CN103023658B - Identity authentication method and identity authentication system based on signature - Google Patents

Abstract

The invention relates to an identity authentication method and an identity authentication system based on signature. The identity authentication method includes the steps of inputting signature information on a touch screen of a computer or a cellphone online, preprocessing the signature information via zooming a signature waveform, spinning, translating and filtering operation; extracting signature characteristics, finding out an optimized time calibration matching path between signature characteristics of a signature to be detected and signature characteristics of a sample signature, calculating distance between the signature to be detected and the sample signature; judging the signature to be true if the calculated distance is smaller than a threshold value, otherwise, judging the signature to be false. The identity authentication method and the identity authentication system have the advantages that the problems about forgetting a password, disclosure of the password and illegal attack and the like of a signature authentication method are avoided, safety in login and identity authentication is improved, and required authentication time is greatly reduced since the authentication is implemented in a one-to-one mode and authentication of a user name/password based authentication method is implemented in a one-to-many mode.

Description

A kind of identity identifying method based on handwritten signature and system

Technical field

The present invention relates to the area of pattern recognition that authentication is carried out in the biological characteristic that utilizes people and handwritten signature, especially a kind of identity identifying method based on handwritten signature and system.

Background technology

In field of identity authentication, most of field adopts fixing password identification authentication mode, the i.e. authentication mode of user name/password.The cardinal principle of the authentication mode of user name/password is compare in the information and date storehouses such as user name/password user inputted, if comparison success, then by authentication, otherwise, do not pass through authentication.Such as, the authentication of computer user, the authentication of the network user, the authentication of bank card user, enterprise staff enters the authentication etc. of Office Area.

User name/password identification authentication mode needs user to remember authenticate password mostly, and user tends to use short password, security of system can be caused on the low side, easily attacked, and when user uses long password, easily forgets; When the same password of user's Long-Time Service, retention cycle is partially long, and the possibility be cracked and day increase severely, but often change password, easily obscure again.Once user forgets authenticate password, then certification cannot be carried out.And disabled user is once obtain the authenticate password of validated user, just legal means can pass through checking, fail safe is poor.Password based on the identification authentication mode of user name/password is easy to forget, reveal and by others' rogue attacks.This mode more and more can not meet the requirement of existing network to fail safe, thus in the urgent need to the higher identity verification scheme of fail safe, the identity verification scheme therefore based on handwritten signature is arisen at the historic moment.

Summary of the invention

Namely brought be easy to forget and easily by the problem of attacking based on logging in of user name/password for prior art, the present invention proposes a kind of identity identifying method based on handwritten signature and system, based on the identity verification scheme of online Handwriting Signature Verification, improve the fail safe of network.The user that the identification authentication mode based on handwritten signature that the present invention proposes is specially adapted on touch-screen logs in, and can replace the identity verification scheme based on user name/password on the products such as computer, mobile phone and USB Key well.

The present invention solves the technical scheme that its technical problem adopts: this identity identifying method based on online Handwriting Signature Verification, the method step is as follows: input handwritten signature information first online in the touch-screen of computer or mobile phone, and preliminary treatment is carried out to signing messages, comprise the convergent-divergent of signature waveform, rotation, translation and filtering operation; Then signature character is extracted, and between signature to be measured and the signature character of sample signature, find mate path a time calibration optimized, finally calculate the distance of signature to be measured and sample signature, if the distance calculated is less than threshold value, then this signature is actual signature; Otherwise this signature is pseudo-signature.

Further, concrete steps are as follows:

(1) touch-screen, is utilized to extract coordinate figure and force value, recycling coordinate figure obtains speed, acceleration, angle and angular speed feature, extract order of strokes observed in calligraphy information during signature, the above feature composition characteristic vector that each moment is extracted, the corresponding characteristic sequence of each like this signature, the characteristic sequence f={f of signature to be measured of a certain moment ₁, f ₂, f ₃..., f _mrepresent; The characteristic sequence μ={ μ of a certain moment sample signature ₁, μ ₂, μ ₃..., μ _mrepresent;

