FACE LIVENESS DETECTION USING BACKGROUND/FOREGROUND MOTION ANALYSIS

• US 20190026544A1
• Filed: 02/08/2017
• Published: 01/24/2019
• Est. Priority Date: 02/09/2016
• Status: Active Grant

• Current Assignee: Aware Incorporated (Crowley Holdings Preferred LLC)
• Original Assignee: Aware Incorporated (Crowley Holdings Preferred LLC)
• Inventors: HUA, Fang, RIOPKA, Taras
• Granted Patent: US 10,990,808 B2
• Time in Patent Office: Days
• CPC Class Codes: A61B 5/0077 Devices for viewing the surface of the body, e.g. camera, magnifying lens
• CPC Class Codes: A61B 5/1128 using image analysis A61B5/1127 takes precedence
• CPC Class Codes: A61B 5/1176 Recognition of faces
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06T 2207/30201 Face
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/194 involving foreground-background segmentation
• CPC Class Codes: G06T 7/246 using feature-based methods, e.g. the tracking of corners or segments
• CPC Class Codes: G06V 10/26 Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
• CPC Class Codes: G06V 40/172 Classification, e.g. identification
• CPC Class Codes: G06V 40/45 Detection of the body part being alive

Abstract

Face recognition systems are vulnerable to the presentation and spoofed faces, which may be presented to face recognition systems, for example, by an unauthorized user seeking to gain access to a protected resource. A face liveness detection method that addresses this vulnerability utilizes motion analysis to compare the relative movement among three regions of interest in a facial image, and based upon that comparison to make a face liveness determination.

Claims

1. A method of determining face liveness, comprising:

receiving, at a processor of a face recognition system and from an image capture device of the face recognition system, a time-stamped frame sequence;

identifying corresponding pixels for each pair of sequential frames in the time-stamped frame sequence;

segmenting one of each pair of sequential frames in the time-stamped frame sequence into regions of interest;

calculating a motion feature for each region of interest of each pair of sequential frames in the time-stamped frame sequence;

generating a preliminary face-liveness determination for each pair of sequential frames in the time-stamped frame sequence, based on a comparison of the calculated motion features for each region of interest of the pair of sequential frames; and

making a final face-liveness determination based on the generated preliminary face liveness determinations.

2. The method of claim 1, wherein the time-stamped frame sequence comprise a series of photographs taken by the image capture device.

3. The method of claim 1, wherein the regions of interest comprise a face region, a near-face region, and a background region.

4. The method of claim 1, wherein the motion feature is calculated based on estimated pixel velocities.

5. The method of claim 1, wherein the final face-liveness determination is made by averaging the preliminary face liveness determinations.

6. The method of claim 1, wherein the final face-liveness determination matches a majority of preliminary face liveness determinations.

7. The method of claim 1, wherein the preliminary face-liveness determination for each pair of sequential frames in the time-stamped frame sequence is further based on liveness detection rules stored in a memory of the face recognition system.

8. The method of claim 7, wherein the liveness detection rules comprise rules configured to detect artificial patterns in the calculated motion features.

9. A method of training a classifier of a face recognition system, comprising:

receiving, at a processor of a face recognition system and from an image capture device associated with the face recognition system, a first plurality of time-stamped frame sequences having a live subject, and a second plurality of time-stamped frame sequences having a spoofed subject;

for each of the time-stamped frame sequences in the first and second pluralities of time-stamped frame sequences;

identifying corresponding pixels for each pair of sequential frames in the time-stamped frame sequence;

segmenting one of each pair of sequential frames in the time-stamped frame sequence into regions of interest; and

calculating a motion feature for each region of interest of each pair of sequential frames in the time-stamped frame sequence;

storing the calculated motion features for each time-stamped frame sequence from the first plurality of time-stamped frame sequences as positive data, and storing the calculated motion features for each time-stamped frame sequence from the second plurality of time-stamped frame sequences as negative data; and

generating training rules based on the positive and negative data.

10. The method of claim 9, wherein at least one of the time-stamped frame sequences comprises a series of photographs taken by the image capture device or a series of frames from a video recorded by the image capture device.

11. The method of claim 9, wherein the regions of interest comprise a face region, a near-face region, and a background region.

12. The method of claim 9, further comprising estimating a pixel velocity for each pixel within each region of interest, and further wherein the estimated pixel velocities are used to calculate the motion feature for each region of interest.

13. The method of claim 9, wherein the spoofed subject of at least one of the second plurality of time-stamped frame sequences is a printed photo spoof, a shaped board spoof, a printed paper spoof, a displayed static digital image spoof, or a smart phone replay video spoof.

14. The method of claim 9, wherein the spoofed subjects of the second plurality of time-stamped frame sequences comprise different types of spoofs.

15. The method of claim 9, wherein machine learning techniques are used to generate the training rules based on the positive and negative data.

16. A face recognition system comprising:

an image capture device;

a processor;

a face detection unit comprising a memory; and

a face matching unit;

wherein the memory stores instructions that, when executed by the processor, cause the processor to;

receive, from the image capture device, a time-stamped frame sequence;

identify corresponding pixels for each pair of sequential frames in the time-stamped frame sequence;

segment one of each pair of sequential frames in the time-stamped frame sequence into regions of interest;

calculate a motion feature for each region of interest of each pair of sequential frames in the time-stamped frame sequence;

generate a preliminary face-liveness determination for each pair of sequential frames in the time-stamped frame sequence, based on a comparison of the calculated motion features for the pair of sequential frames; and

make a final face-liveness determination based on the generated preliminary face liveness determinations.

17. The system of claim 16, wherein the memory stores additional instructions that, when executed by the processor, further cause the processor to estimate a pixel velocity for each pixel within each region of interest, and further wherein the estimated pixel velocities are used to calculate the motion feature for each region of interest.

18. The system of claim 16, wherein the regions of interest comprise a face region, a near-face region, and a background region.

19. The system of claim 16, wherein the memory stores additional instructions that, when executed by the processor, further cause the processor to provide a face associated with the time-stamped frame sequence to the face matching unit when the final face liveness determination indicates that the face is live.

20. The system of claim 16, wherein the face matching unit comprises a second memory, the second memory storing second instructions that, when executed by the processor, cause the processor to compare a face associated with the time-stamped frame sequence to a plurality of stored faces to identify a matching face.