JP6820580B2 - Pen input personal recognition method - Google Patents

Description

The present invention relates to a pen input personal authentication method. More specifically, the present invention relates to a method of authenticating whether or not a person is a person based on handwriting information input to a plane by an electronic pen.

There is an increasing demand for personal identification technology in the field of "security" in order to prevent unauthorized operations on information terminals and unauthorized intrusion into buildings. Personal identification methods include (1) those that depend on personal memory such as passwords and PINs, (2) those that depend on personal property such as driver's licenses and credit cards, and (3) individuals such as signatures and fingerprints. (4) There are those that depend on the biological characteristics of the individual such as fingerprints and retina (see Table 1). Of these methods, (3) and (4) are called biometrics methods. As the conventional personal identification method, the above methods (1) and (2) are known.

However, methods that rely on possessions and stored information are not always perfect in recent years when computer crimes such as the knowledge, theft, and counterfeiting of third parties by fraudulent methods are increasing, and these conventional methods are complemented. The biometrics method is expected as a substitute for it. There is a demand for a biometrics-type personal identification method that is a relatively simple method for the person to be identified and that can perform personal identification at low cost.

As an individual identification method based on the above biometrics method, for example, there is a pen input personal authentication method using a dynamic programming method (hereinafter, referred to as “DP method”) according to Patent Document 1. This recognizes the handwriting information when signing a tablet with an electronic pen as (1) position trajectory information of the pen tip on the tablet plane and (2) pressure trajectory information, and these are information as necessary. After compression, the evaluation function created for this purpose is used to calculate the distance between the person's template prepared in advance, and the obtained calculated value is compared with a preset threshold to sign the signature. It authenticates whether or not the person who made the request is the person himself / herself. As a result, not only the visible handwriting but also the writing pressure is used, so that it is very difficult for a third party to imitate the signature.

The number of the person's template (signature of the person to be registered) used for personal authentication at this time is arbitrary, but if the number increases, more accurate authentication becomes possible, while the identification target at the time of signature registration. There is a problem that the burden on the person increases. Therefore, there is a demand for a technique for reducing the number of templates as much as possible, preferably three or less.

For example, if the number of templates is reduced in the center of gravity method, for example, when the position of the center of gravity overlaps the signature character, the distance and angle from the center of gravity to each point on the locus of the overlapped character fluctuate greatly, so that it is stable. There is a problem that it becomes difficult to create a template.

Furthermore, in order to use the signature written by the person as a template for authentication, instead of using a template for all the person's signatures, a template is used that meets certain conditions, and a template that does not meet certain conditions is not used as a template. However, the condition is also required to be optimized by reducing the number of templates.

In addition, as the number of templates decreases, a more elaborate method is required for the judgment at the time of authentication.

Patent No. 4603675

An object of the present invention is to enable accurate personal authentication with a smaller number of templates.

In order to achieve the above object, the present invention uses a center of gravity method (hereinafter referred to as "center of gravity method") using vector information from the position of the center of gravity of handwriting information when signing a tablet with an electronic pen. Data analysis is performed using a virtual center of gravity that is moved by a certain non-zero distance in the vertical direction from the plane.

Further, according to the present invention, at the stage of registering a template, the DP distance between each signature by the person is obtained, and the DP distance is registered as a template only when a certain condition is satisfied.

Further, according to the present invention, when there are three templates in the authentication stage, a triangular pyramid composed of a signature to be authenticated, three DP distances between each template, and three distances between each template. The volume is calculated, and if it is less than or equal to a certain value, the signature to be authenticated is authenticated as the person's.

Then, when the templates are aggregated into one, for example, three signatures are given by thinning out or adding points on the handwriting trajectory based on a certain rule in the algorithm for calculating the DP method. Create one representative signature template from the trajectory of.

That is, the present invention provides the following method.