(2), between signature to be measured and the signature character of sample signature, find mate path a time calibration optimized, go out the beeline between signature to be measured and sample signature by this coupling path computing; Specific implementation process is: alignd by two signature signals before matching, uneven distortion is carried out on a timeline with bending relative to sample signature signal by signature signal to be measured, make signature character to be measured point and the alignment of sample signature characteristic point, and constantly carry out the coupling path computing that makes two vector distances minimum between; Suppose from signature to be measured and sample signature, to gather several characteristic points respectively, Optimum Matching path is that the characteristic point of signature to be measured and the characteristic point of sample signature are set up mapping relations; Suppose that m characteristic point of signature to be measured and the n-th characteristic point of sample signature are mapping relations, by calculating the characteristic sequence f={f of m Feature point correspondence ₁, f ₂, f ₃..., f _mand the characteristic sequence μ={ μ of the n-th Feature point correspondence ₁, μ ₂, μ ₃..., μ _mbetween distance, obtain two signatures in certain any distance;

(3), according to the characteristic vector in the characteristic vector of signature to be measured extracted and sample database, the distance of signature to be measured and sample signature is calculated; According to the mapping relations of mating path the time calibration of previous step and obtaining between signature to be measured and sample signature characteristic point, the distance between any two mapping points is first calculated according to mapping relations, then to be got up the distance just obtained between two signatures by these distance superpositions, the distance metric formula wherein between any two mapping points is:

$D=\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{(f_i-{\mathrm{\μ}}_i)}^2$

As shown in above formula, D is the distance between two signature mapping points, and M is characteristic, f _irepresent i-th feature of signature character collection to be measured, μ _irepresent i-th feature of sample signature feature set;

(4), by the distance calculated and threshold value compare, if be less than specific threshold, then this signature is actual signature; Otherwise this signature is pseudo-signature.

This identity authorization system based on online Handwriting Signature Verification of the present invention, comprises pre-stored device, digital library and authenticate device; Described pre-stored device, for completing the registration of handwritten signature, namely gathers handwritten signature information by input equipment, and extracts sample signature feature; Described database is the sample signature feature extracted by registration phase, is kept at sample database, so that follow-up coupling; Described authenticate device, mainly comprises collecting unit, pretreatment unit, feature extraction unit and matching unit; Collecting unit: the input being realized handwritten signature by the touch-screen on computer and mobile phone; Pretreatment unit: the handwritten signature waveform of collecting unit collection is carried out series of preprocessing, is convenient to follow-up feature extraction and matching; Feature extraction unit: signature character is here on the one hand from the on-line signature feature of handwritten signature input, and the sample signature feature stated from user in sample database on the other hand, compares the two and can verify signature; Matching unit: compare inputting signature online with its signature stated in sample database, the distance of both calculating; The distance calculated and matching threshold are compared, if be less than threshold value, this signature is actual signature, otherwise is pseudo-signature.

The effect that the present invention is useful is: 1, Handwritten signature verfication method of the present invention instead of traditional authentication mode based on user name/password and logs in, avoid the forgeing of password, password reveals and by problems such as rogue attacks, improve the fail safe logged in; There is very low rate of by mistake refusing (FRR) and rate (FAR) of by mistake receiving, improve the fail safe of authentication system; 2, verification mode of the present invention is for verify one to one, and what adopt based on the verification mode of user name/password is one-to-many checking, greatly reduces the time needed for checking; 3, the generality, uniqueness, stability and the collection property that have of the present invention, ensure that its unique advantage in authentication.

Accompanying drawing explanation

Fig. 1 is based on the identity authorization system structured flowchart of line Handwritten signature verfication;

Fig. 2 is based on the identity identifying method flow chart of line Handwritten signature verfication;

Fig. 3 is the handwritten signature identity authorization system structured flowchart of bank system of web.

Embodiment

Below in conjunction with drawings and Examples, the invention will be further described:

As shown in Figure 1, pre-stored device, digital library and authenticate device is comprised based on the identity authorization system of online Handwriting Signature Verification in present embodiment.Pre-stored device, for completing the registration of handwritten signature, namely gathers handwritten signature information by input equipment, and extracts sample signature feature; Database is the sample signature feature extracted by registration phase, is kept at sample database, so that follow-up coupling; Authenticate device mainly comprises collecting unit, pretreatment unit, feature extraction unit and matching unit.Collecting unit: the input being realized handwritten signature by the touch-screen on the product such as computer and mobile phone; Pretreatment unit: the handwritten signature waveform of collecting unit collection is carried out series of preprocessing, is convenient to follow-up feature extraction and matching; Feature extraction unit: signature character is here on the one hand from the on-line signature feature of handwritten signature input, and the sample signature feature stated from user in sample database on the other hand, compares the two and can verify signature.Signature character fully should reflect the writing style of signer, should have metastable characteristic simultaneously; Matching unit: compare inputting signature online with its signature stated in sample database, the distance of both calculating.The distance calculated and matching threshold are compared, if be less than threshold value, this signature is actual signature, otherwise is pseudo-signature.This process is man-to-man matching process, and namely whether the identity of the people of checking input signature is true.

