Method and apparatus for verifying static signatures using dynamic information

• US 5,930,380 A
• Filed: 02/11/1997
• Issued: 07/27/1999
• Est. Priority Date: 02/11/1997
• Status: Expired due to Term

• Current Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Original Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Inventors: Wilfong, Gordon Thomas, Kashi, Ramanujan S., Nelson, Winston Lowell
• Primary Examiner(s): Boudreau, Leo H.
• Assistant Examiner(s): Werner, Brian P.
• Application Number:
  US08/798,981
• Time in Patent Office: 896 Days
• Field of Search: 382/119, 382/120, 382/123, 382/186, 382/197, 382/195, 382/202, 382/206, 382/215, 382/217, 382/218, 382/259, 382/275, 348/161, 434/155
• US Class Current: 382/119
• CPC Class Codes: G06V 40/30 Writer recognition; Reading and verifying signatures

Abstract

In a process of signature verification, an off-line test signature is compared to one or more on-line reference signatures. Dynamic data obtained from the reference signature or signatures may be used to improve the thinning of the test signature. Such data may also be used for segmenting the test signature and sequencing the resulting strokes. In some embodiments, the invention includes evaluating a dynamic error, based at least in part on such sequencing of the test signature. In some further embodiments, the invention includes producing a rendering of at least one on-line reference signature as a two-dimensional image, and evaluating a static error based, at least in part, on shape-matching between the rendered image and the test signature.

Claims

1. A method for verifying a test signature provided as a static two-dimensional pattern, comprising:

digitizing the test signature;

comparing the test signature to at least one template signature, thereby to obtain at least one error score; and

comparing at least one said error score to a threshold value;

CHARACTERIZED IN THAT the template signature is derived from real-time data obtained during the entry of at least one on-line reference signature, and the method further comprises;

a) providing a sequenced listing of discrete strokes of the template signature, said listing derived from said real-time data, said listing to be referred to as a "stroke ordering;

"

b) segmenting the test signature into discrete strokes; and

c) before or during the comparing step, sequencing said test signature strokes to correspond, at least in part, to the stroke ordering of the template signature, wherein;

the comparing step is carried out, at least in part, by treating the template signature and the test signature as one-pixel-wide curves each having at least a partial temporal ordering, said treatment leading to an error score D_DYNAMIC ;

the error score D_DYNAMIC is evaluated, at least in part, as a combination of error scores D_LOCAL and D_GLOBAL ;

the error score D_LOCAL is evaluated, at least in part, by matching sequenced strokes of the test signature to sequenced strokes of the template signature in accordance with the methods of dynamic programming; and

the error score D_GLOBAL is evaluated, at least in part, by calculating at least one global feature on the test signature and on the reference signature, and comparing resulting corresponding global feature values.

2. The method of claim 1, further comprising:

deriving at least one further template signature as a static two-dimensional rendering of at least one said on-line reference signature;

calculating at least one feature, to be referred to as a static feature, on the static template signature;

calculating at least one static feature on the test signature as a static two-dimensional pattern; and

comparing at least one static feature value of the static template signature with a corresponding static feature value of the test signature, thereby to obtain at least one further error score, to be referred to as a static error score D_STATIC.

3. A method for verifying a test signature provided as a static two-dimensional pattern, comprising:

digitizing the test signature;

comparing the test signature to at least one template signature, thereby to obtain at least one error score; and

comparing at least one said error score to a threshold value;

CHARACTERIZED IN THAT the template signature is derived from real-time data obtained during the entry of at least one on-line reference signature, and the method further comprises;

a) providing a sequenced listing of discrete strokes of the template signature, said listing derived from said real-time data, said listing to be referred to as a "stroke ordering;

"

b) segmenting the test signature into discrete strokes;

c) before or during the comparing step, sequencing said test signature strokes to correspond, at least in part, to the stroke ordering of the template signature;

d) deriving at least one further template signature as a static two-dimensional rendering of at least one said on-line reference signature;

e) calculating at least one feature, to be referred to as a static feature, on the static template signature;

f) calculating at least one static feature on the test signature as a static two-dimensional pattern; and

g) comparing at least one static feature value of the static template signature with a corresponding static feature value of the test signature, thereby to obtain at least one further error score, to be referred to as a static error score D_STATIC ;

wherein;

the step of comparing the test signature to at least one template signature is carried out, at least in part, by treating the template signature and the test signature as one-pixel-wide curves each having at least a partial temporal ordering, said treatment leading to an error score D_DYNAMIC ;

the error score D_DYNAMIC is evaluated, at least in part, as a combination of error scores D_LOCAL and D_GLOBAL ;

the error score D_LOCAL is evaluated, at least in part, by matching sequenced strokes of the test signature to sequenced strokes of the template signature in accordance with the methods of dynamic programming;

the error score D_GLOBAL is evaluated, at least in part, by calculating at least one global feature on the test signature and on the reference signature, and comparing resulting corresponding global feature values; and

the step of comparing at least one error score to a threshold value comprises combining D_DYNAMIC and D_STATIC into a single error score D, and comparing D to a threshold value.