First, the present invention is a pen input personal authentication method for authenticating whether or not a person is a person based on handwriting information output from an input device including trajectory information drawn on a plane by the pen tip.
A step of taking in the coordinates and pressure information of the pen tip as the handwriting information in chronological order,
The step of finding the position of the center of gravity of the entire handwriting from the captured coordinates,
A step of moving the obtained center of gravity to a position away from the plane of the handwriting by a certain distance and using it as the center of gravity for calculation.
Steps to find the vector from the position of the center of gravity for calculation to each point of the handwriting coordinates,
Steps to obtain the length component and angle component of the vector,
Dynamic using an evaluation function that is the product of the difference between the length component, angle component, and pen pressure information in the person's template prepared in advance, and the corresponding length component, angle component, and pen pressure information of the signature to be collated. The step of measuring the DP distance between the signature to be collated and the template of the person prepared in advance by the programming method,
Provided is a pen input personal authentication method including a step of identifying whether or not the signature is the person's signature by comparing the DP distance measured by the evaluation function with a threshold value prepared in advance.

Further, the present invention is a pen input personal authentication method for authenticating whether or not a person is a person based on handwriting information output from an input device including trajectory information drawn on a plane by the pen tip.
A step of capturing the coordinates and pressure information of the pen tip as the handwriting information,
The step of finding the position of the center of gravity of the entire handwriting from the captured coordinates,
The step of finding the vector from the position of the center of gravity to each point of the coordinates of the handwriting,
Steps to find the length component and angle component from the vector,
The step of finding the distance by the DP method between two of three or more signatures of the person, and
In order to determine the signature to be registered as a template of the person, select three of the DP distances between those signatures, obtain the area of the triangle formed by the three DP distances, and the area is constant. If it is within the range, the template decision step to be adopted for the template and
An evaluation function that is the product of the respective differences between the length component, the angle component, and the pen pressure information of the signature to be collated with the length component, the angle component, and the pen pressure information in the person's template determined in the template determination step. The step of measuring the DP distance between the signature to be collated and the template of the person prepared in advance by the dynamic programming method using
Provided is a pen input personal authentication method including a step of identifying whether or not the signature is the person's signature by comparing the DP distance measured by the evaluation function with a predetermined threshold value.

Further, the present invention is a pen input personal authentication method for authenticating whether or not a person is a person based on handwriting information output from an input device including trajectory information drawn on a plane by the pen tip.
A step of capturing the coordinates and pressure information of the pen tip as the handwriting information,
The step of finding the position of the center of gravity of the entire handwriting from the captured coordinates,
The step of finding the vector from the position of the center of gravity to each point of the coordinates of the handwriting,
Steps to find the length component and angle component from the vector,
The step of finding the distance by the DP method between two of three or more signatures of the person, and
In order to determine the signature to be registered as a template of the person, select three of the DP distances between those signatures, obtain the area of the triangle formed by the three DP distances, and the area is constant. If it is within the range, the template decision step to be adopted for the template and
It is the product of the difference between the length component, the angle component, and the pen pressure information in the person's template determined in the template determination step, and the corresponding length component, the angle component, and the pen pressure information of the signature to be collated. A step to measure the DP distance between the signature to be collated and the template of the person prepared in advance by the dynamic programming method using the evaluation function, and
Provided is a pen input personal authentication method including a step of identifying whether or not the signature is the person's signature by comparing the DP distance measured by the evaluation function with a predetermined threshold value.

Further, the present invention is a pen input personal authentication method for authenticating whether or not the person is the person based on the handwriting information output from the input device including the locus information drawn on the plane by the pen tip.
A step of capturing the coordinates and pressure information of the pen tip as the handwriting information,
The step of finding the position of the center of gravity of the entire handwriting from the captured coordinates,
The step of finding the vector from the position of the center of gravity to each point of the coordinates of the handwriting,
Steps to find the length component and angle component from the vector,
An intermediate signature trajectory between two signatures by thinning out or adding coordinate points based on certain rules during analysis by the DP method between two of each of the three or more signatures. Intermediate trajectory determination step to find
A template determination step that seeks a single representative signature template by repeating the intermediate trajectory determination step,
It is the product of the difference between the length component, the angle component, and the pen pressure information in the person's template determined in the template determination step, and the corresponding length component, the angle component, and the pen pressure information of the signature to be collated. A step to measure the DP distance between the signature to be collated and the template of the person prepared in advance by the dynamic programming method using the evaluation function, and
Provided is a pen input personal authentication method including a step of identifying whether or not the signature is the person's signature by comparing the DP distance measured by the evaluation function with a predetermined threshold value.

According to the present invention, when extracting the handwriting information of the person from the signature input to the tablet by the electronic pen while obtaining the center of gravity of the entire signature and comparing this with the template of the person prepared in advance, the calculation is performed. A smaller number of templates can be used to accurately determine whether or not the person is the person.

