CN104573748A - Dynamic signature recognition method based on touch screen mobile device - Google Patents

Abstract

The invention relates to a dynamic signature recognition method based on a touch screen mobile device, and belongs to the technical field of bioidentification. On the basis of easily-obtained signature polar angle characteristics, user signatures are matched to achieve dynamic signature recognition on a mobile terminal. The dynamic signature recognition method comprises the following steps: firstly, preprocessing an initial user signature data obtained by sampling to enable two adjacent points to have equal time interval, and on the basis of a centroid of a signature in an X-Y plane and a golden centroid, normalizing the size of a character shape; secondly, selecting a pole to establish a polar coordinate system, calculating characteristic sequences of an polar angle and a polar radius of the signature, and extracting stable polar value points from the polar angle characteristics as separating points; finally, on the basis of sample signatures to be measured and a separating point sequence of a template, searching and matching to obtain a global optimal matching scheme and the best similarity so as to judge the identity of a signer. According to the technical scheme, signature recognition on the terminal has the advantages of low cost, high accuracy, high calculating speed and the like.

Description

A kind of dynamic signature recognition methods based on touch-screen mobile device

Technical field

The present invention relates to a kind of dynamic signature recognition methods, particularly a kind of dynamic signature recognition methods based on touch-screen mobile device, belongs to technical field of biometric identification.

Background technology

Dynamic signature recognition technology utilizes the various characteristic informations in special digitizing handwriting equipment collection signature writing process, through processing further, surface---font and internal feature---pressure, speed, the acceleration etc. of the person's individual information that is processed into identification signature, and by the similarity between the tested signature of corresponding method comparison and reference template, thus judge the true and false of signing according to result, and with the legitimacy of this certifying signature people identity in real time.Therefore, dynamic signature authentication technique can be used in the occasion using and the mobile terminal of touch panel device needs to carry out authentication, that is: judge current whether what using mobile device be legal user, this is of great significance tool mobile terminal.Although some terminal devices are integrated with fingerprint recognition chip at present, utilize fingerprint identification technology to identify identity, but this needs to increase special hardware cost, and existing touch-screen has on the terminal device become basic configuration, touch-screen has been utilized to obtain signing messages and carry out the authentication means that authentication has become a kind of quick cheapness.

Existing dynamic signature discrimination method can be divided into two large classes: parametric method sum function method.Parametric method is few for the data volume characterizing signing messages, make calculating easy, quick, but error rate is higher; The Major Difficulties of function method is the Proper Match between signature, namely by effective method, two signature time serieses must be carried out Nonlinear Mapping, find the reasonable corresponding relation between signature parts, then just can compare, this process more complicated, operand are larger.

Summary of the invention

The object of the invention is to the above-mentioned defect overcoming prior art, a kind of high-efficiency dynamic signature identifying method based on mobile terminal of touch screen equipment is provided, the method is characterized as foundation with the signature polar angle being easy to obtain, the signature of user is mated, realizes the dynamic signature identification on mobile terminal.

Thought of the present invention is by first carrying out pre-service to the initial user signed data that sampling obtains, the sampling of putting on the touchscreen user due to each operating system (as: android system, IOS system etc.) of mobile terminal is not equifrequent, therefore manually adjustment makes between adjacent two points is constant duration, again with signature barycenter on an x-y plane and gold barycenter for foundation, size normalizing is carried out to font; Then, choose corresponding limit on this basis, set up polar coordinate system, the polar angle of compute signature and footpath, pole characteristic sequence, and the stable extremal point extracting polar angle feature is as separation; Finally, with sample to be tested signature and template separation sequence for according to carry out search mate, obtain global optimum's matching scheme and Best similarity degree, to judge the identity of signer.

