System and method for distortion characterization in fingerprint and palm-print image sequences and using this distortion as a behavioral biometrics

• US 7,054,470 B2
• Filed: 09/03/2003
• Issued: 05/30/2006
• Est. Priority Date: 12/02/1999
• Status: Expired due to Term

• Current Assignee: Uniloc 2017 LLC (SoftBank Group Corp.)
• Original Assignee: International Business Machines Corporation
• Inventors: Ratha, Nalini K., Dorai, Chitra, Bolle, Rudolf Maarten
• Primary Examiner(s): AHMED, SAMIR ANWAR
• Application Number:
  US10/653,804
• Publication Number:
  US 20040042642A1
• Time in Patent Office: 1,000 Days
• Field of Search: 382/120, 382/122, 382/125, 382/173, 382/190, 382/191, 382/199, 382/201, 382/276, 382124-127, 382/115, 382/116, 382/166, 356/71, 345/173, 713/186
• US Class Current: 382/124
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06V 40/12 Fingerprints or palmprints
• CPC Class Codes: G06V 40/30 Writer recognition; Reading and verifying signatures
• CPC Class Codes: G07C 9/32 in combination with an identity check

Abstract

This invention uses a novel biometrics, called resultant fingerprints and palm-prints, for authentication. The novel biometrics are consecutive traditional print images where the subject physically changes the appearance of the print images by rotating or rotating and translating, or rotating, translating, and shearing the finger or palm. That is, it is a sequence of finger or palm-print images over a short interval of time where the images are modified according to the rotation or a combination of rotation and translation or a combination of rotation, translation, and shear. The rotational and translational and shear components of the motion in the sequence of print images are determined from the image-to-image flow. This flow is either computed from motion-compensation vectors of the sequence compressed in MPEG formats or directly from the uncompressed images. The global image-to-image flow is expressed in terms of an affine transformation, computed from the local flow in blocks around a non-moving central region. The rotational and translational components of this affine transformation are smoothed over a temporal neighborhood resulting in a function of time. This function of time is a behavioral biometrics which can be changed by the user when compromised. Matching of this function for authentication purposes is achieved very much as is done in legacy signature matching authentication systems where two temporal signals are compared.

Claims

1. A biometrics system comprising:

an acquisition device for acquiring and storing a sequence of discrete print images from a part of a hand moving during a time period;

a trajectory process that determines the position and orientation of the images of said part of the hand as a function of time during the time period; and

an estimator process that determines a distortion of the discrete print images as a function of time due to the change in position and orientation, wherein the estimator process determines distortion by determining at least a motion of an image pattern occurring in one or more blocks of at least two of the discrete print images.

2. A system, as in claim 1, where the part of the hand includes one or more of the following:

a fingerprint and a palm-print.

3. A system, as in claim 1, where the distortion is caused by one or more of the following:

rotation, translation, and shear.

4. A system, as in claim 1, where the motion is interframe motion and where the estimator process comprises the steps of:

determining one or more blocks of interframe motion between consecutive pairs of the images in the sequence;

determining a proportion of blocks with no motion to blocks with some motion;

using the proportion to select a set of candidate distorted images;

identifying a largest stationary and spatially contiguous block in each candidate distorted image in the set;

estimating a global affine transformation between every pair of candidate distorted images in the set about the stationary and contiguous block;

determining a curl and translation from the global affine transformation between every pair of candidate distorted images in the set; and

using the change of the curl over the time period to indicate the distortion.

5. A system, as in claim 4, where, when the change in curl over the time period is greater than a threshold, the distortion is caused by one or more of the following:

rotation, translation, and shear.

6. A system, as in claim 4, where the distortion is primarily translation when curl is within a second threshold of zero and the translation exceeds a third threshold.

7. A method for detecting the distortion of a fingerprint or palm-print, comprising the steps of:

acquiring and storing a sequence of discrete print images from a part of a hand moving during a time period;

determining the position and orientation of the images of said part of the hand as a function of time during the time period; and

determining a distortion of the discrete print images as a function of time due to the change in position and orientation, wherein the step of determining a distortion further comprises the step of determining at least a motion of an image pattern occurring in one or more blocks of at least two of the discrete print images.

8. The method of claim 7, where the part of the hand includes one or more of the following:

a fingerprint and a palm-print.

9. The method of claim 7, where the distortion is caused by one or more of the following:

rotation, translation, and shear.

10. A method for determining a distortion of a set of images as a function of time due to the change in position and orientation of a hand comprising the steps of:

determining one or more blocks of interframe motion between consecutive pairs of images in a sequence of images from a part of a hand moving during a time period;

determining a proportion of blocks with no motion to blocks with some motion;

using the proportion to select a set of candidate distorted images;

identifying a largest stationary and spatially contiguous block in each candidate distorted image in the set;

estimating a global affine transformation between every pair of candidate distorted images in the set about the stationary and contiguous block;

determining a curl and translation from the global affine transformation between every pair of candidate distorted images in the set; and

using the change of the curl over the time period to indicate the distortion.

11. The method of claim 10, where, when the change in curl over the time period is greater than a threshold, the distortion is caused by one or more of the following:

pure rotation, translation, and shear.

12. The method of claim 10, where the distortion is primarily translation when curl is within a second threshold of zero and the translation exceeds a third threshold.

13. A biometrics system comprising:

an acquisition device for acquiring and storing a sequence of discrete images from a part of a hand moving during a time period;

a trajectory process that determines the position and orientation of the images of said part of the hand as a function of time during the time period;

an estimator process that determines a distortion of the discrete images as a function of time due to the change in position and orientation, wherein the estimator process determines distortion by determining at least a motion of an image pattern occurring in one or more blocks of at least two of the discrete print images; and

identifying a person utilizing at least said determined distortion.

14. A biometrics system comprising:

an acquisition device for acquiring and storing a sequence of discrete images from a part of a hand moving during a time period;

a trajectory process that determines the position and orientation of the images of said part of the hand as a function of time during the time period;

an estimator process that determines a distortion of the discrete images as a function of time due to the change in position and orientation, wherein the estimator process determines distortion by determining at least a motion of an image pattern occurring in one or more blocks of at least two of the discrete print images; and

authenticating a person utilizing at least said determined distortion.