The flow diagram of the algorithm of the personal authentication method of this invention is shown. Indicates x _min , x _max , y _min , y _max for the entire signature. The standardized center of gravity position G obtained from the entire signature is shown. The vector representation from the center of gravity position G to the dots is shown. The flow chart up to the type 1 error output is shown. The triangular pyramid conceived in the third embodiment of the present invention is shown. It is a diagram which shows the DP matching method when the signature data time is the same and the trajectory is very close. It is a diagram which shows the state of DP matching when the time of two signature data is different. It is a table which shows the relationship between two signature data obtained as a result of DP matching.

According to the present invention, vector information including a length component and an angle component of a vector from the position of the center of gravity obtained from the handwriting information to each point of the handwriting coordinates, and pressure information at each point of the handwriting coordinates are prepared in advance. The following evaluation function D is applied to the template information of the person who made the request.
(In the formula, θ _i is the angle information of the vector from the position of the center of gravity of the handwriting information to each handwriting coordinate (x _i , y _i ), and η _l is each handwriting coordinate (x _i , y _i ) from the handwriting information obtained from the person's template. It is the angle information of the vector to x _l , y _l ), d is a function of the input pressure information p and the corresponding template pressure information q, and ρ is the length Δf of the above vector and it. is a function of the length △ g of the vector of the corresponding template, S is a positive integer, i _s, l _{s, s} uses an index integer.).

Specifically, for some signature information collected by the tablet, each point (x (t _i ), y (t _i )) and pen pressure (p (t)) obtained as the signature information on the tablet. Create (x (t _i ), y (t _i ), p (t _i )) data from the information related to _i )). Then, using (x (t _i ), y (t _i )), the standardized centroid position G of the entire signature or the computational centroid located a certain distance away from the signature plane from that centroid position. Find the position Gc, find the vector data from this G or Gc to each point (x (t _i ), y (t _i ), p (t _i )) of the signature position, and find the angle of the vector to each point, It is decomposed into three feature quantities of length and pen pressure information. Then, DP matching is performed on two pairs of signature information of some persons collected in advance, and the value of the evaluation function D is obtained. This value will be called the DP distance.

As a threshold value for determining personal authentication, for example, the signature information obtained from some DP distance values obtained by performing DP matching on some signature information of the person collected in advance as described above. If there are five, the average value of the third, fourth, and fifth values excluding the maximum value and the second largest value is taken and used as the threshold value Th. In addition, as the judgment value at the time of the person's judgment, the judgment value c · Th optimized for the person who signed the signature by the coefficient c which changes in increments of 0.1 within the range of 0.5 to 2.0 set for each person who signed the signature. Can be used. In order to reduce the type 1 error, DP matching is performed with the input multiple signature information of the person and some of the above-mentioned templates of the person, and the minimum value of the obtained DP distance D is D. If D _min ≤ c · Th with respect to (= D _min ), the person is determined to be the person. For type 2 errors, as with the above type 1 errors, DP matching is performed for multiple signatures of others and templates of some individuals, and when D _min > c · Th, it is judged to be a false writing. Reject. If there are three signature information obtained, for example, three DP distances between the three are obtained and the maximum value is set to the threshold value Th, or twice the minimum value (for example) is set as the threshold value. You can think of various options, such as whether to set Th or the larger of them to the threshold Th.

More specifically, the present invention is a pen input personal authentication method for authenticating whether or not a person is a person based on handwriting information output from an input device including trajectory information drawn on a plane by the pen tip. A step of taking in the pen tip coordinates and pen pressure information at the coordinates as handwriting information, a step of obtaining the position of the center of gravity of the entire handwriting from the coordinates, a step of obtaining a vector from the position of the center of gravity toward each point of the handwriting coordinates, and the above. A step of obtaining the length and angle components from the vector, and a step of analyzing the length and angle components by a dynamic programming method using an evaluation function including a predetermined weighting in order to collate the length and angle components with a prepared personal template. The present invention provides a pen input personal authentication method including a step of comparing the result of the analyzed evaluation function with a preset threshold value to determine whether or not the person is the person himself / herself.