When carrying out the similarity mode of corresponding pen section, for font style characteristic, shorter polar angle characteristic sequence entirety is mated successively with the different parts of longer polar angle characteristic sequence, calculates similarity, and using the similarity of the maximum similarity of acquisition as these two sequences.For behavioral characteristics, because the time interval in sequence between two consecutive point is all equal, therefore the length of sequence just represents the average velocity of pen section corresponding to this sequence, so the similarity of calculate two sequences be multiplied by compared with the length ratio of short data records with longer sequence, finally obtain the comprehensive similarity of these two sequences above.This similarity had both contained the font information representated by two sequences, contained again the behavioral characteristics such as the speed of pen section corresponding to two sequences, therefore, it is possible to be applicable to the coupling of dynamic signature pen section preferably.

The object of the invention is to be achieved through the following technical solutions:

Based on a dynamic signature recognition methods for touch-screen mobile device, comprise the following steps:

The point of the whole signature pen section of step one, input user obtains point coordinate sequence; If the signature pen section of input is template, go to step three; Otherwise the signature pen section that input is described is signature sample, goes to step two;

The relevant information of step 2, reading signature template;

Step 3, carry out pre-service point coordinate sequence is converted to polar angle sequence angle [N] to described signature pen section, N+1 represents the number of point coordinate sequence mid point;

Step 4, described polar angle sequence carried out to extreme points extraction and extreme point concavity judges to obtain polar angle feature;

If the signature pen section of step 5 input is template, then preserves the relevant information of signature template, go to step eight; Otherwise, illustrate that the signature pen section of input is signature sample, go to step six;

Step 6, the polar angle feature of described signature template and the polar angle feature of signature sample being matched obtains signing the similarity of template and signature sample;

Step 7, described similarity and threshold value T to be contrasted, if described similarity is more than or equal to T, then recognition result to be signature sample be my signature, otherwise recognition result is signature sample is not my signature; Export recognition result;

Step 8, end.

Beneficial effect

Contrast existing signature recognition technology, the inventive method has the advantages such as cost is low, accuracy rate is high, fast operation.

Accompanying drawing explanation

Fig. 1 is the schematic flow sheet of embodiment of the present invention dynamic signature recognition methods.

Fig. 2 is embodiment of the present invention user's signature schematic diagram before treatment.

Fig. 3 is the schematic diagram after the sampling of embodiment of the present invention user's signature.

Fig. 4 is the schematic diagram after embodiment of the present invention user's signature interpolation.

Fig. 5 is embodiment of the present invention user's signature pen section extreme point schematic diagram.

Embodiment

Below in conjunction with drawings and Examples, the present invention is described in detail, also describe technical matters and the beneficial effect of technical solution of the present invention solution simultaneously, it is pointed out that described embodiment is only intended to be convenient to the understanding of the present invention, and any restriction effect is not play to it.

The present embodiment gives according to the inventive method and utilizes the touch-screen of terminal device to carry out dynamic signature knowledge method for distinguishing.

The present embodiment requires the condition possessed:

1. there is the mobile terminal device of touch-screen input

The step implemented:

The present embodiment specific algorithm flow process as shown in Figure 1, comprises the following steps:

The point of the whole signature pen section of step one, input user obtains point coordinate sequence; If the signature pen section of input is template, go to step three; Otherwise the signature pen section that input is described is signature sample, goes to step two;

The relevant information of step 2, reading signature template;

Step 3, carry out pre-service point coordinate sequence is converted to polar angle sequence angle [N] to described signature pen section, N+1 represents the number of point coordinate sequence mid point;

Step 4, described polar angle sequence carried out to extreme points extraction and extreme point concavity judges to obtain polar angle feature;

If the signature pen section of step 5 input is template, then preserves the relevant information of signature template, go to step eight; Otherwise, illustrate that the signature pen section of input is signature sample, go to step six;

Step 6, the polar angle feature of described signature template and the polar angle feature of signature sample being matched obtains signing the similarity of template and signature sample;

Step 7, described similarity and threshold value T to be contrasted, if described similarity is more than or equal to T, then recognition result to be signature sample be my signature, otherwise recognition result is signature sample is not my signature; Export recognition result;

Step 8, end.