Fig. 2 shows the method flow adopting system shown in Figure 1 to carry out authentication, the method comprises and inputs handwritten signature information first online in the touch-screen of computer or mobile phone, and preliminary treatment is carried out to signing messages, comprise the convergent-divergent of signature waveform, rotation, translation and filtering operation etc.Then signature character is extracted, and between signature to be measured and the signature character of sample signature, find mate path a time calibration optimized, finally calculate the distance of signature to be measured and sample signature, if the distance calculated is less than specific threshold, then this signature is actual signature; Otherwise this signature is pseudo-signature.

Preliminary treatment: even if same person writes same word, speed when writing also can change, and time period when writing shared by each stroke generally also can be different, are therefore necessary first to carry out preliminary treatment to signature, make it in the scope of allowing, do some and correct.Preliminary treatment comprises the signature convergent-divergent of waveform, rotation, translation and filtering operation etc.Also must remove the interval between stroke in the process, these Gap responses signer does not touch the time of touch-screen when signing and leaves touch-screen and the position contacting touch-screen, also needs the normalized of carrying out size, length etc. of signing simultaneously.

The extraction of signature character comprises: utilize touch-screen to extract coordinate figure and force value, recycling coordinate figure obtains other features as speed, acceleration, angle (direction of writing stroke) and angular speed, need the order of strokes observed in calligraphy information sequencing of stroke (when signature is write) when extracting signature in addition, the above feature composition characteristic vector that each moment is extracted, the corresponding characteristic sequence of each like this signature.The characteristic sequence f={f of signature to be measured of a certain moment ₁, f ₂, f ₃..., f _mrepresent; The characteristic sequence μ={ μ of a certain moment sample signature ₁, μ ₂, μ ₃..., μ _mrepresent.

Find Optimum Matching path in Fig. 2, between signature to be measured and the signature character of sample signature, namely find mate path a time calibration optimized, the beeline between signature to be measured and sample signature can be calculated by this coupling path.Be: due to signature to be measured and the writing speed of sample signature and the difference of writing position be difficult to two signatures directly to mate according to specific implementation process.Therefore two signature signals are needed to align before matching, uneven distortion is carried out on a timeline with bending relative to sample signature signal by signature signal to be measured, make signature character to be measured point and the alignment of sample signature characteristic point, and constantly carry out the coupling path computing that makes two vector distances minimum between.Suppose from signature to be measured and sample signature, to gather several characteristic points respectively, the object finding Optimum Matching path is the characteristic point of signature to be measured and the characteristic point of sample signature to set up mapping relations.Suppose that m characteristic point of signature to be measured and the n-th characteristic point of sample signature are mapping relations, by calculating the characteristic sequence f={f of m Feature point correspondence ₁, f ₂, f ₃..., f _mand the characteristic sequence μ={ μ of the n-th Feature point correspondence ₁, μ ₂, μ ₃..., μ _mbetween distance, two signatures can be obtained in certain any distance.

According to the characteristic vector in the characteristic vector of the signature to be measured extracted and sample database, calculate the distance (distance here adopt be Euclidean distance) of signature to be measured and sample signature.According to the mapping relations of mating path the time calibration of previous step and obtaining between signature to be measured and sample signature characteristic point, first calculate the distance between any two mapping points according to mapping relations, the distance just obtained between two signatures of then being got up by these distance superpositions.Distance metric formula wherein between any two mapping points is:

$D=\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{(f_i-{\mathrm{\μ}}_i)}^2$

As shown in above formula, D is the distance between two signature mapping points, and M is characteristic, f _irepresent i-th feature of signature character collection to be measured, μ _irepresent i-th feature of sample signature feature set.

The distance calculated and threshold value are compared, if be less than specific threshold, then this signature is actual signature; Otherwise this signature is pseudo-signature.As far as possible the prerequisite of Threshold selection here keeps lower rate of by mistake refusing (FRR) and rate (FAR) of by mistake receiving.