4. A method for verifying a test signature provided as a static two-dimensional pattern, comprising:

digitizing the test signature;

comparing the test signature to at least one template signature, thereby to obtain at least one error score; and

comparing at least one said error score to a threshold value;

CHARACTERIZED IN THAT the template signature is derived from real-time data obtained during the entry of at least one on-line reference signature, and the method further comprises;

a) providing a sequenced listing of discrete strokes of the template signature, said listing derived from said real-time data, said listing to be referred to as a "stroke ordering;

"

b) segmenting the test signature into discrete strokes;

c) before or during the comparing step, sequencing said test signature strokes to correspond, at least in part, to the stroke ordering of the template signature;

d) deriving at least one further template signature as a static two-dimensional rendering of at least one said on-line reference signature;

e) calculating at least one feature, to be referred to as a static feature, on the static template signature;

f) calculating at least one static feature on the test signature as a static two-dimensional pattern; and

g) comparing at least one static feature value of the static template signature with a corresponding static feature value of the test signature, thereby to obtain at least one further error score, to be referred to as a static error score D_STATIC ;

wherein;

the step of comparing the test signature to at least one template signature is carried out, at least in part, by treating the template signature and the test signature as one-pixel-wide curves each having at least a partial temporal ordering, said treatment leading to an error score D_DYNAMIC ;

the error score D_DYNAMIC is evaluated, at least in part, as a combination of error scores D_LOCAL and D_GLOBAL ;

the error score D_LOCAL is evaluated, at least in part, by matching sequenced strokes of the test signature to sequenced strokes of the template signature in accordance with the methods of dynamic programming;

the error score D_GLOBAL is evaluated, at least in part, by calculating at least one global feature on the test signature and on the reference signature, and comparing resulting corresponding global feature values; and

the step of comparing at least one error score to a threshold value comprises comparing D_DYNAMIC to a first threshold value and comparing D_STATIC to a second threshold value.

5. A method for verifying a test signature provided as a static two-dimensional pattern, comprising:

digitizing the test signature;

comparing the test signature to at least one template signature, thereby to obtain at least one error score; and

comparing at least one said error score to a threshold value;

CHARACTERIZED IN THAT the template signature is derived from real-time data obtained during the entry of at least one on-line reference signature, and the method further comprises;

a) providing a sequenced listing of discrete strokes of the template signature, said listing derived from said real-time data, said listing to be referred to as a "stroke ordering;

"

b) segmenting the test signature into discrete strokes; and

c) before or during the comparing step, sequencing said test signature strokes to correspond, at least in part, to the stroke ordering of the template signature and seeking at least one location of directional ambiguity in the test signature, and responsive to detection of said at least one location of directional ambiguity in the test signature, obtaining from said stroke ordering a listing of directions of strokes emanating from a location in said dynamic template signature corresponding to said at least one location of directional ambiguity in the test signature, and only if such a corresponding location exists, arranging the strokes of said test signature to emanate in substantially a same direction as corresponding strokes of said dynamic template signature;

wherein the comparing step comprises matching sequenced strokes of the test signature to sequenced strokes of the template signature; and

wherein said stroke-matching step is carried out by dynamic programming.

6. A method for verifying a test signature provided as a static two-dimensional pattern, comprising digitizing the test signature;

comparing the test signature to at least one reference signature, thereby to obtain at least one error score (D); and

comparing said at least one error score (D) to a threshold value;

CHARACTERIZED IN THAT a dac template signature is derived from real-time data obtained during an entry of said least one on-line reference signature, wherein the method further comprises;

a) providing and storing, as part of said dynamic template signature, a sequenced listing of discrete strokes of the reference signature, said listing derived from said real-time data, said listing to be referred to as a "stroke ordering;

"

b) normalizing the test signature;

c) thinning the test signature, after the step of normalizing, by seeking at least one location of directional ambiguity in the test signature, and responsive to detection of said at least one location of directional ambiguity in the test signature, obtaining from said stroke ordering a listing of directions of strokes emanating from a location in said dynamic template signature corresponding to said at least one location of directional ambiguity in the test signature; and

only if such a corresponding location exists, then modifying the step of thinning so that resulting thinned strokes of said test signature emanate in substantially a same direction as corresponding strokes of said dynamic template signature;

d) computing a global error score (D_GLOBAL) by global feature matching of the tat signature, after it has undergone the step of thinning, with the dynamic template signature;

e) computing a dynamic error score (D_DYNAMIC), at least in part, from said global error score (D_GLOBAL);

f) computing combined error score (D), at least in part from said dynamic error score (D_DYNAMIC); and

g) comparing the combined error score (D) with threshold value to either accept or reject the test signature.