The present invention also includes, among the values of the plurality of evaluation functions obtained by calculating the value of the evaluation function with other templates for each of the templates when creating a plurality of templates from the signature of the person himself / herself. The average value of the evaluation function values obtained by removing at least the maximum value and the second largest value is used as the threshold value, and a plurality of the evaluations are performed between the input handwriting information and the template created from the signature of the person. It also provides a pen input personal authentication method characterized in that the value of a function is calculated and the smallest value among the obtained values of the plurality of evaluation functions is collated with the threshold value.

The pen and tablet or the electronic device for input referred to here are a general term for an input device for obtaining handwriting information necessary for carrying out the present invention. Pens and tablets are representative of devices that can obtain real-time and dynamic data of handwriting, but for example, a simple short stick and a camera that captures its movement can also be used. Examples of the pen and tablet that can obtain the position information of the tip of the pen and the pen pressure information include a general tablet-type personal computer and a commercially available tablet made by Wacom.

The evaluation function referred to here is a type of handwriting information multiplied by a weighting based on different handwriting information. For example, the handwriting information is multiplied by a function of the pen pressure information as a weight, or the pen pressure information obtained from the handwriting information is the length and angle which are vector components from the center of gravity coordinates obtained from the coordinate information to the coordinates of the handwriting. The evaluation function obtained by weighting the function of (that is, multiplying the weighting function having each component of the vector as an element by the function of the corresponding pen pressure information).

The method of the present invention can be practiced in many ways. It is most common to use a computer including a personal computer and connect a tablet to it, but it is also possible to use a computer's liquid crystal display or touch panel instead of the tablet. In addition, when using a simple short stick and a camera, it can be carried out by connecting a computer and a camera. Further, the method of the present invention can also be carried out on a computer by reading an execution program stored in a computer-readable storage medium such as a ROM, RAM, or a hard disk and executing the execution program. Further, the method of the present invention can also be provided in the form of a computer or a program incorporated in a device and an electronic device controlled by the computer. For example, it can be incorporated into automatic cash payment machines, access control devices, and any other device or electronic device that requires personal authentication.

FIG. 1 shows a flow chart of the algorithm of the signature verification method according to the present invention. In the signature verification method according to the present invention, the signature input 101 is divided into the position information on the tablet at the tip of the pen and P steps (P is not particularly limited as long as it is a positive integer value), and the pen pressure is detected. Use a combination of an electronic pen and a tablet that can be used. Data feature extraction 102 is performed from the signature input 101 obtained by signing this tablet. The data obtained by the feature extraction 102 includes position information and pen pressure information on the tablet at the tip of the pen. The pattern matching 104 between the feature extraction 102 data and the template 103 prepared in advance is performed. Next, the result (value of the evaluation function) obtained from the pattern collation 104 is compared 105 with the threshold value set in advance using the template 103, and after determining whether or not the person is the person, the result output 106 is performed. ..

The data input from the tablet includes the two-dimensional position information x (t _i ), y (t _i ) and the pen pressure information p (t _i ) at that time. this,
Write. I also represents the number of dot data obtained from the tablet. Here, t _i is a parameter for distinguishing one point from the next, and can be regarded simply as an index indicating the order of points, but it is usually convenient to think of it as representing a time value.

Next, the operation for extracting the features of the input data will be described below. As step 1, the center of gravity, which is a characteristic point for the entire signature, is obtained. If the desired center of gravity is G (X _g , Y _g ),
Here, x _min 301 and x _max 302 are the minimum value and the maximum value in the x coordinate of the obtained signature, respectively. Similarly, y _min 303 and y _max 304 are the minimum and maximum values in the y coordinate of the obtained signature, respectively (see FIG. 2). Moreover, these equations of number 4 are standardized by the value of α. The center of gravity obtained by these is the center of gravity G401 as shown in FIG. Here, as an example, since α = 1000, the vertical and horizontal values are standardized at 1000. Then, in the normalized coordinates, as z-coordinate direction a distance Z _g, for example 100 from the signature of the plane (xy coordinate plane), defines the calculation centroid Gc402.

Next, as step 2, the vector from the obtained center of gravity Gc402 to each coordinate is obtained. First, the vector connecting the two points of each dot is obtained from the calculation center of gravity Gc402 (see FIG. 4). The center of gravity G and the center of gravity Gc for calculation appear to overlap in order to be represented in a plan view. here,
And.