Above-mentioned steps is summed up and can be divided into pre-service, polar angle feature extraction, polar angle characteristic matching and signature identification four part.Fig. 2 is user's signature before treatment, herein by with the signature shown in Fig. 2 exemplarily combination principle the result after each several part process is described.

One, Image semantic classification

The concrete steps of Image semantic classification are as follows:

1. obtain the point of the whole signature pen section of user's input, in view of the technology in the present invention realizes based on mobile terminal, not equifrequent between adjacent two points in the signature pen section obtained, therefore need first to process the point coordinate sequence got in this step, the time interval between every 2 is adjusted to a certain set time, as 50ms.

Without loss of generality, can adopt and adjust with the following method: using the time of signature pen section first some correspondence as start time, from start time point, generate a new virtual subscription point every 50ms.The generation method of virtual subscription point is: search two signature points nearest apart from this time point, carries out interpolation obtain virtual subscription point coordinate corresponding to this time point to its coordinate.

Especially, lift point corresponding in the time period for user, the equal assignment of its coordinate is (-1 ,-1), waits until subsequent treatment.So, in a pen section, the number of point can represent the average velocity of this section.After this step process, the result of user's signature as shown in Figure 3.

2. in pair sequence, user lifts point corresponding in a time period and carries out interpolation, and carries out filtering to whole signature pen section and disturb and noise to remove; 7 are removed heart mean filter (1,2,1,0,1,2,1);

Owing to directly cannot be lifted the movement locus in the time period by mobile device acquisition user, being the integrality ensureing user's signature pen section, needing the point to lifting in the time period to carry out interpolation.Consider user lift at every turn a process time shorter and between 2 straight line the shortest, the pen section that therefore the present invention's acquiescence carries out interpolation is straight-line segment and is at the uniform velocity.Concrete grammar is: lift point corresponding in the time period for user, finding its two ends coordinate is not (-1,-1) point, and using these 2 as two end points of straight-line segment carrying out interpolation, be that the point of (-1 ,-1) is distributed on this straight-line segment by coordinate according to the uniform length of this time period again.

The interference and noise that bring in signed data gatherer process can be reduced to the filtering of signature pen section.The present invention adopts and goes heart Mean Filtering Algorithm to carry out filtering, specifically uses at 7 and removes heart mean filter (1,2,1,0,1,2,1), by this wave filter filtering interfering and noise.Specific practice is described below: for sequence X=(X ₁, X ₂..., X _t..., X _n), make Y _t=(X _t-3+ 2*X _t-2+ X _t-1+ X _t+1+ 2*X _t+2+ X _t+3)/8, wherein X _trepresent t element in the sequence X before filtering, Y _trepresent t element in filtered sequence Y.For the equal assignment 0 of the element exceeding X sequence context, this step to the result of user's signature as shown in Figure 4.

3. normalization

(1) extract center of mass point M and the gold center of mass point M1 of sample S and template T respectively according to the point sequence of sample and template according to computing formula below, obtain Ms (xs, ys), M1s (x1s, y1s), Mt (xt, yt), M1t (x1t, y1t), wherein Ms and M1s is respectively center of mass point coordinate and the gold center of mass point coordinate of sample S, Mt and M1t is respectively center of mass point coordinate and the gold center of mass point coordinate of template T; The coordinate computing formula of center of mass point M and gold center of mass point M1 is as follows:

$x_M=\frac{{\mathrm{\Σ}}_{i=0}^N{x}_i}{N+1},y_M=\frac{{\mathrm{\Σ}}_{i=0}^N{y}_i}{N+1},i\∈[\mathrm{0,1,2},...,N];---\left[1\right]$

$x_{M_1}=\frac{{\mathrm{\Σ}}_{i=0}^{N_1}{x}_i}{N_1+1},y_{M_1}=\frac{{\mathrm{\Σ}}_{i=0}^{N_1}{y}_i}{N_1+1},i\∈[\mathrm{0,1,2},...,N_1].---\left[2\right]$

Wherein, (N+1) represents counting of whole signature pen section, N ₁=0.618*N.

