Spoofed face detection

Patent metadata

Field	Value
Publication number	US-10452935-B2
Application number	US-201515763619-A
Country	US
Kind code	B2
Filing date	Oct 30, 2015
Priority date	Oct 30, 2015
Publication date	Oct 22, 2019
Grant date	Oct 22, 2019

Abstract

Examples are disclosed herein that relate to detecting spoofed human faces. One example provides a computing device comprising a processor configured to compute a first feature distance between registered image data of a human face in a first spectral region and test image data of the human face in the first spectral region, compute a second feature distance between the registered image data and test image data of the human face in a second spectral region, compute a test feature distance between the test image data in the first spectral region and the test image data in the second spectral region, determine, based on a predetermined relationship, whether the human face to which the test image data in the first and second spectral regions corresponds is a real human face or a spoofed human face, and modify a behavior of the computing device.

First claim

The invention claimed is:

1. A computing device, comprising:
a processor configured to:
compute a first feature distance between registered image data of a human face in a first spectral region and test image data of the human face in the first spectral region;
compute a second feature distance between the registered image data of the human face in the first spectral region and test image data of the human face in a second spectral region;
compute a test feature distance between the test image data of the human face in the first spectral region and the test image data of the human face in the second spectral region;
determine, based on a predetermined relationship differentiating between real human face image data and spoofed human face image data, whether the human face to which the test image data in the first and second spectral regions corresponds is a real human face or a spoofed human face; and
modify a behavior of the computing device based on whether the human face is determined to be the real human face or the spoofed human face.

2. The computing device of claim 1 , wherein the predetermined relationship is determined by training a classifier to differentiate between real human face image data and spoofed human face image data classes in a two-dimensional feature distance space.

3. The computing device of claim 2 , wherein a first axis of the two-dimensional feature distance space plots a range of test feature distances, and
wherein a second axis of the two-dimensional feature space plots a range of differences between corresponding pairs of first and second feature distances.

4. The computing device of claim 2 , wherein the predetermined relationship compares a linear combination of the first, second, and test feature distances to a threshold, and
wherein the classifier is configured to determine that the human face is a real human face when the comparison exceeds the threshold, and to determine that the human face is a spoofed human face when the comparison falls below the threshold.

5. The computing device of claim 1 , wherein each feature distance is computed as a likelihood ratio of a probability of a respective pair of feature vectors belonging to the real human face and a probability of the respective pair of feature vectors belonging to the spoofed human face.

6. The computing device of claim 5 , wherein the feature vectors are computed based on local binary pattern analysis.

7. The computing device of claim 1 , wherein the first spectral region is an infrared spectral region, and
wherein the second spectral region is a visible spectral region.

8. The computing device of claim 1 , wherein it is determined that the human face to which the test image data in the first and second spectral regions corresponds is the real human face when the test feature distance exceeds a threshold.

9. The computing device of claim 1 , wherein it is determined that the human face to which the test image data in the first and second spectral regions corresponds is the real human face when the second feature distance exceeds the first feature distance.

10. The computing device of claim 1 , wherein computing the first, second, and test feature distances includes computing a respective pair of feature vectors, each feature vector encoding, for corresponding image data, one or more of a histogram of texture values, an edge, a point, a shape, and a brightness.

11. A method of biometric verification, comprising:
computing a first feature distance between registered image data of a human face in a first spectral region and test image data of the human face in the first spectral region;
computing a second feature distance between the registered image data of the human face in the first spectral region and test image data of the human face in a second spectral region;
computing a test feature distance between the test image data of the human face in the first spectral region and the test image data of the human face in the second spectral region;
determining, based on a predetermined relationship differentiating between real human image data and spoofed human image data, whether the human face to which the test image data in the first and second spectral regions corresponds is a real human face or a spoofed human face; and
modifying a behavior of a computing device based on whether the human face is determined to be the real human face or the spoofed human face.

12. The method of claim 11 , wherein the predetermined relationship is determined by training a classifier to differentiate between real human face image data and spoofed human face image data classes in a two-dimensional feature distance space.

13. The method of claim 12 , wherein a first axis of the two-dimensional feature distance space plots a range of test feature distances, and
wherein a second axis of the two-dimensional feature space plots a range of differences between corresponding pairs of first and second feature distances.

14. The method of claim 12 , wherein the predetermined relationship compares a linear combination of the first, second, and test feature distances to a threshold, and
wherein the classifier is configured to determine that the human face is a real human face when the comparison exceeds the threshold, and to determine that the human face is a spoofed human face when the comparison falls below the threshold.

15. The method of claim 11 , wherein each feature distance is computed as a likelihood ratio of a probability of a respective pair of feature vectors belonging to the real human face and a probability of the respective pair of feature vectors belonging to the spoofed human face.

16. The method of claim 15 , wherein the feature vectors are computed based on local binary pattern analysis.

17. The method of claim 11 , wherein the first spectral region is an infrared spectral region, and
wherein the second spectral region is a visible spectral region.

18. The method of claim 11 , wherein it is determined that the human face to which the test image data in the first and second spectral regions corresponds is the real human face when the test feature distance exceeds a threshold.

19. The method of claim 11 , wherein it is determined that the human face to which the test image data in the first and second spectral regions corresponds is the real human face when the second feature distance exceeds the first feature distance.

20. A computing device, comprising:
a processor; and
memory holding instructions executable by the processor to:
receive registered image data of a human face in a first spectral region;
receive test image data of the human face in the first spectral region;
receive test image data of the human face in a second spectral region;
for each of the registered image data, the test image data in the first spectral region, and the test image data in the second spectral region:
identify a plurality of regions of interest;
extract at least one feature from each of the plurality of regions of interest; and
assemble the extracted features into a feature vector;

compute a first feature distance between the feature vector of the registered image data and the feature vector of the test image data in the first spectral region;
compute a second feature distance between the feature vector of the registered image data and the feature vector of the test image data in the second spectral region;
compute a test feature distance between the feature vector of the test image data in the first spectral region and the

Assignees

• Microsoft Technology Licensing Llc

Inventors

• Li Jinyu
• Wen Fang
• Wei Yichen
• Conrad Michael John
• Chu Chun-Te
• Jawaid Aamir

Classifications

• G06V10/764

  using classification, e.g. of video objects · CPC title

• G06V10/50 Primary

  by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis · CPC title

• G06V40/40

  Spoof detection, e.g. liveness detection · CPC title

• G06F18/2411

  based on the proximity to a decision surface, e.g. support vector machines · CPC title

• G06F18/21

  Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation · CPC title