selected from the group consisting of a HOG classifier, an LBP classifier, and a Haar classifier.
Feature extraction and biometric identifier generation from fingertip segments
Extracting, with the processor from the identified respective fingertip segment for one or more fingers, features of the one or more fingers; generating, with the processor, a biometric identifier including the extracted features; and storing the generated biometric identifier in the memory.
CNN-based minutia extraction with quality scoring and chain filtering
Extracting fingerprint minutia from a fingertip segment using a convolutional neural network (CNN), calculating quality scores for a set of minutia points, identifying chains of minutia points based on measured versus expected characteristics for falsely detected minutia points, filtering the identified chains of minutia points, and selecting at least a subset of the minutia points for inclusion in the biometric identifier based on the quality scores.
CNN-based matching of generated and stored biometric identifiers
Matching a generated biometric identifier to a previously stored biometric identifier using a processor and a convolutional neural network (CNN).
3D representation-based liveness determination from multi-position images
Capturing first and second camera images of a subject’s plurality of fingers at different positions relative to the camera, aligning the images, generating a three-dimensional representation of at least one finger from the aligned images, and determining the subject’s liveness based on the three-dimensional representation.
CNN liveness determination using live versus spoof image classes
Determining a subject’s liveness by processing an identified fingertip segment image with a convolutional neural network (CNN) trained on live-finger and spoofed-finger image classes.
Overall, the claims cover distance-based multi-finger capture on a mobile device, fingertip segment identification using trained classifiers selected from HOG, LBP, and Haar, and generating and storing a biometric identifier from extracted finger features. Dependent claim refinements further add CNN-based minutia extraction with quality scoring and chain filtering, CNN-based matching to a stored identifier, and liveness determination using either multi-position 3D representation or a CNN trained on live versus spoof image classes.
Stated Advantages
Not explicitly described in patent.
Documented Applications
Not explicitly described in patent.
Interested in licensing this patent?