(2) calculate the centroid distance of sample and template, computing formula is as follows:

Sample centroid distance $d_s=\sqrt{{(x_s-x_{1s})}^2+{(y_s-y_{1s})}^2}---\left[3\right]$

Template centroid distance $d_t=\sqrt{{(x_t-x_{1t})}^2+{(y_t-y_{1t})}^2}---\left[4\right]$

(3) according to the distance of M to M1, signature S is carried out normalizing to T.That is: the x coordinate of point each in sample sequence and y coordinate are multiplied by (ds/dt) simultaneously.

This step has only carried out convergent-divergent to signature, does not change the shape of signature.

4. polar coordinates conversion

The computing formula of limit is as follows:

$x_P=x_{\min }-|x_M-x_{M_1}|.---\left[5\right]$

$y_P=y_M+\frac{(x_P-x_M)(y_{M_1}-y_M)}{x_{M_1}-x_M}.---\left[6\right]$

Wherein, x _minrepresent the minimum value of x direction coordinate in the point of signature pen section.

The computing formula of footpath, pole and polar angle is as follows:

R _i＝[(x _i-x _P) ²+(y _i-y _P) ²] ^1/2， [7]

$A_i=\frac{(y_i-y_P)(x_M-x_P)-(x_i-x_P)(y_M-y_P)}{R_i{d}_{\mathrm{PM}}}.---\left[8\right]$

Wherein, x _iand y _isubmeter represents the coordinate of an i in x direction and y direction; d _pMcalculated by following formula:

$d_{\mathrm{PM}}=\frac{|x_P-x_M|}{|x_{M_1}-x_M|}d..---\left[9\right]$

Wherein, d represents centroid distance;

By above formula, obtain the limit (x of sample _ps, y _ps), the limit (x of template _pt, y _pt), the polar angle sequence anglet of the polar angle sequence angles [N] of sample and template [N '].Then, carried out to the polar angle sequence of sample and template at 7 and remove heart mean filter (1,2,1,0,1,2,1), to reduce the noise in sequence.

Two, polar angle feature extraction

For sample and template, from sequence, find all extreme points respectively, and record convexity corresponding to all extreme points.

The method of extreme point is looked for be: for polar angle sequence angle [N], if angle [i] >angle [i-1] and angle [i] >angle [i+1], then angle [i] is extreme point, and its convexity is convex; If angle [i] <angle [i-1] and angle [i] <angle [i+1], then angle [i] is extreme point, and its convexity is recessed.The sequence number of all extreme points in sample sequence is recorded in Es [] sequence together with head and the tail 2, and the convexity of corresponding extreme point is recorded in Fs [] sequence; Similarly, the sequence number of extreme points all in template sequence is recorded in Et [] sequence together with head and the tail 2, and corresponding extreme point convexity is recorded in Ft [] sequence.E [] and F [] segmentation signature pen section will be used in subsequent step, thus calculate similarity.The part extreme point extracted after this step process as shown in Figure 5.

Three, polar angle characteristic matching

Use different modes to split pen section according to extreme point, calculate the pen section similarity obtained with often kind of partitioning scheme, the mode choosing pen section similarity maximum is split signature.

This step completes further by following steps:

First the implication of each variable is explained:

CorDTW: overall similarity, initial value is 0;

I0: the starting point subscript of the extreme point sequence of the current sample pen section correspondence for coupling, initial value is 0;

Es (I0): the initial subscript of the current sample pen section for coupling;

I: the terminal subscript of the extreme point sequence of the current sample pen section correspondence for coupling, initial value is 1;

Es (I): the terminal subscript of the current sample pen section for coupling;

J0: the starting point subscript of the extreme point sequence of the current template pen section correspondence for coupling, initial value is 0;

Et (J0): the initial subscript of the current template pen section for coupling;

J: the terminal subscript of the extreme point sequence of the current template pen section correspondence for coupling, initial value is 1;

Et (J): the terminal subscript of the current template pen section for coupling;

Cor (Es (I0), Es (I), ET (J0), ET (J)): sample pen section angles [Es (I0)] ... angles [Es (I)] and template pen section anglet [Et (J0)] ... the pen section similarity of anglet [Et (J)].