The angle and length of the converted vector using the center of gravity are obtained by the following equations 6 and 7.
Here, θ (t _i ) represents the angle of the vector, and Δf (t _i ) represents the length of the vector. In this way, the following features are calculated while maintaining the time series information.
Here, p (t _i ) indicates the pen pressure at the tip of the vector.

Similarly, the center of gravity of the feature-extracted vector is obtained and compared with the feature-extracted template vector. As a feature of the template
Is getting. In order to calculate the distance between them, the following number tens is used.
Here, d (p (t _i ), q (t _l )) is the weight for considering the pen pressure information, and ρ (Δf (t _i ), Δg (t _l )) is the length of the vector. Represents the weight to consider.

As a result, the evaluation function at each time is obtained. To define the total distance between two signatures, the sum of i and l should be taken. However, while i = 1,2, ..., I, l = 1,2, ..., L, and generally I ≠ L, so find the sub-indexes i _s and l _s , and s = 1. , 2, ..., It is necessary to take the sum up to S. Here, the sub-index i _s and l _s is required to satisfy the following relation.
Also, the converted data (θ (t _i ), Δf (t _i ), p (t _i )), (η (t _l ), Δf (t _l ), p (t _l )) may not be skipped. Because it is desirable
To restrain. To summarize the above points, the distance between the two can be given by the following equation.
However,
Is fixed.

It is calculated by the dynamic programming method (DP method) from the nature of the minimization problem of the above number 13. In dynamic programming, the solution of sequential minimization shown below is the solution of global minimization.
here,
And. The DP distance D (I, L), which is the final value obtained here, is compared with the threshold value for collation.

Next, a method of determining the template will be described. First, as shown in FIG. 5, m ₀ signatures of the person are obtained (601). Then, this for creating template signature m ₁ piece 601 (here, extracted with a rate of one to five), distributed to experimental signature m ₂ pieces for calculating the type 1 error (602 and 603). For each of these signatures, the above-mentioned center of gravity point or calculation center of gravity point is calculated and the feature amount is calculated (604,605,611,612). Next, the round-robin of _one template creation signature m is performed. Make all pairs (10 pairs if m ₁ is 5) and perform DP matching between the pairs (606,701), resulting in another signature for each m ₁ signature. Three templates are selected as templates in ascending order of the sum of the values of the DP distances with and (607,702,703).

Furthermore, DP matching is performed between the three templates obtained in this way and the remaining (m _1-3 ) templates. Desirably, the 3rd to 5th positions excluding the 1st and 2nd places are extracted from the maximum value of the DP distance, and the average value thereof is used as the threshold value (802). The reason for doing this is to ignore signatures that are far from the template and make appropriate threshold decisions (610,803).

Next, in order to calculate the type 1 error, DP matching is performed between 3 templates and ₂ experimental signatures as shown in FIG. 5 (613). Since there are three templates here, three DP distance values can be obtained for each experiment signature. The smallest evaluation function D (hereinafter referred to as “D _min ”) among the three is compared with the above-mentioned threshold value (614). At this time, if the above-mentioned threshold value is set to Th, the following two conditions of equations 18 and 19 can be obtained by using the coefficient c in which 0.5 to 2.0 is in 0.1 increments.
When the equation of Eq. 18 is satisfied, the entered signature will be rejected as the signature of another person. Further, when the equation of Eq. 19 is satisfied, the entered signature is accepted as the signature of the person himself / herself. As described above, it is possible to determine the type 1 error (615).

Next, the type 2 error will be described. The difference from the type 1 error described above is that signatures written by others are used instead of m ₂ experimental signatures in FIG. After that, DP matching can be performed with 3 templates as in the case of type 1 error. The evaluation function D _min , which is the minimum value among the obtained three evaluation functions, and the threshold value Th are compared. If the inequality of Eq. 18 is satisfied, the entered signature will be rejected as false. Further, when the inequality of the equation 19 is satisfied, the entered signature is accepted as the signature of the person himself / herself.

In the present invention, it is important to set an appropriate threshold value in order to prevent a type 1 error (error of refusal of the person) and a type 2 error (error of acceptance of another person) at the time of signature verification. When the threshold is set strictly, the occurrence of type 2 errors decreases, but the occurrence of type 1 errors increases. On the other hand, when the threshold value is set loosely, the occurrence of type 1 errors decreases, but the occurrence of type 2 errors increases.

