System and method for rapid face recognition

• US 9,275,309 B2
• Filed: 08/01/2014
• Issued: 03/01/2016
• Est. Priority Date: 08/01/2014
• Status: Active Grant

• Current Assignee: TCL Research America, Inc. (TCL Technology Group Corp.)
• Original Assignee: TCL Research America, Inc. (TCL Technology Group Corp.)
• Inventors: Wang, Haohong, Iliadis, Michael
• Primary Examiner(s): Park, Chan
• Assistant Examiner(s): Kholdebarin, Iman K
• Application Number:
  US14/449,352
• Publication Number:
  US 20160034789A1
• Time in Patent Office: 578 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06F 18/28 Determining representative reference patterns, e.g. by averaging or distorting; Generating dictionaries
• CPC Class Codes: G06V 10/513 Sparse representations
• CPC Class Codes: G06V 10/772 Determining representative reference patterns, e.g. averaging or distorting patterns; Generating dictionaries
• CPC Class Codes: G06V 40/169 Holistic features and representations, i.e. based on the facial image taken as a whole
• CPC Class Codes: G06V 40/172 Classification, e.g. identification

Abstract

A face recognition method is provided to use sparse representation and regularized least squares-based classification on a computing device. The method includes obtaining an image to be recognized as a test sample y and a set of training images of certain subjects as training sample matrix T, obtaining a sparse representation of the test sample and the training samples including an initial estimation of a sparse vector a, and constructing a new face dictionary comprising training samples with non-zero corresponding coefficients in the sparse vector a for the initial estimation. The method also includes obtaining new coefficients by solving a regularized least squares problem based on the constructed new face dictionary, and determining a face identity of the test sample based on minimum class residual calculated by using the new coefficients.

Claims

1. A face recognition method using sparse representation and regularized least squares-based classification on a computing device, the method comprising:

obtaining an image to be recognized as a test sample y and a set of training images of certain subjects as training sample matrix T;

obtaining a sparse representation of the test sample and the training samples including an initial estimation of a sparse vector a;

constructing a new face dictionary comprising training samples with non-zero corresponding coefficients in the sparse vector a for the initial estimation;

obtaining new coefficients by solving a regularized least squares problem based on the constructed new face dictionary; and

determining a face identity of the test sample based on minimum class residual calculated by using the new coefficients.

2. The face recognition method according to claim 1, further including:

presenting the face identity of the test sample to a user of the computing device.

3. The face recognition method according to claim 1, further including:

determining whether to use a standard sparse coding optimization problem or to use an approximated sparse coding optimization problem to obtain the initial estimation of the sparse vector a,

wherein the standard sparse coding optimization problem uses an l₁ minimization algorithm and the approximated sparse coding optimization problem requires that a least squares problem is first solved and a threshold is used to suppress most values to zero.

4. The face recognition method according to claim 3, wherein:

the test sample y is represented as a sparse linear combination of samples in T as;

y=Ta+e, wherein eε

^d is dense noise and aε

ⁿ is the sparse vector with nonzero elements corresponding to few samples in T.

5. The face recognition method according to claim 4, wherein:

when the standard sparse coding optimization problem is used, the coefficients of the sparse vector a is estimated by solving the sparse coding optimization problem by

6. The face recognition method according to claim 4, wherein:

when the approximated sparse coding optimization problem is used, the coefficients of the sparse vector a is estimated by solving the approximated sparse coding optimization problem by

7. The face recognition method according to claim 1, wherein constructing the new face dictionary further includes:

provided the function ƒ

(a_i), where a_i is the segment of a associated with class i, be given as

8. The face recognition method according to claim 7, wherein obtaining the new coefficients further includes:

obtaining new estimation vector can be obtained by solving the regularized least squares (RLS) problem

9. The face recognition method according to claim 8, wherein:

the minimum class residual of each class i is calculated as e_i(y)=∥

y−

T_iƒ

_i∥

₂, wherein ƒ

_i is the coding coefficient vector associated with class i; and

the face identity of the test sample y is determined as

Identity(y)=argmin_i{e_i}.

10. The face recognition method according to claim 2, wherein presenting the face identity further includes:

alerting the user of one or more events based on the face identity.

11. A non-transitory computer-readable medium having computer program for, when being executed by a processor, performing a face recognition method using sparse representation and regularized least squares-based classification on a computing device, the method comprising:

obtaining an image to be recognized as a test sample y and a set of training images of certain subjects as training sample matrix T;

obtaining a sparse representation of the test sample and the training samples including an initial estimation of a sparse vector a;

constructing a new face dictionary comprising training samples with non-zero corresponding coefficients in the sparse vector a for the initial estimation;

obtaining new coefficients by solving a regularized least squares problem based on the constructed new face dictionary; and

determining a face identity of the test sample based on minimum class residual calculated by using the new coefficients.

12. The non-transitory computer-readable medium according to claim 11, the method further including:

presenting the face identity of the test sample to a user of the computing device.

13. The non-transitory computer-readable medium according to claim 11, the method further including:

determining whether to use a standard sparse coding optimization problem or to use an approximated sparse coding optimization problem to obtain the initial estimation of the sparse vector a,

wherein the standard sparse coding optimization problem uses an l₁ minimization algorithm and the approximated sparse coding optimization problem requires that a least squares problem is first solved and a threshold is used to suppress most values to zero.

14. The non-transitory computer-readable medium according to claim 13, wherein:

the test sample y is represented as a sparse linear combination of samples in T as;

y=Ta+e, wherein eε

^d is dense noise and aε

ⁿ is the sparse vector with nonzero elements corresponding to few samples in T.

15. The non-transitory computer-readable medium according to claim 14, wherein:

when the standard sparse coding optimization problem is used, the coefficients of the sparse vector a is estimated by solving the sparse coding optimization problem by

16. The non-transitory computer-readable medium according to claim 14, wherein:

when the approximated sparse coding optimization problem is used, the coefficients of the sparse vector a is estimated by solving the approximated sparse coding optimization problem by

17. The non-transitory computer-readable medium according to claim 11, wherein constructing the new face dictionary further includes:

provided the function ƒ

(a_i), where a_i is the segment of a associated with class i, be given as

18. The non-transitory computer-readable medium according to claim 17, wherein obtaining the new coefficients further includes:

obtaining new estimation vector can be obtained by solving the regularized least squares (RLS) problem

19. The non-transitory computer-readable medium according to claim 18, wherein:

the minimum class residual of each class i is calculated as e_i(y)=λ

y−

T_iƒ

_i∥

₂, wherein ƒ

_i is the coding coefficient vector associated with class i; and

the face identity of the test sample y is determined as

Identity(y)=argmin_i{e_i}.

20. The non-transitory computer-readable medium according to claim 12, wherein presenting the face identity further includes:

alerting the user of one or more events based on the face identity.