Concrete steps are as follows:

(1) initialization initial value: I0=J0=0, I=J=1; CorDTW=0;

(2) if Es (I) is consistent with the convexity of Et (J), then forward to (3); Otherwise, forward to (6)

(3) CorDTW is upgraded by following formula:

CorDTW＝CorDTW+cor(Es(I0),Es(I),Et(J0),Et(J))*(Et(J)-Et(J0)+1)；

(4) if I exceedes the extreme point sequence length (i.e. J>=Et.length) that the extreme point sequence length (i.e. I>=Es.length) of sample or J exceed template, forward to (9);

(5) I0=I, J0=J, I=I+1, J=J+1; Forward to (2);

(6) if cor (Es (I0), Es (I+1), Et (J0), Et (J)) >=

Cor (Es (I0), Es (I), Et (J0), Et (J+1)), then CorDTW=CorDTW+

Cor (Es (I0), Es (I+1), Et (J0), Et (J)) * (Et (J)-Et (J0)+1); Otherwise CorDTW=CorDTW+

cor(Es(I0),Es(I),Et(J0),Et(J+1))*(Et(J+1)-Et(J0)+1)；

(7) if I exceedes the extreme point sequence length that the extreme point sequence length of sample or J exceed template, forward to (9);

(8) if cor is (Es (I0), Es (I+1), Et (J0), Et (J)) >=cor (Es (I0), Es (I), Et (J0), Et (J+1)), then I0=I+1, J0=J, I=I+1+1, J=J+1, forward to (2); Otherwise I0=I, J0=J+1, I=I+1, J=J+1+1, forward to (2);

(9)CorDTW＝CorDTW/(Et(J)+1)；

Wherein, following method is adopted to calculate the similarity cor (Es (I0), Es (I), Et (J0), Et (J)) of each section:

For the sample sequence angles [Es (I0)] for coupling ... angles [Es (I)] and template sequence anglet [Et (J0)] ... anglet [Et (J)], because they may be Length discrepancy, the present invention adopts the algorithm of following similarity, and idiographic flow is as follows:

Invention provides for the similarity on three levels: be the similarity that two isometric sequences are obtained by similarity formulae discovery at the bottom, it is long-pending that the computing method of this similarity are that two isometric sequences do the 2-norm of inner product then divided by these two sequences, claims this similarity to be the matching similarity of two isometric sequences; Be for two Length discrepancy sequences in middle layer, with the calculating carrying out repeatedly said matching similarity above compared with short data records and multiple parts of longer sequence, and get the similarity of its maximal value as these two Length discrepancy sequences, claim this similarity to be direct similarity; Be only for given two sequences at most top layer:

Angles [Es (I0)] ... angles [Es (I)] and anglet [Et (J0)] ... anglet [Et (J)] calculates similarity, in order to reduce the interference of sample sequence head and the tail part further, multiple repairing weld should be carried out to sample sequence, obtain different secondary samples by the head and the tail of reduced sample sequence at every turn, and the multiple secondary samples obtained are carried out the calculating of direct similarity with template sequence respectively, finally get maximal value and the length ratio be multiplied by compared with short data records and longer sequence as sample sequence angles [Es (I0)] ... angles [Es (I)] and template sequence anglet [Et (J0)] ... the final similarity of anglet [Et (J)].

Four, signature identifies

After obtaining the similarity of signature sample and template, by checking whether similarity exceedes the threshold value T of setting in advance, judges whether signature sample and template come from the hand of same people.

As preferably, T=0.83; That is: if the similarity of signature sample and template is more than or equal to 0.83, then think that signature sample and template come from the hand of same people; Otherwise think that signature sample is generated by different two people from template.

