Measure of distinguishability for signature verification

Claim 2

A machine method of verifying a signature of a putative signer, comprising the steps of: (a) measuring a set of physical writing characteristics of the putative signer when the putative signer writes his signature; (b) determining input measures (M 2 , M 2 , . . . ) for the signature of the from the measurements of the physical writing characteristics in step (a); (c) determining similarity components (Sm 2 , Sm 2 , . . . ) from the input measure obtained in step (b), each similarity component being defined by the relationship ##EQU19## where MIDi is approximately the value of the i-th measure at the intersection of the normalized distributions for the i-th measure for verifies and forgeries, where HIi is chosen as a value at which a relatively large preselected percentage of forgeries have a similarity component of less than a preselected maximum, where LOi is chosen as a value at which a relatively large preselected percentage of verifies have a similarity component value greater than a preselected minimum, and where m is a preselected number ≧1; (d) determining a preliminary similarity measure from the similarity components of step (c), the preliminary similarity measure being defined as the sum of the similarity components; (e) comparing the preliminary similarity measure with a preselected threshold to verify the signature.

Claim 3

A machine method as recited in claim 1 wherein the step of measuring the set of physical writing characteristic comprises the step of measuring electrical signals produced by a pen when the signature is written with the pen.

Claim 4

A machine method as recited in claim 2, wherein the electrical signals produced by the pen comprise three sample signals, the first sample signal being proportional to the x-acceleration, the second sample signal being proportional to the y-acceleration and finally the third signal being proportional to rate of change of axial pen pressure, all three sample signals being produced by the writing of the signature by the putative signer.

Claim 5

A machine method as recited in claim 2, wherein the input measures determined comprise the following input measures: segment alignment, pressure correlation, acceleration correlation, pressure coherence, and acceleration of coherence.

Claim 6

A machine method as recited in claim 1, wherein the physical writing characteristics comprise the characteristics of the rate of change of pen pressure and pen acceleration.

Claim 7

A machine method as recited in claim 5, wherein the rate of change of pen pressure is measured in the axial direction of the pen and the acceleration, the acceleration being measured in two approximately orthogonal directions in a plane perpendicular to the pen axis.

Claim 8

A machine method of verifying a signature of a putative signer, comprising the steps of: (a) measuring a set of physical writing characteristics of the putative signer when the putative signer writes his signature; (b) determining input measures (M 1 ,M 2 , . . . ) for the signature of the putative signer from the measurements of the physical writing characteristics in step (a); (c) determining similarity components (Sm 1 ,SM 2 , . . . ) from the input measures obtained in step (b), each similarity component being defined by the relationship ##EQU20## where MIDi is approximately the value of the i-th measure at the intersection of the normalized distributions for the i-th measure for verifies and forgeries, where HIi is chosen as a value at which a relatively large preselected percentage of forgeries have a similarity component of less than a preselected maximum, where LOi is chosen as a value at which a relatively large preselected percentage of verifies have a similarity component value greater than a preselected minimum, and where m is a preselected number ≧3; (d) determining a preliminary similarity measure (Smp) from the similarity components of step (c), the preliminary similarity measure being defined as the sum of the similarity components; (e) determining a similarity measure from the preliminary similarity measure, the similarity measure (SM) being defined as SM=Smp-REF where REF is an individualized small adjustment parameter determined during enrollment by REF=Smp×SENSITIVITY-EXPECTED where the bar over Smp indicates that average input measures for a set of enrollment signatures and standard similarity parameters are used to determine the preliminary similarity measure, sensitivity approximately =0.588 and expected approximately=0.0053; (f) comparing the similarity measure with a preselected threshold to verify the signature.

Claim 9

A machine method of verifying a signature of a putative signer, comprising the steps of: (a) measuring a set of physical writing characteristics of the putative signer when the putative signer writes his signature; (b) determining input measures (M 1 ,M 2 , . . . ) for the signature of the putative signer from the measurements of the physical writing characteristics in step (a); (c) determining similarity components (Sm 1 ,SM 2 , . . . ) from the input measures obtained in step (b), each similarity component being defined by the relationship ##EQU20## where MIDi is approximately the value of the i-th measure at the intersection of the normalized distributions for the i-th measure for verifies and forgeries, where HIi is chosen as a value at which a relatively large preselected percentage of forgeries have a similarity component of less than a preselected maximum, where LOi is chosen as a value at which a relatively large preselected percentage of verifies have a similarity component value greater than a preselected minimum, and where m is a preselected number ≧3; (d) determining a preliminary similarity measure (Smp) from the similarity components of step (c), the preliminary similarity measure being defined as the sum of the similarity components; (e) determining a similarity measure from the preliminary similarity measure, the similarity measure (SM) being defined as SM=Smp-REF where REF is an individualized small adjustment parameter determined during enrollment by REF=Smp×SENSITIVITY-EXPECTED where the bar over Smp indicates that average input measures for a set of enrollment signatures and standard similarity parameters are used to determine the preliminary similarity measure, sensitivity approximately =0.588 and expected approximately=0.0053; (f) comparing the similarity measure with a preselected threshold to verify the signature.