7. A method according to claim 6, further comprising:

tracing the test sigma after it has undergone the steps of normalization and thinning, wherein, when an intersection or branch point is encountered, obtaining from said stroke ordering a listing of directions of strokes emanating from a location in said dynamic template signal corresponding to said intersection or branch point; and

only if such a corresponding location exists, then modifying the step of tracing so that resulting traced strokes of said test signature emanate in substantially a same direction as corresponding strokes of said dynamic template signature;

computing a local error score (D_LOCAL) by dynamic matching of the test signature, after it has undergone the step of tracing, with the dynamic template signature; and

computing the dynamic error score (D_DYNAMIC) from said global error score (D_GLOBAL) and said local error score (D_LOCAL).

8. A method according to claim 6, further comprising:

deriving a static template signature from said dynamic template signature;

computing a static error score (D_STATIC) by static 2D image matching of the test signature, after it has undergone the step of normalizing, with the static template signature; and

computing the combined error score (D) from said dynamic error score (D_DYNAMIC) and said static error score (D_STATIC).

9. A method according to claim 8, further comprising:

tracing the test signature after it has undergone the steps of normalization and thinning, wherein, when an intersection or branch point is encountered, obtaining from said stroke ordering a listing of directions of strokes emanating from a location in said dynamic template signature corresponding to said intersection or branch point; and

only if such a corresponding location exists, then modifying the step of tracing so that resulting traced strokes of said test signature emanate in substantially a same direction as corresponding strokes of said dynamic template signature;

computing a local error score (D_LOCAL) by dynamic matching of the test signature, after it has undergone the step of tracing, with the dynamic template signature;

computing the dynamic error score (D_DYNAMIC) from said global error score (D_GLOBAL) and said local error score (D_LOCAL); and

computing the combined error score (D) from said dynamic error score (D_DYNAMIC) and said static error score (D_STATIC).

10. A method for verifying a test signature provided as a static two-dimensional pattern, comprising digitizing the test signature;

comparing the test signature to at least one reference signature, thereby to obtain at least one error score (D); and

comparing said at least one error score (D) to a threshold value;

CHARACTERIZED IN THAT a dynamic template signature is derived from real-time data obtained during an entry of said least one on-line reference signature, wherein the method further comprises;

a) providing and storing, as part of said dynamic template signature, a sequenced listing of discrete strokes of the reference signature, said listing derived from said real-time data, said listing to be referred to as a "stroke ordering;

"

b) normalizing the test signature;

c) thinning the test signature, after the step of normalizing, by seeking at least one location of directional ambiguity in the test signature, and responsive to detection of said at least one location in the test signature, obtaining from said stroke ordering a listing of directions of strokes emanating from a location in said dynamic template signature corresponding to said at least one location in the test signature; and

only if such a corresponding location exists, then modifying the step of thinning so that resulting thinned strokes of said test signature emanate in substantially a same direction as corresponding strokes of said dynamic template signature;

d) computing a global error score (D_GLOBAL) by global feature matching of the test signature, after it has undergone the step of thinning, with the dynamic template signature;

c) tracing the test signature after it has undergone the steps of normalization and thinning, wherein, when an intersection or branch point is encountered, obtaining from said stroke ordering a listing of directions of strokes emanating from a location in said dynamic template signature corresponding to said intersection or branch point, and only if such a corresponding location exists, then modifying the step of tracing so that resulting traced strokes of said test signature emanate in substantially a same direction as corresponding strokes of said dynamic template signature;

f) computing a local error score (D_LOCAL) by dynamic matching of the test signature, after it has undergone the step of tracing, with the dynamic template signature; and

g) comparing said local error score (D_LOCAL) to a fit threshold and comparing said global error score (D_GLOBAL) to a second threshold, wherein the test signature is rejected if either said local error score (D_LOCAL) is less than said first threshold or said global error score (D_GLOBAL) is less than said second threshold, or both.

11. A method according to claim 10, further comprising:

h) deriving a static template signature from said dynamic template signature;

i) computing a static error score (D_STATIC) by static 2D image matching of the test signature, after it has undergone the stop of normalizing, with the static template signature; and

j) comparing said static or score (D_STATIC) to a third threshold, wherein the test signature is rejected if said static error score (D_STATIC) is less than said third threshold.