Method and apparatus for parametric signature verification using global features and stroke-direction codes

• US 5,828,772 A
• Filed: 12/27/1995
• Issued: 10/27/1998
• Est. Priority Date: 12/27/1995
• Status: Expired due to Term

• Current Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Original Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Inventors: Kashi, Ramanujan S., Turin, William, Nelson, Winston Lowell
• Primary Examiner(s): Moore, David K.
• Assistant Examiner(s): Werner, Brian P.
• Application Number:
  US08/579,534
• Time in Patent Office: 1,035 Days
• Field of Search: 382/119, 382/120, 382/122, 382/123, 382/186, 382/197
• US Class Current: 382/119
• CPC Class Codes: G06V 40/30 Writer recognition; Reading and verifying signatures
• CPC Class Codes: G07C 9/35 by means of a handwritten signature

Abstract

A signature verification method operates by comparing the numerical values of parameters evaluated on a trial signature with stored reference data derived from previously entered reference signatures. The parameters include global features of the signature, and also include a stroke-direction code (SDC) of the signature. An SDC is derived by subdividing the signature into a sequence of time-ordered, spatially oriented line segments, each segment extending between a pair of discrete points along the signature. Each line segment has a stroke-direction value. The SDC is the ordered sequence of these stroke-direction values.

Claims

1. A method for encoding the handwritten signature of a person, comprising:

(a) digitally sampling the trajectory of a stylus thereby to provide a digitized signature record;

(b) storing said record in a digital memory;

(c) digitally processing said record, thereby to derive a numerical value for each of a plurality of global features of the signature;

(d) storing said numerical values in a digital memory;

(e) subdividing the signature into a time-ordered sequence of oriented links;

(f) assigning to each link a quantized stroke-direction value corresponding to the respective orientation of that link, thereby to obtain for the signature as a whole an ordered sequence of stroke-direction values, said sequence to be referred to as a stroke-direction code (SDC);

(g) storing the SDC in a digital memory;

(h) repeating steps (a)-(g) for each of a plurality of signatures provided by the same person, thereby to provide, in a digital memory;

(i) a reference set of feature vectors, each of said feature vectors consisting of the numerical global-feature values relating to a respective one of the signatures; and

(ii) a reference set of SDCs, each said SDC relating to a respective one of the signatures;

(i) averaging the feature vectors, thereby to derive an average and a deviation measure of each of the respective global features over the feature-vector reference set;

(j) storing the averages and the deviation measures in a digital memory;

(k) evaluating a distance function, to be referred to as the SDC distance, between each pair of SDCs in the SDC reference set;

(l) identifying as an average or representative SDC, to be referred to as the SDC template, that one SDC that has the least mean-square distance to all of the other SDCs in the SDC reference set; and

(m) storing the SDC template in a digital memory.

2. The method of claim 1, wherein the distance-function-evaluating step is performed by dynamic programming.

3. A method for encoding the handwritten signature of a person, comprising:

(a) digitally sampling the trajectory of a stylus thereby to provide a digitized signature record;

(b) storing said record in a digital memory;

(c) digitally processing said record, thereby to derive a numerical value for each of a plurality of global features of the signature; and

(d) storing said numerical values in a digital memory;

CHARACTERIZED IN THAT the digital processing step further comprises;

(e) taking a Fourier transform of the signature, thereby to derive a sequence of Fourier descriptors;

(f) normalizing the Fourier descriptors for size, position, and orientation;

(g) after (f), subdividing the signature into a time-ordered sequence of oriented links;

(h) assigning to each link a quantized stroke-direction value corresponding to the respective orientation of that link, thereby to obtain for the signature as a whole an ordered sequence of stroke-direction values, said sequence to be referred to as a stroke-direction code (SDC); and

(i) storing the SDC in a digital memory.

4. A method for verifying the handwritten trial signature of a person, comprising:

(a) digitally sampling the trajectory of a stylus, thereby to provide a digitized signature record;

(b) storing said record in a digital memory;

(c) digitally processing said record, thereby to derive a numerical value for each of a plurality of global features of the signature; and

(d) in a digital data-processing unit, comparing said numerical global-feature values to global-feature reference data retrieved from a digital memory;

CHARACTERIZED IN THAT the method further comprises;

(e) taking a Fourier transform of the signature, thereby to derive a sequence of Fourier descriptors;

(f) normalizing the Fourier descriptors for size, position, and orientation;

(g) after (f), subdividing the signature into a time-ordered sequence of oriented links;

(h) assigning to each link a quantized stroke-direction value corresponding to the respective orientation of that link, thereby to obtain for the trial signature as a whole an ordered sequence of stroke-direction values, said sequence to be referred to as a trial stroke-direction code (SDC); and

(i) comparing said trial SDC to SDC reference data retrieved from a digital memory.