Claim 10

A machine method as recited in claim 7, wherein the step of measuring the set of physical writing characteristic comprises the step of measuring electrical signals produced by a pen when the signature is written with the pen.

Claim 11

A machine method as recited in claim 8, wherein the electrical signals produced by the pen comprise three sample signals, the first sample signal being proportional to the x-acceleration, the second sample signal being proportional to the y-acceleration and finally the third signal being proportional to rate of change of axial pen pressure, all three sample signals being produced by the writing of the signature by the putative signer.

Claim 12

A machine method as recited in claim 8, wherein the input measures determined comprise the following input measures: segment alignment, pressure correlation, acceleration correlation, pressure coherence, and acceleration of coherence.

Claim 13

A machine method as recited in claim 7, wherein the physical writing characteristics comprise the characteristics of the rate of change of pen pressure and pen acceleration.

Claim 14

A machine method as recited in claim 11, wherein the rate of change of pen pressure is measured in the axial direction of the pen and wherein, the acceleration is measured in two approximately orthogonal directions in a plane perpendicular to the pen axis.

Claim 15

A machine method as recited in claim 7, wherein there is at least two input measures.

Claim 16

A machine method as recited in claim 7, wherein the input measures determined are the five input measures in claim 13.

Claim 17

A machine method as recited in claim 7, wherein the integer m equals 3.

Claim 18

A machine method of verifying a signature of a putative signer, comprising the steps of: (a) measuring the rate of change of axial pen pressure and two approximately orthogonal components of pen acceleration, the acceleration components being in a plane orthogonal to the axis of the pen, the acceleration components and the rate of change pressure being produced on a pen by the writing of the signature by the putative signer with the pen of the writing surface; (b) determining input measures (M 1 ,M 2 , . . . ) from the rate of change of axial pen pressure and the two approximately orthogonal acceleration components, the input measures comprising: a segment alignment input measure, a pressure correlation input measure, an acceleration correlation input measure, a pressure coherence input measure and an acceleration coherence input measure; (c) determining similarity components (Sm i ) for each of the input measures obtained in step (b), each similarity component being defined by the relationship ##EQU21## where MIDi is approximately the value of the i-th measure at the intersection of the normalized distributions for the i-th measure for verifies and forgeries, where HIi is chosen as a value at which a relatively large preselected percentage of forgeries have a similarity component of less than a preselected maximum, where LOi is chosen as a value at which a relatively large preselected percentage of verifies have a similarity component value greater than a preselected minimum, and where m is a preselected number ≧3; (d) determining a preliminary similarity measure (Smp) which is the sum of the similarity components of step (c); (e) determining a similarity measure from the preliminary similarity measure, the similarity measure (SM) being defined as SM=Smp-REF where REF is an individualized small adjustment parameter determined during enrollment by REF=Smp×SENSlTIVITY-EXPECTED where the bar over Smp indicates that average input measures for a set of enrollment signatures and standard similarity parameters are used to determine the preliminary similarity measure, sensitivity approximately =0.588 and expected approximately =0.0053; and (f) comparing the similarity measure with a preselected threshold to verify the signature.

