Robust Human Authentication Using Holistic Anthropometric and Appearance-Based Features and Boosting

• US 20110135165A1
• Filed: 06/01/2010
• Published: 06/09/2011
• Est. Priority Date: 06/02/2009
• Status: Active Grant

• Current Assignee: George Mason Intellectual Properties, Inc.
• Original Assignee: George Mason Intellectual Properties, Inc.
• Inventors: Ramanathan, Venkatesh, Wechsler, Harry
• Granted Patent: US 8,379,940 B2
• Time in Patent Office: Days
• US Class Current: 382/118
• CPC Class Codes: G06F 18/211 Selection of the most significant subset of features
• CPC Class Codes: G06F 18/2135 based on approximation criteria, e.g. principal component analysis
• CPC Class Codes: G06F 18/214 Generating training patterns; Bootstrap methods, e.g. bagging or boosting
• CPC Class Codes: G06F 18/256 of results relating to different input data, e.g. multimodal recognition
• CPC Class Codes: G06V 10/771 Feature selection, e.g. selecting representative features from a multi-dimensional feature space
• CPC Class Codes: G06V 10/7715 Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods
• CPC Class Codes: G06V 10/774 Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting

Abstract

A new robust human authentication system, device, and instructions, embeddable in a physical and tangible computer readable medium, for determining if at least one test image obtained using an imaging device matches at least one training image in an enrollment database, are disclosed. This invention applies the concepts of appearance (PCA or PCA+LDA) and holistic anthropometrics that include head, face, neck, and shoulder linear and non-linear geometric measurements. The appearance (“eigen”) coefficients and holistic anthropometric measurements selected may be used as feature vectors. A boosting algorithm ranks features as “weak learners” and combines their outputs for “strong” recognition.

Claims

1. A robust human authentication system for determining if at least one test image obtained using an imaging device matches at least one training image in an enrollment database comprising:

a. feature extraction modules comprising;

1) an appearance-based feature extraction module configured for extracting appearance-based features from the face of the test image;

2) a holistic anthropometric feature extraction module configured for extracting holistic anthropometric features from the following elements from the test image;

a) head;

b) face;

c) neck; and

d) at least one shoulder;

b. feature selector modules comprising;

1) an appearance-based feature selector module configured for selecting the appearance-based features by using energy based cut-off criteria; and

2) a holistic anthropometric feature selector module configured for selecting the holistic anthropometric features that are predictive of the identification of the test image; and

c. at least one data fusion combiner module.

2. The robust human authentication system according to claim 1, wherein the holistic anthropometric feature selector module uses a filter-based feature selection approach.

3. The robust human authentication system according to claim 2, wherein the filter-based feature selection approach is driven by the mutual information between the test image'"'"'s data and the test image'"'"'s label.

4. The robust human authentication system according to claim 3, wherein the filter-based feature selection approach is implemented by correlation-based feature selection.

5. The robust human authentication system according to claim 1, wherein the data fusion combiner module is a decision-level neural network combiner module configured for combining the appearance-based features with the holistic anthropometric features using neural networks.

6. The robust human authentication system according to claim 1, wherein the data fusion combiner module is a feature-level boosting combiner module configured for combining the appearance-based features with the holistic anthropometric features using boosting.

7. The robust human authentication system according to claim 1, further including an authentication module, configured for matching and authenticating results from the data fusion combiner with the training image.

8. The robust human authentication system according to claim 1, further including a training image preprocessor module configured for preprocessing at least one training image by:

a. converting 256 gray levels into floating points;

b. using geometric normalization that lines up chosen eye coordinates;

c. cropping the training image using an elliptical mask;

d. equalizing a histogram of the training image; and

e. normalizing pixel values to mean zero and variance of one.

9. A robust human authentication device for determining if at least one test image obtained using an imaging device matches at least one training image in an enrollment database comprising:

a. feature extraction modules comprising;

1) an appearance-based feature extraction module configured for extracting appearance-based features from the face of the test image;

2) a holistic anthropometric feature extraction module configured for extracting holistic anthropometric features from the following elements from the test image;

a) head;

b) face;

c) neck; and

d) at least one shoulder;

b. feature selector modules comprising;

1) an appearance-based feature selector module configured for selecting the appearance-based features by using energy based cut-off criteria; and

2) a holistic anthropometric feature selector module configured for selecting the holistic anthropometric features that are predictive of the identification of the test image; and

c. at least one data fusion combiner module.

10. The robust human authentication device according to claim 9, wherein the holistic anthropometric feature selector module uses a filter-based feature selection approach.

11. The robust human authentication device according to claim 10, wherein the filter-based feature selection approach is driven by the mutual information between the test image'"'"'s data and the test image'"'"'s label.

12. The robust human authentication device according to claim 11, wherein the filter-based feature selection approach is implemented by correlation-based feature selection.

13. The robust human authentication device according to claim 9, wherein the data fusion combiner module is a decision-level neural network combiner module configured for combining the appearance-based features with the holistic anthropometric features using neural networks.

14. The robust human authentication device according to claim 9, wherein the data fusion combiner module is a feature-level boosting combiner module configured for combining the appearance-based features with the holistic anthropometric features using boosting.

15. The robust human authentication device according to claim 9, further including an authentication module, configured for matching and authenticating results from the data fusion combiner with the training image.

16. The robust human authentication device according to claim 9, further including a training image preprocessor module configured for preprocessing at least one training image by:

a. converting 256 gray levels into floating points;

b. using geometric normalization that lines up chosen eye coordinates;

c. cropping the training image using an elliptical mask;

d. equalizing a histogram of the training image; and

e. normalizing pixel values to mean zero and variance of one.

17. A physical and tangible computer readable medium encoded with instructions for determining if at least one test image obtained using an imaging device matches at least one training image in an enrollment database, wherein execution of the instructions by one or more processors causes the one or more processors to perform the steps comprising:

a. extracting

1) appearance-based features from the face of the test image;

2) holistic anthropometric features from the following elements from the test image;

a) head;

b) face;

c) neck; and

d) at least one shoulder;

b. selecting

1) the appearance-based features by using energy based cut-off criteria; and

2) the holistic anthropometric features that are predictive of the identification of the test image by using a filter-based feature selection approach; and

c. combining the appearance-based features with the holistic anthropometric features.

18. The physical and tangible computer readable medium according to claim 17, wherein the appearance-based features are combined with the holistic anthropometric features using neural networks.

19. The physical and tangible computer readable medium according to claim 17, wherein the appearance-based features are combined with the holistic anthropometric features using boosting.

20. The physical and tangible computer readable medium according to claim 17, further including matching and authenticating the results of combining the appearance-based features with the holistic anthropometric features, with the training image.