As a second embodiment of the present invention, a method of determining a template different from the above will be described, but this is only an example. For example, get three signatures from you. Then, the DP distance between the three pairs of the three signatures is obtained by using the DP matching method as described above. Then, the area A of the triangle is defined as follows, where the three DP distances are a, b, and c.
However, as a premise, as a condition for the triangle to be established, the largest DP distance is set to a, and then
If this condition is not met, it is necessary to urge the collection of at least another signature and prepare three signatures that satisfy this condition. The reason why the formation of the triangle is required is that the DP distance follows a different rule from the Euclidean distance, but the reason is that the trajectory of each signature does not have an appropriate distribution (unnaturally biased). Can be.

Then, when the value of A is η ₂ or more and η ₁ or less, these three signatures are templated, accepted, registered, and used for the subsequent signature authentication. Here, η ₂ and η ₁ are constants that determine the lower limit and the upper limit, respectively, and are experimentally optimized. In other words, if the area of this triangle is extremely small, that is, if it is smaller than η ₂ , there is a risk that it will be a rigid signature of the gold nail style, for example, the name "Tanaka Ichi" written in a group of vertical and horizontal sides. Because there is, encourage them to write their name softer and more naturally. On the contrary, if the area of this triangle is extremely large, that is, if it is larger than η ₁ , there is a risk that, for example, you may write in your signature in a different stroke order, or write extremely early or late ( (Type 2 errors can be large), you can encourage them to write their name in a consistent manner as they normally would, and then collect signatures.

Further, as another criterion, it can be obtained that the value obtained by dividing the area of this triangle by the above s is larger than the experimentally obtained constant of η ₃ . A / s is the radius of the inscribed circle, and its extremely small value means that it is a collapsed triangle. That is, it means that there is one signature that is extremely far apart from the other two signatures (for example, if there is a missed signature), so in order to eliminate such signatures, This is a standard set to request rewriting.

As a further condition, it is also possible to further satisfy a <min {a ₁ xt _s xα ₂ , b ₁ }, 1 <= α ₂ for each of a, b, and c. Here, a can be replaced with b or c. This seeks to eliminate large variations in the time it takes to sign. a ₁ is the number of data points per unit time of handwriting obtained from the input device, and is generally 50 points per second or 200 points per second. t _s is the time required for signing, and can be defined as, for example, t _s = arg min {a + b, a + c, b + c}. The signature length b ₁ can be arbitrarily set as the upper limit value. For example, assuming that a ₁ = 200, if we want to eliminate variations in the time required for signing for 6 seconds or more, b ₁ = 2400.

As an additional condition, if an extreme jump occurs in the middle of calculating the DP distance, it is assumed that a problem such as a difference in stroke order has occurred, even if the cumulative DP distance is less than the threshold value. , It is also possible to refuse to register the signature set as a template as if an error occurred.

As a further different third embodiment, the conditions at the time of authentication can be changed. For example, when there are three templates, the DP distance between the template and the signature to be authenticated is calculated. That is, there are three DP distances. Then, since there are also three DP distances for three pairs among the three templates, the volume of the triangular pyramid can be considered from the total of six DP distances. That is, as shown in FIG. 6, it is possible to consider the DP distances u, v, w between the U, V, W, which are the DP distances between the three templates, and the signatures to be authenticated, which are represented by circles. it can. In other words, if you define it as follows,
At that time, the volume Vol of the triangular pyramid can be expressed by the following equation 23.

In this case as well, it is a condition that the inside of the radical symbol of Equation 23 is positive. It is considered that this condition is not a very high hurdle condition if the condition of the number 21 when registering the signature is satisfied.

Then, if the value of this Vol is less than or equal to a certain threshold value (for example, a constant) that can be optimized experimentally, it is accepted as the signature of the person, and if it is larger than the threshold value, it is rejected as a false writing. Can be done. Here, if the signature to be authenticated exactly matches the template, the calculation of Equation 22 may fail, but such a possibility is extremely low and it is considered that there is no problem in actual authentication.

At this time, if there are two templates, a triangle and its area can be considered instead of the triangular pyramid, and if there are four templates, the quadrangular pyramid and its volume can be considered in the same manner.

By calculating the volume of the triangular pyramid in this way, applying a threshold value to it, and determining the authenticity of the signature, it is possible to carry out more accurate personal authentication even with a smaller number of templates.

