System and methods for transforming biometric image data to a consistent angle of inclination

• US 7,916,907 B2
• Filed: 11/30/2009
• Issued: 03/29/2011
• Est. Priority Date: 06/14/2004
• Status: Expired due to Fees

• Current Assignee: Biocrypt Access, LLC
• Original Assignee: Mark A. Kelty, Rodney Beatson
• Inventors: Kelty, Mark A., Beatson, Rodney
• Primary Examiner(s): Bali, Vikkram
• Assistant Examiner(s): Brooks, Julian D
• Application Number:
  US12/627,413
• Publication Number:
  US 20100142763A1
• Time in Patent Office: 484 Days
• Field of Search: None
• US Class Current: 382/119
• CPC Class Codes: G06V 30/1478 of characters or characters lines
• CPC Class Codes: G06V 40/30 Writer recognition; Reading and verifying signatures

Abstract

Biometric data are obtained through a biometric input device (120) and subsequently pixelated via a pixelator (130). The pixelator (130) creates an image of the biometric data. Via a processing unit (110), a relationship between pixels of a transformed version of the image is asserted. Thus, the biometric data is rotated to a consistent inclination based on the relationship between pixels regardless of an orientation in which the biometric data were captured in the original image. Once the image has been transformed, features of the biometric data may be extracted and either stored in a data storage unit (140) or compared with previously stored feature values for validation of the biometric data.

Claims

1. A method for spatially transforming biometric data for invariant feature extraction comprising the steps of:

using a processor to perform the following functions;

acquiring and digitizing biometric information from a user and generating the biometric data responsive thereto;

pixelating the biometric data to form a biometric image thereof, said image including a plurality of pixels at a corresponding plurality of pixel coordinates of an orthogonal coordinate system, said biometric data oriented in said image at an arbitrary submission angle with respect to a predetermined axis of said image;

applying a coordinate transformation to said plurality of pixels of said biometric image to form a transformed biometric image, said transformed biometric image remaining substantially invariant for different submissions of said biometric information at variant submission angles, said transformed biometric image being defined according to one of the following characteristics;

var(x_i)=k₁*var(y_i), 

 

1)

where “

var”

is a variance, k₁ is a scalar multiple, and x_i, y_i represent a plurality of transformed pixel coordinates, such that;

x_i=(X_i+M₁Y_i)/(1+M₁²)^1/2

and

y_i=(Y_i−

M₁X_i)/(1+M₁²)^1/2

wherein X_i, Y_i represent said plurality of pixel coordinates of said biometric image; and

,

wherein for k₁=1;

M₁=[2ra+or−

Sqrt(4r²a²+a⁴−

2a²+1)]/(a²−

1), if a≠

1 and,

M₁=0, if a=1

where a=sqrt[var(X_i)/var(Y_i)]; and

, r is a correlation coefficient of X_i and Y_i;

cov(x_i,y_i)=k₂ 

 

2)

where “

cov”

is a covariance, k₂ is a scalar multiple, and x_i, y_i represent a plurality of transformed pixel coordinates such that;

x_i=(X_i+M₂Y_i)/(1+M₂²)^1/2

and

y_i=(Y_i−

M₂X_i)/(1+M₂²)^1/2

wherein X_i, Y_i represent said plurality of pixel coordinates of said biometric image; and

wherein for k₂=0;

M₂=[(1−

a²)+or−

Sqrt(4r²a²+a⁴−

2a²+1)]/2ra, if r≠

0

and,

M₂=0, if r=0

where a=sqrt[var(X_i)/var(Y_i)]; and

, r is a correlation coefficient of X_i and Yi.

2. The method for spatially transforming biometric data for invariant feature extraction as recited in claim 1, wherein said plurality of pixel coordinates and said plurality of transformed pixel coordinates correspond to Cartesian coordinates of said orthogonal coordinate system.

3. The method for spatially transforming biometric data for invariant feature extraction as recited in claim 1, wherein k₂ equals zero.

4. The method for spatially transforming biometric data for invariant feature extraction as recited in claim 1 further including the steps of:

fitting a line to said pixelated biometric data of said transformed biometric image; and

rotating said transformed biometric image so that said fitted line is oriented in said image at a predetermined angle with respect to a predetermined axis of said transformed biometric image.

5. The method for spatially transforming biometric data for invariant feature extraction as recited in claim 4 where said line fitting step includes the step of fitting said line using a regression procedure.

6. The method for spatially transforming biometric data for invariant feature extraction as recited in claim 5 where said line fitting step includes the step of establishing a least square estimation as said regression procedure;

said estimation including minimizing a summation of squared perpendicular distances from the pixilated biometric data of said transformed biometric image, to said line.

7. The method for spatially transforming biometric data for invariant feature extraction as recited in claim 1, where said acquiring step includes the step of providing a handwriting sample as said biometric information.