Above-described specific descriptions; the object of inventing, technical scheme and beneficial effect are further described; be understood that; the foregoing is only specific embodiments of the invention; the protection domain be not intended to limit the present invention; within the spirit and principles in the present invention all, any amendment made, equivalent replacement, improvement etc., all should be included within protection scope of the present invention.

Claims (10)

1., based on a dynamic signature recognition methods for touch-screen mobile device, it is characterized in that: comprise the following steps:

The point of the whole signature pen section of step one, input user obtains point coordinate sequence; If the signature pen section of input is template, go to step three; Otherwise the signature pen section that input is described is signature sample, goes to step two;

The relevant information of step 2, reading signature template;

Step 3, carry out pre-service point coordinate sequence is converted to polar angle sequence angle [N] to described signature pen section, N+1 represents the number of point coordinate sequence mid point;

Step 4, described polar angle sequence carried out to extreme points extraction and extreme point concavity judges to obtain polar angle feature;

If the signature pen section of step 5 input is template, then preserves the relevant information of signature template, go to step eight; Otherwise, illustrate that the signature pen section of input is signature sample, go to step six;

Step 6, the polar angle feature of described signature template and the polar angle feature of signature sample being matched obtains signing the similarity of template and signature sample;

Step 7, described similarity and threshold value T to be contrasted, if described similarity is more than or equal to T, then recognition result to be signature sample be my signature, otherwise recognition result is signature sample is not my signature; Export recognition result;

Step 8, end.

2. a kind of dynamic signature recognition methods based on touch-screen mobile device according to claim 2, is characterized in that: before described point coordinate sequence is converted to polar angle sequence, needs first signature sample to be normalized according to signature template.

3. a kind of dynamic signature recognition methods based on touch-screen mobile device according to claim 3, is characterized in that: described normalization obtains x and the y coordinate after described signature sample normalization by x and the y coordinate of described signature sample being multiplied by respectively the centroid distance of sample with the ratio of the centroid distance of template.

4. according to the arbitrary described a kind of dynamic signature recognition methods based on touch-screen mobile device of right 1-3, it is characterized in that: described pre-service also comprises to be carried out process to described point coordinate sequence and make between every 2 as Fixed Time Interval, as 50ms, then point corresponding in the time period is lifted to user and carry out interpolation, finally carried out to whole signature pen section at 7 and remove heart mean filter.

5. a kind of dynamic signature recognition methods based on touch-screen mobile device according to claim 1, is characterized in that: describedly carry out extreme points extraction and extreme point concavity to described polar angle sequence and judge to obtain polar angle feature, and discrimination standard is as follows:

Polar angle as fruit dot i is not only greater than the polar angle of an i-1 but also is greater than the polar angle of an i+1, then putting i is extreme point, and its convexity is convex;

Polar angle as fruit dot i is not only less than the polar angle of an i-1 but also is less than the polar angle of an i+1, then putting i is extreme point, and its convexity is recessed;

Polar angle is characterized as extreme point sequence and convexity extreme point sequence, and described extreme point sequence comprises the head and the tail 2 of polar angle sequence and the sequence number of all extreme points, and this sequence number is for arrange from low to high.

6. a kind of dynamic signature recognition methods based on touch-screen mobile device according to claim 1, is characterized in that: the relevant information of described signature template comprises centroid distance and polar angle feature.

7. a kind of dynamic signature recognition methods based on touch-screen mobile device according to claim 1, it is characterized in that: described the polar angle feature of described signature template and the polar angle feature of signature sample are matched to obtain signing the similarity of template and signature sample, matching process is: use different modes to split pen section according to extreme point, calculate the pen section similarity obtained with often kind of partitioning scheme, the mode choosing pen section similarity maximum is split signature.

