Age invariant face recognition using convolutional neural networks and set distances

Patent metadata

Field	Value
Publication number	US-10565433-B2
Application number	US-201815936525-A
Country	US
Kind code	B2
Filing date	Mar 27, 2018
Priority date	Mar 30, 2017
Publication date	Feb 18, 2020
Grant date	Feb 18, 2020

Abstract

Time lapse, characteristic of aging, is a complex process that affects the reliability and security of biometric face recognition systems. Systems and methods use deep learning, in general, and convolutional neural networks (CNN), in particular, for automatic rather than hand-crafted feature extraction for robust face recognition across time lapse. A CNN architecture using the VGG-Face deep (neural network) learning produces highly discriminative and interoperable features that are robust to aging variations even across a mix of biometric datasets. The features extracted show high inter-class and low intra-class variability leading to low generalization errors on aging datasets using ensembles of subspace discriminant classifiers.

First claim

The invention claimed is:

1. An age invariant face recognition system comprising:
an input device for acquiring a representation of a facial image;
a pre-trained convolutional neural network receiving said representation of the facial image and extracting compact, highly discriminative and interoperable feature descriptors;
a biometric gallery built during an enrollment process when biometric traits of all known subjects are extracted as feature descriptors and stored as sets of image templates in a database;
a classifier receiving said feature descriptors from the convolutional neural network and comparing said feature descriptors with feature descriptors of subjects enrolled in the biometric gallery using set similarity distances chosen by the classifier from the group consisting of minimum distance, directed Hausdorff distance, undirected Hausdorff distance, directed modified Hausdorff distance, and undirected modified Hausdorff distance; and
an output device generating, based on said comparing, a match or no match of face recognition subject to aging, the match being used for authentication or identification.

2. The age invariant face recognition system of claim 1 , wherein the input device includes image preprocessing functions including face detection and subsequent image normalization in terms of pose and image size.

3. The age invariant face recognition system of claim 1 , wherein the pre-trained convolutional neural network is a visual geometry group (VGG)-Face convolutional neural network composed of a sequence of convolutional, pool and fully-connected layers, activation of a first fully-connected layer producting feature descriptors which are used for both face identification and face verification.

4. The age invariant face recognition system of claim 1 , wherein the match of face recognition is used for authentication.

5. The age invariant face recognition system of claim 1 , wherein the match of face recognition is used for identification.

6. The age invariant face recognition system of claim 1 , wherein the set similarity distances chosen by the classifier are minimum distances.

7. The age invariant face recognition system of claim 6 , wherein the minimum set similarity distances are defined as follows:

h

m
⁢

⁢
i
⁢

⁢
n

⁡

(

A
,
B

)

=

min

a
∈
A

b
∈
B

⁢

d
⁡

(

a
,
b

)

.

8. The age invariant face recognition system of claim 1 , wherein the set similarity distances chosen by the classifier are directed Hausdorff distances.

9. The age invariant face recognition system of claim 8 , wherein the directed Hausdorff set similarity distances are defined as follows:

h
d

⁡

(

A
,
B

)

=

max

a
∈
A

⁢

{

min

b
∈
B

⁢

{

d
⁡

(

a
,
b

)

}

}

.

10. The age invariant face recognition system of claim 1 , wherein the set similarity distances chosen by the classifier are undirected Hausdorff distances.

11. The age invariant face recognition system of claim 10 , wherein the undirected Hausdorff set similarity distances are defined as follows:
h u ( A,B )=max( h d ( A,B ), h d ( B,A )).

12. The age invariant face recognition system of claim 1 , wherein the set similarity distances chosen by the classifier are directed modified Hausdorff distances.

13. The age invariant face recognition system of claim 12 , wherein the directed modified Hausdorff set similarity distances are defined as follows:

H

d
⁢

⁢
m

⁡

(

A
,
B

)

=

1

N
a

⁢

∑

a
∈
A

⁢

{

min

b
∈
B

⁢

{

Assignees

• Univ George Mason

Inventors

• Wechsler Harry
• El Khiyari Hachim

Classifications

• G06V10/82

  using neural networks · CPC title

• G06V10/764

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

• G06V40/172

  Classification, e.g. identification · CPC title

• G06F21/32

  using biometric data, e.g. fingerprints, iris scans or voiceprints · CPC title

• G06N3/02

  Neural networks · CPC title