The fourth embodiment of the present invention will be described below. The ultimate reduction in the number of templates is to make the number of templates one. This is a challenging task, but it's not entirely impossible if you create that single template based on multiple templates. For example, the following method can be considered.

7 and 8 show that when there are two signatures, the time required for each signature can be regarded as the same under a simple rule (FIG. 7) and different (FIG. 8), respectively. It shows. If the signing time is the same under some rules (for example, within plus or minus half of the sampling interval to absorb the error), then, for example, at a given time interval If it is considered that each point on the sampled signature also corresponds under a certain rule, the average value can be calculated between the corresponding points. Here, the explanation is made using the xy coordinate system, but the same is true even if this becomes a polar coordinate system.

In FIG. 8, the time required for signature is different, and the sampling interval for signature is 1, 12 for signature data 1 and 15 for signature data 2. The actual signature is much larger (a typical tablet can sample 200 points per second, and it usually takes a few seconds to complete one signature), but it's dare to understand the principle. An example is shown.

That is, in the case shown in FIG. 8, it is necessary to make the lengths of the signature data 1 and the signature data 2 uniform, and for that purpose, the concept of DP matching is adopted. Which one is aligned with which one is arbitrary, but here, the signature data 1 is aligned. Then, it can be seen that t ₃ and t ₄ corresponding to r ₃ are missing in the signature data 2. Similarly, t ₉ , t ₁₀ , and t ₁₁ corresponding to r ₉ are missing in the signature data 2. On the contrary, r ₆ and r ₇ corresponding to t ₈ are missing in the signature data 1. Therefore, in order to align the length of the signature data 2 with the signature data 1, it is necessary to delete the missing portion on the signature data 2 side and make up for the missing portion of the signature data 1. In this example, both x _rm and x _rn can be, for example, the average value of x _r5 and x _r8 , or x _rm = x _r5 and x _rn = x _r8 . It is also possible to approximate x _rm and x _rn using more advanced calculation methods such as polynomial fitting. In any case, based on a simple rule, the approximation will be made with numerical values that are generally considered appropriate.

In this way, the arithmetic mean can be obtained between each point of the signature data 1 and the signature data 2 whose correspondence relationship has been clarified by using the concept of DP matching, and this can be calculated among, for example, three signature data. By executing this, representative signature data can be obtained. This will be called "representative signature data". As described above, although the xy coordinate system is used in this example, the same can be done in the polar coordinate system, and the same is true even when the signature data includes the pen pressure value. In addition, even if the concept of DP matching is not used, the time length of each signature is arithmetically compressed or decompressed and aligned to complement the points between the points on the same signature, and the average. It is possible to obtain a signature, but if you do so, there is a strong risk that slow and slow information such as the rhythm of the signature will not be captured correctly. It is important to use the concept of DP matching in obtaining the representative signature data in order to capture such slow and fast information as accurately as possible.

101 Signature input 102 Data feature extraction 103 Template 104 Pattern collation 105 Comparison with threshold value 106 Result output 301 Signature position x Coordinate minimum value (x _min )
302 Signature position x maximum coordinate (x _max )
303 Signature position Minimum value of y coordinate (y _min )
304 Signature position y coordinate maximum (y _max )
401 Position of the center of gravity of the entire standardized signature information 601 Signature of the person m ₀ 602 Signature for template creation m ₁ 603 Experimental signature m ₂ 604 Calculation of the center of gravity of the entire signature 605 Calculation of feature amount 606 DP matching 607 3 template 608 (m _1-3 ) signatures 609 DP matching 610 Threshold determination 611 Calculation of the center of gravity of the entire signature 612 Calculation of features 613 DP matching 614 Comparison of magnitude relationship between D _min and c · Th 615 types 1 Error output

Claims (6)