5. A method for verifying the handwritten trial signature of a person, comprising:

(a) digitally sampling the trajectory of a stylus, thereby to provide a digitized signature record;

(b) storing said record in a digital memory;

(c) digitally processing said record, thereby to derive a numerical value for each of a plurality of global features of the signature;

(d) in a digital data-processing unit, comparing said numerical global-feature values to global-feature reference data retrieved from a digital memory;

(e) subdividing the signature into a time-ordered sequence of oriented links;

(f) assigning to each link a quantized stroke-direction value corresponding to the respective orientation of that link, thereby to obtain for the trial signature as a whole an ordered sequence of stroke-direction values, said sequence to be referred to as a trial stroke-direction code (SDC); and

(g) comparing said trial SDC to SDC reference data retrieved from a digital memory;

wherein;

the global-feature reference data comprise the mean value and standard deviation of each of the respective global features taken over a plurality of reference signatures provided by the person during a registration procedure;

step (d) comprises;

determining a difference between each numerical global-feature value and the corresponding one of said mean values; and

combining said differences, thereby to obtain a global-feature error value;

the SDC reference data comprise an average or representative SDC taken over said plurality of reference signatures, to be referred to as an SDC template; and

step (g) comprises;

by dynamic programming, determining a distance, to be referred to as the SDC error value, between the trial SDC and the SDC template.

6. The method of claim 5, further comprising:

combining the global-feature error value with the SDC error value, thereby to obtain a total error;

comparing the total error to a threshold; and

if the total error is smaller than the threshold, activating an electric circuit, thereby to indicate that the trial signature is verified.

7. The method of claim 6, wherein:

each difference between a numerical global-feature value and the corresponding mean value is normalized to the corresponding standard deviation before combining said differences.

8. The method of claim 5, further comprising:

comparing the global-feature error value with a first threshold;

comparing the SDC error value with a second threshold; and

if the global-feature error value is less than the first threshold, and also the SDC error value is less than the second threshold, activating an electric circuit, thereby to indicate that the trial signature is verified.

9. Apparatus comprising:

a digital memory for receiving and storing digitized human signatures from a signature digitizing device;

a global feature extractor in receiving relationship to said memory, for;

(i) deriving, from each digitized signature received from said memory, numerical values of at least two global features of said digitized signature, and (ii) storing said values in a data storage device as a global feature vector;

a stroke-direction coder, in receiving relationship to said memory, for;

(i) deriving, from each digitized signature received from said memory, an ordered sequence of numerical stroke-direction values, to be referred to as an SDC vector, and (ii) storing said values in a data storage device;

an averager for;

(i) receiving plural global feature vectors and deriving from said vectors an average global feature vector, and (ii) storing said average global feature vector in a data storage device;

a processing element for;

(i) receiving plural SDC vectors and deriving from said vectors an average or representative SDC vector, and (ii) storing said average or representative SDC vector in a data storage device; and

a correlating element for deriving, by dynamic programming methods, a distance between each pair of said plural SDC vectors.

10. Apparatus of claim 9, wherein said processing element further comprises:

a selection element for identifying, from said distances, that SDC vector that lies nearest a centroid of said plural SDC vectors.

11. Apparatus comprising:

a digital memory for receiving and storing digitized human signatures from a signature digitizing device;

a global feature extractor in receiving relationship to said memory, for;

(i) deriving, from each digitized signature received from said memory, numerical values of at least two global features of said digitized signature, and (ii) storing said values in a data storage device as a global feature vector;

a stroke-direction coder, in receiving relationship to said memory, for;

(i) deriving, from each digitized signature received from said memory, an ordered sequence of numerical stroke-direction values, to be referred to as an SDC vector, and (ii) storing said values in a data storage device;

a first error element for;

(i) comparing the global feature vector of a designated signature to a stored global feature reference vector, and (ii) deriving a first error measure of discrepancy between said compared feature vectors;

a second error element for;

(i) comparing the SDC vector of a designated signature to a stored average or representative SDC vector, and (ii) applying dynamic programming methods to derive a second error measure of distance between said compared SDC vectors; and

a threshold element for;

(i) applying a threshold condition to said first and second error measures, and (ii) if said threshold condition is satisfied, activating an electric circuit, thereby to indicate that the designated signature has been verified.