Claim 19

A machine method of verifying a signature of a putative signer, comprising the steps of: (a) measuring the rate of change of axial pen pressure and two approximately orthogonal components of pen acceleration, the acceleration components being in a plane orthogonal to the axis of the pen, the acceleration components and the rate of change pressure being produced on a pen by the writing of the signature by the putative signer with the pen of the writing surface; (b) determining input measures (M 1 ,M 2 , . . . ) from the rate of change of axial pen pressure and the two approximately orthogonal acceleration components, the input measures comprising: a segment alignment input measure, a pressure correlation input measure, an acceleration correlation input measure, a pressure coherence input measure and an acceleration coherence input measure; (c) determining similarity components (Sm i ) for each of the input measures obtained in step (b), each similarity component being defined by the relationship ##EQU21## where MIDi is approximately the value of the i-th measure at the intersection of the normalized distributions for the i-th measure for verifies and forgeries, where HIi is chosen as a value at which a relatively large preselected percentage of forgeries have a similarity component of less than a preselected maximum, where LOi is chosen as a value at which a relatively large preselected percentage of verifies have a similarity component value greater than a preselected minimum, and where m is a preselected number ≧3; (d) determining a preliminary similarity measure (Smp) which is the sum of the similarity components of step (c); (e) determining a similarity measure from the preliminary similarity measure, the similarity measure (SM) being defined as SM=Smp-REF where REF is an individualized small adjustment parameter determined during enrollment by REF=Smp×SENSlTIVITY-EXPECTED where the bar over Smp indicates that average input measures for a set of enrollment signatures and standard similarity parameters are used to determine the preliminary similarity measure, sensitivity approximately =0.588 and expected approximately =0.0053; and (f) comparing the similarity measure with a preselected threshold to verify the signature.

Claim 20

A machine method of verifying a signature of a putative signer, comprising the steps of: (a) measuring the rate of change of axial pen pressure and two approximately orthogonal components of acceleration, the acceleration components being a plane parallel to a writing surface, the components and the pressure being produced on a pen by the writing of the signature by the putative signer, the pen producing electrical signals, each respective signal being proportional to the pressure or to one of the acceleration components; (b) determining input measures (M 1 ,M 2 , . . . ) from the rate of change of axial pen pressure and the two approximately orthogonal acceleration components, the input measures comprising: a segment alignment input measure, a pressure correlation input measure, an acceleration correlation input measure, a pressure coherence input measure and an acceleration coherence input measure; (c) determining similarity components (Sm i ) for each of the input measures obtained in step (b), each similarity component being defined by the relationship ##EQU22## where MIDi is approximately the value of the i-th measure at the intersection of the normalized distributions for the i-th measure for verifies and forgeries, where HIi is chosen as a value at which a relatively large preselected percentage of forgeries have a similarity component of less than a preselected maximum, where LOi is chosen as a value at which a relatively large preselected percentage of verifies have a similarity component value greater than a preselected minimum, and where m is a preselected number ≧3; (d) determining a preliminary similarity measure (Smp) which is the sum of the similarity components of step (c); (e) determining a similarity measure from the preliminary similarity measure, the similarity measure (SM) being defined as SM=Smp-REF where REF is an individualized small adjustment parameter determined during enrollment by REF=Smp·SENSITIVITY-EXPECTED where the bar over Smp indicates that average input measures for a set of enrollment signatures and standard similarity parameters are used to determine the preliminary similarity measure, sensitivity approximately =0.588 and expected approximately =0.0053; and (f) comparing the similarity measure with a preselected threshold to verify the signature.

Claim 21

A machine method of verifying a signature of a putative signer, comprising the steps of: (a) measuring the rate of change of axial pen pressure and two approximately orthogonal components of acceleration, the acceleration components being a plane parallel to a writing surface, the components and the pressure being produced on a pen by the writing of the signature by the putative signer, the pen producing electrical signals, each respective signal being proportional to the pressure or to one of the acceleration components; (b) determining input measures (M 1 ,M 2 , . . . ) from the rate of change of axial pen pressure and the two approximately orthogonal acceleration components, the input measures comprising: a segment alignment input measure, a pressure correlation input measure, an acceleration correlation input measure, a pressure coherence input measure and an acceleration coherence input measure; (c) determining similarity components (Sm i ) for each of the input measures obtained in step (b), each similarity component being defined by the relationship ##EQU22## where MIDi is approximately the value of the i-th measure at the intersection of the normalized distributions for the i-th measure for verifies and forgeries, where HIi is chosen as a value at which a relatively large preselected percentage of forgeries have a similarity component of less than a preselected maximum, where LOi is chosen as a value at which a relatively large preselected percentage of verifies have a similarity component value greater than a preselected minimum, and where m is a preselected number ≧3; (d) determining a preliminary similarity measure (Smp) which is the sum of the similarity components of step (c); (e) determining a similarity measure from the preliminary similarity measure, the similarity measure (SM) being defined as SM=Smp-REF where REF is an individualized small adjustment parameter determined during enrollment by REF=Smp·SENSITIVITY-EXPECTED where the bar over Smp indicates that average input measures for a set of enrollment signatures and standard similarity parameters are used to determine the preliminary similarity measure, sensitivity approximately =0.588 and expected approximately =0.0053; and (f) comparing the similarity measure with a preselected threshold to verify the signature.