It is a pen input personal authentication method that authenticates the identity of the person based on the handwriting information output from the input device including the trajectory information drawn on the plane by the pen tip.
A step of taking in the coordinates and pressure information of the pen tip as the handwriting information in chronological order,
The step of finding the position of the center of gravity of the entire handwriting from the captured coordinates,
A step of moving the obtained center of gravity to a position away from the plane of the handwriting by a certain distance and using it as the center of gravity for calculation.
Steps to find the vector from the position of the center of gravity for calculation to each point of the handwriting coordinates,
Steps to obtain the length component and angle component of the vector,
Dynamic using an evaluation function that is the product of the difference between the length component, angle component, and pen pressure information in the person's template prepared in advance, and the corresponding length component, angle component, and pen pressure information of the signature to be collated. The step of measuring the DP distance between the signature to be collated and the template of the person prepared in advance by the programming method,
A pen-input personal authentication method including a step of identifying whether or not the signature is the person's signature by comparing the DP distance measured by the evaluation function with a threshold value prepared in advance. It is a pen input personal authentication method that authenticates the identity of the person based on the handwriting information output from the input device including the trajectory information drawn on the plane by the pen tip.
A step of capturing the coordinates and pressure information of the pen tip as the handwriting information,
The step of finding the position of the center of gravity of the entire handwriting from the captured coordinates,
The step of finding the vector from the position of the center of gravity to each point of the coordinates of the handwriting,
Steps to find the length component and angle component from the vector,
The step of finding the distance by the DP method between two of three or more signatures of the person, and
In order to determine the signature to be registered as a template of the person, select three of the DP distances between those signatures, obtain the area of the triangle formed by the three DP distances, and the area is constant. If it is within the range, the template decision step to be adopted for the template and
An evaluation function that is the product of the respective differences between the length component, the angle component, and the pen pressure information of the signature to be collated with the length component, the angle component, and the pen pressure information in the person's template determined in the template determination step. The step of measuring the DP distance between the signature to be collated and the template of the person prepared in advance by the dynamic programming method using
A pen-input personal authentication method including a step of identifying whether or not the signature is the person's signature by comparing the DP distance measured by the evaluation function with a predetermined threshold value. It is a pen input personal authentication method that authenticates the identity of the person based on the handwriting information output from the input device including the trajectory information drawn on the plane by the pen tip.
A step of capturing the coordinates and pressure information of the pen tip as the handwriting information,
The step of finding the position of the center of gravity of the entire handwriting from the captured coordinates,
The step of finding the vector from the position of the center of gravity to each point of the coordinates of the handwriting,
Steps to find the length component and angle component from the vector,
The step of finding the distance by the DP method between two of three or more signatures of the person, and
In order to determine the signature to be registered as a template of the person, select three of the DP distances between those signatures, obtain the area of the triangle formed by the three DP distances, and the area is constant. If it is within the range, the template decision step to be adopted for the template and
It is the product of the difference between the length component, the angle component, and the pen pressure information in the person's template determined in the template determination step, and the corresponding length component, the angle component, and the pen pressure information of the signature to be collated. A step to measure the DP distance between the signature to be collated and the template of the person prepared in advance by the dynamic programming method using the evaluation function, and
A pen-input personal authentication method including a step of identifying whether or not the signature is the person's signature by comparing the DP distance measured by the evaluation function with a predetermined threshold value. It is a pen input personal authentication method that authenticates the identity of the person based on the handwriting information output from the input device including the trajectory information drawn on the plane by the pen tip.
A step of capturing the coordinates and pressure information of the pen tip as the handwriting information,
The step of finding the position of the center of gravity of the entire handwriting from the captured coordinates,
The step of finding the vector from the position of the center of gravity to each point of the coordinates of the handwriting,
Steps to find the length component and angle component from the vector,
An intermediate signature trajectory between two signatures by thinning out or adding coordinate points based on certain rules during analysis by the DP method between two of each of the three or more signatures. Intermediate trajectory determination step to find
A template determination step that seeks a single representative signature template by repeating the intermediate trajectory determination step,
It is the product of the difference between the length component, the angle component, and the pen pressure information in the person's template determined in the template determination step, and the corresponding length component, the angle component, and the pen pressure information of the signature to be collated. A step to measure the DP distance between the signature to be collated and the template of the person prepared in advance by the dynamic programming method using the evaluation function, and
A pen-input personal authentication method including a step of identifying whether or not the signature is the person's signature by comparing the DP distance measured by the evaluation function with a predetermined threshold value. In the pen input personal authentication method according to any one of claims 2 to 4, the determined center of gravity is moved to a position away from the plane of the handwriting by a certain distance as the position of the center of gravity. A pen-input personal authentication method characterized by using the position of the center of gravity for calculation. A computer-readable storage medium comprising a program for carrying out any one of claims 1-5.