Fig. 3 is the example block diagram of the handwritten signature identity authorization system application of bank system of web.As shown in Figure 3, when user to log in bank server complete operate accordingly time, the handwritten signature of oneself first need be utilized to carry out authentication, only have and just can proceed corresponding operating by authentication, otherwise by the information of return authentication failure.

Typical apply of the present invention comprises: (1) bank system of web; (2) intranets systems; (3) securities exchange system; (4) office automation system.

Handwritten signature identity authorization system is except above typical apply, and every place that user name/password can be used to carry out certification, all can use handwritten signature to replace.The visible identity authorization system based on handwritten signature has very large actual application value.

Terminological interpretation

Refuse rate: be called for short FRR (False Rejection Rate) by mistake.I signs and knows for my signature non-and refuse acceptance, i.e. FRR=a/c by mistake by system.Wherein, a is the number of samples by mistake known for my signature non-of I being signed in discrimination process; C is the total number of sample of discrimination process.

Receive rate: be called for short FAR (FalseAcceptance Rate) by mistake.My signature non-is known for my signature and accepting, i.e. FAR=b/c by system by mistake.Wherein, my signature non-is known the number of samples for my signature in discrimination process by b by mistake; C is the total number of sample of discrimination process.

In addition to the implementation, the present invention can also have other execution modes.All employings are equal to the technical scheme of replacement or equivalent transformation formation, all drop on the protection range of application claims.

Claims (1)

1. the identity identifying method based on online Handwriting Signature Verification, it is characterized in that: the method step is as follows: input handwritten signature information first online in the touch-screen of computer or mobile phone, and preliminary treatment is carried out to signing messages, comprise the convergent-divergent of signature waveform, rotation, translation and filtering operation; Then signature character is extracted, and between signature to be measured and the signature character of sample signature, find mate path a time calibration optimized, finally calculate the distance of signature to be measured and sample signature, if the distance calculated is less than threshold value, then this signature is actual signature; Otherwise this signature is pseudo-signature, and concrete steps are as follows:

(1) touch-screen, is utilized to extract coordinate figure and force value, recycling coordinate figure obtains speed, acceleration, angle and angular speed feature, extract order of strokes observed in calligraphy information during signature, the above feature composition characteristic vector that each moment is extracted, the corresponding characteristic sequence of each like this signature, the characteristic sequence f={f of signature to be measured of a certain moment ₁, f ₂, f ₃..., f _mrepresent; The characteristic sequence μ={ μ of a certain moment sample signature ₁, μ ₂, μ ₃..., μ _mrepresent;

(2), between signature to be measured and the signature character of sample signature, find mate path a time calibration optimized, go out the beeline between signature to be measured and sample signature by this coupling path computing; Specific implementation process is: alignd by two signature signals before matching, uneven distortion is carried out on a timeline with bending relative to sample signature signal by signature signal to be measured, make signature character to be measured point and the alignment of sample signature characteristic point, and constantly carry out the coupling path computing that makes two vector distances minimum between; Suppose from signature to be measured and sample signature, to gather several characteristic points respectively, Optimum Matching path is that the characteristic point of signature to be measured and the characteristic point of sample signature are set up mapping relations; Suppose that m characteristic point of signature to be measured and the n-th characteristic point of sample signature are mapping relations, by calculating the characteristic sequence f={f of m Feature point correspondence ₁, f ₂, f ₃..., f _mand the characteristic sequence μ={ μ of the n-th Feature point correspondence ₁, μ ₂, μ ₃..., μ _mbetween distance, obtain two signatures in certain any distance;

(3), according to the characteristic vector in the characteristic vector of signature to be measured extracted and sample database, the distance of signature to be measured and sample signature is calculated; According to the mapping relations of mating path the time calibration of previous step and obtaining between signature to be measured and sample signature characteristic point, the distance between any two mapping points is first calculated according to mapping relations, then to be got up the distance just obtained between two signatures by these distance superpositions, the distance metric formula wherein between any two mapping points is:

$D=\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{(f_i-{\mathrm{\μ}}_i)}^2$

As shown in above formula, D is the distance between two signature mapping points, and M is characteristic, f _irepresent i-th feature of signature character collection to be measured, μ _irepresent i-th feature of sample signature feature set;

(4), by the distance calculated and threshold value compare, if be less than specific threshold, then this signature is actual signature; Otherwise this signature is pseudo-signature.