8. a kind of dynamic signature recognition methods based on touch-screen mobile device according to claim 7, is characterized in that: described matching process arthmetic statement is as follows:

(1) initialization initial value: I0=J0=0, I=J=1; CorDTW=0;

(2) if Es (I) is consistent with the convexity of Et (J), then forward to (3); Otherwise, forward to (6)

(3) CorDTW is upgraded by following formula:

CorDTW＝CorDTW+cor(Es(I0),Es(I),Et(J0),Et(J))*(Et(J)-Et(J0)+1)；

(4) if I exceedes the extreme point sequence length (i.e. J>=Et.length) that the extreme point sequence length (i.e. I>=Es.length) of sample or J exceed template, forward to (9);

(5) I0=I, J0=J, I=I+1, J=J+1; Forward to (2);

(6) if cor is (Es (I0), Es (I+1), Et (J0), Et (J)) >=cor (Es (I0), Es (I), Et (J0), Et (J+1)), then CorDTW=CorDTW+cor (Es (I0), Es (I+1), Et (J0), Et (J)) * (Et (J)-Et (J0)+1); Otherwise CorDTW=CorDTW+cor (Es (I0), Es (I), Et (J0), Et (J+1)) * (Et (J+1)-Et (J0)+1);

(7) if I exceedes the extreme point sequence length that the extreme point sequence length of sample or J exceed template, forward to (9);

(8) if cor is (Es (I0), Es (I+1), Et (J0), Et (J)) >=cor (Es (I0), Es (I), Et (J0), Et (J+1)), then I0=I+1, J0=J, I=I+1+1, J=J+1, forward to (2); Otherwise I0=I, J0=J+1, I=I+1, J=J+1+1, forward to (2);

(9)CorDTW＝CorDTW/(Et(J)+1)；

Wherein, the implication of above-mentioned symbol is as described below:

CorDTW: overall similarity, initial value is 0;

I0: the starting point subscript of the extreme point sequence of the current sample pen section correspondence for coupling, initial value is 0;

Es (I0): the initial subscript of the current sample pen section for coupling;

I: the terminal subscript of the extreme point sequence of the current sample pen section correspondence for coupling, initial value is 1;

Es (I): the terminal subscript of the current sample pen section for coupling;

J0: the starting point subscript of the extreme point sequence of the current template pen section correspondence for coupling, initial value is 0;

Et (J0): the initial subscript of the current template pen section for coupling;

J: the terminal subscript of the extreme point sequence of the current template pen section correspondence for coupling, initial value is 1;

Et (J): the terminal subscript of the current template pen section for coupling;

Cor (Es (I0), Es (I), ET (J0), ET (J)): pen section angles [Es (I0)] in sample polar angle sequence ... angles [Es (I)] and pen section anglet [Et (J0)] in template polar angle sequence ... the pen section similarity of anglet [Et (J)].

9. a kind of dynamic signature recognition methods based on touch-screen mobile device according to claim 8, it is characterized in that: the similarity cor (Es (I0) of described each section, Es (I), Et (J0), Et (J)) be similarity on three levels, concrete computation process is as follows:

The similarity that two isometric sequences are obtained by similarity formulae discovery at the bottom, it is long-pending that the computing method of this similarity are that two isometric sequences do the 2-norm of inner product then divided by these two sequences, claims this similarity to be the matching similarity of two isometric sequences;

Be for two Length discrepancy sequences in middle layer, with the calculating carrying out repeatedly said matching similarity above compared with short data records and multiple parts of longer sequence, and get the similarity of its maximal value as these two Length discrepancy sequences, claim this similarity to be direct similarity;

Be only for given two sequence: angles [Es (I0)] at most top layer ... angles [Es (I)] and anglet [Et (J0)] ... anglet [Et (J)] calculates similarity, in order to reduce the interference of sample sequence head and the tail part further, multiple repairing weld should be carried out to sample sequence, obtain different secondary samples by the head and the tail of reduced sample sequence at every turn, and the multiple secondary samples obtained are carried out the calculating of direct similarity with template sequence respectively, finally get maximal value and the length ratio be multiplied by compared with short data records and longer sequence as sample sequence angles [Es (I0)] ... angles [Es (I)] and template sequence anglet [Et (J0)] ... the final similarity of anglet [Et (J)].

10. a kind of dynamic signature recognition methods based on touch-screen mobile device according to claim 1, is characterized in that: described T=0.83.