8. A method for verifying the validity of biometric data by invariant feature extraction, the method comprising the steps of:

using a processor to perform the following functions;

providing an input device for acquiring and digitizing biometric information from a user;

providing a storage unit for storing features of said biometric information;

acquiring a first sample of said biometric information from said user and generating a first sample of biometric data responsive thereto;

pixelating said first sample of biometric data to form a first sample biometric image thereof, said first sample biometric image including a plurality of pixels at a corresponding plurality of pixel coordinates of an orthogonal coordinate system, said first sample of biometric data oriented in said first sample biometric image at an arbitrary first submission angle with respect to a predetermined axis of said first sample biometric image;

applying a coordinate transformation to said plurality of pixels of first sample biometric image to form a first transformed biometric image, said first transformed biometric image remaining substantially invariant for different submissions of said biometric information at variant submission angles, said transformed biometric image being defined according to one of the following characteristics;

var(x_i)=k₁*var(y_i), 

 

1)

where “

var”

is a variance, k₁ is a scalar multiple, and x_i, y_i represent a plurality of transformed pixel coordinates, such that;

x_i=(X_i+M₁Y_i)/(1+M₁²)^1/2

and

y_i=(Y_i−

M₁X_i)/(1+M₁²)^1/2

wherein X_i, Y_i represent said plurality of pixel coordinates of said biometric image; and

,

wherein for k₁=1;

M₁=[2ra+or−

Sqrt(4r²a²+a⁴−

2a²+1)]/(a²−

1), if a≠

1 and,

M₁=0, if a=1

where a=sqrt[var(X_i)/var(Y_i)]; and

, r is a correlation coefficient of X_i and Yi;

cov(x_i,y_i)=k₂ 

 

2)

where “

cov”

is a covariance, k₂ is a scalar multiple, and x_i, y_i represent a plurality of transformed pixel coordinates such that;

x_i=(X_i+M₂Y_i)/(1+M₂²)^1/2

and

y_i=(Y_i−

M₂X_i)/(1+M₂²)^1/2

wherein X_i,Y_i represent said plurality of pixel coordinates of said biometric image; and

wherein for k₂=0;

M₂=[(1−

a²)+or−

Sqrt(4r²a²+a⁴−

2a²+1)]/2ra, if r≠

0

and,

M₂=0, if r=0

where a=sqrt[var(X_i)/var(Y_i)]; and

, r is a correlation coefficient of X_i and Yi;

extracting biometric features from said first transformed biometric image and storing said features in said storage unit;

acquiring a second sample of said biometric information from said user and generating a second sample of biometric data responsive thereto;

pixelating said second sample of biometric data to form a second sample biometric image thererof, said second sample biometric image including a plurality of pixels at a corresponding plurality of pixel coordinates of said orthogonal coordinate system, said second sample of biometric data oriented in said second sample biometric image at an arbitrary second submission angle with respect to a predetermined axis of said second sample biometric image;

applying said transformation to said plurality of pixels of said second sample biometric image to form a second transformed biometric image, said second transformed biometric remaining substantially invariant for different submissions of said biometric information at variant submission angles angle of;

extracting said biometric features from said second transformed biometric image; and

comparing said biometric features of said second transformed biometric image to corresponding ones of said features stored in said storage unit.

9. The method for verifying the validity of biometric data by invariant feature extraction as recited in claim 8, wherein said plurality of pixel coordinates and said plurality of said transformed pixel coordinates correspond to Cartesian axes of said orthogonal coordinate system.

10. The method for verifying the validity of biometric data by invariant feature extraction as recited in claim 8, wherein k₂ equals zero.

11. The method for verifying the validity of biometric data by invariant feature extraction as recited in claim 8 further including the steps of:

fitting a line to each of said pixelated first biometric data sample and said pixelated second biometric data sample of respectively said first transformed biometric image and second transformed biometric image;

rotating said first transformed biometric image so that said respective fitted line is oriented in said first transformed biometric image at an angle with respect to a predetermined axis thereof prior to said first transformed biometric image feature extraction step; and

rotating said second transformed biometric image so that said respective fitted line is oriented in said second transformed biometric image at an angle with respect to a predetermined axis thereof prior to said second transformed biometric data image feature extraction step.

12. The method for verifying the validity of biometric data by invariant feature extraction as recited in claim 11 where said line fitting step includes the step of fitting each said respective line using a regression procedure.

13. The method for verifying the validity of biometric data by invariant feature extraction as recited in claim 12 where said line fitting step includes the step of establishing a least square estimation as said regression procedure;

said estimation including minimizing a summation of squared perpendicular distances from the pixilated first and second sample of biometric data to said line.

14. The method for verifying the validity of biometric data by invariant feature extraction as recited in claim 8, where said first sample acquiring step and said second sample acquiring step each includes the step of providing a handwriting sample as said first sample and said second sample, respectively.

15. A system for spatially transforming biometric data for invariant feature extraction, the system comprising:

an input device operable to acquire and digitize biometric information data from a user and generate the biometric data responsive thereto;

a pixelator operable to pixelate said biometric data into an image thereof, said image including a plurality of pixels at a corresponding plurality of pixel coordinates in an orthogonal coordinate system, said biometric data oriented in said image at an arbitrary submission angle with respect to a predetermined axis of said image;

a storage unit operable to store sequences of computer program instructions for a transformation process executable to transform said plurality of pixels of said image into a transformed image, said transformation process involving the solution of a quadratic equation where the equation involves the ratio of variances of the pixel coordinates and a correlation coefficient of the pixel coordinates, said transformed image remaining substantially invariant for different submissions of said biometric information at a plurality of submission angles; and

a processing unit coupled to said input device and said storage unit, said processing unit operable to execute said transformation process.

16. The system for spatially transforming biometric data for invariant feature extraction as recited in claim 15, wherein said sequences of computer program instructions executable by the processing unit include:

a line fitting process for fitting a line to said pixelated biometric data of said transformed biometric image; and

an image rotation process for rotating said transformed biometric image so that said fitted line is oriented in said image at an angle with respect to a predetermined axis of said rotated transformed biometric image.