CN102592142A - Computer-system-based handwritten signature stability evaluation method - Google Patents

Abstract

A method for evaluating the stability of handwritten signatures based on a computer system. In the present invention, the signature to be tested and a sample set of signatures from the same writer are obtained by collecting or extracting from an existing signature database by a handwriting input device. The signature to be tested needs to be stabilized. The analyzed handwritten signature, the signature sample set is the handwritten signature set for stability calculation with the signature to be tested. The signature similarity evaluation between each signature sample in the signature sample set and the signature to be tested is completed by a signature segment matching calculation and local matching segment difference calculation, and then according to the signature similarity evaluation, the signature sample set is similar to the signature to be tested The number of signature samples, and compare it with the preset threshold to judge whether the signature to be tested is stable, and finally complete the stability evaluation of the signature to be tested. This method is both meticulous and comprehensive in assessing the stability of handwritten signatures, and effectively solves the problem of signature detail information extraction that is ignored by existing handwritten signature stability assessment methods.

Description

A computer-based method for evaluating the stability of handwritten signatures

technical field

The invention relates to the technical field of online handwritten signature authentication, in particular to a method for evaluating the stability of handwritten signatures based on a computer system.

Background technique

As one of the important contents of biometric identification technology, handwritten signature authentication utilizes personalized writing features to realize identity identification. According to different data acquisition methods, signature authentication can be divided into offline signature authentication and online signature authentication. Compared with offline signature verification using static images, online signature verification uses handwriting pads and other devices to collect handwriting information generated during the writing process, which can obtain richer information and has higher accuracy. Therefore, it has broad application prospects.

The precondition for using biometrics as the basis for identity authentication is that it has certain stability. However, affected by the level of writing, the stability of signatures written by people is not the same. For example, writers who often carry out writing activities have higher writing skills, fluent writing, and higher stability of signatures; For early children, or for some elderly people suffering from diseases, the stability of their signatures is poor. For signatures with poor stability, the difference between signatures is large, which will cause two problems in the signature-based authentication system: one is that while tolerating greater intra-class differences, more forged signatures will inevitably be accepted ; The other is that while rejecting more forged signatures, more authentic signatures will inevitably be rejected.

In order to solve this problem, Joan et al. pointed out in the literature "On-line signature verification system with failure toenrol management" (Pattern Recognition, 2009) that in practical applications, the system needs to automatically discover and handle the failure of user registration, that is, the user submits The stability of the biometric feature does not meet the requirements for identity authentication. Enrollment failures are not uncommon in the field of biometric authentication, for example, about 4% of people cannot successfully enroll in a fingerprint authentication system because they cannot provide high-quality personal fingerprints. Since the signature itself has a large intra-class variation, the proportion of users who fail to register in the signature verification field may be larger.

For the handling of registration failures, Joan et al. proposed the idea of "smart registration": perform stability checks during the registration phase and automatically reject low-quality signature samples, while requiring users to submit new registration samples. In this way, although it will cause inconvenience to some users, the authentication performance of the system can be guaranteed to a certain extent.

The quality of signature samples is determined by their stability. Joan et al. evaluated the stability of registration signatures through pairwise matching results between samples. For each pair of compared signatures, seven types of time series signals are collected, and the parameter features are extracted by discrete cosine transform. After feature selection, a secondary discriminant classifier is used for matching. Where three registration samples are required, if one sample does not match the other two, the system will reject the sample submission and require the user to submit a new sample.

After analyzing the above prior art, the inventor found that:

The stability assessment method adopted by Joan et al. is based on the global parameter feature between samples. The global parameter feature is a statistical feature obtained from the entire signature signal as a feature extraction unit, which cannot effectively reflect the detailed information of the signature. An effective stability evaluation should at least meet two points: ① Meticulousness, that is, the signature features should give an accurate description of the subtle differences in the signature, and local features should be used; ② Comprehensiveness, that is, the signature features should give the stability of the signature To develop a comprehensive description, both static and dynamic characteristics of signatures need to be considered.

Invention content:

The technical problem to be solved by the present invention is to propose a method for assessing the stability of handwritten signatures in view of the meticulousness and comprehensiveness of the stability analysis, especially by using a signature segment matching calculation and a local matching segment difference value calculation Handwritten signature stability assessment method to solve the problem of quantitative assessment of handwriting stability

Technical scheme of the present invention is:

A method for evaluating the stability of a handwritten signature based on a computer system, in particular: the signature to be tested and a sample set of signatures from the same writer are obtained through collection of a handwriting input device, the signature to be tested is a handwritten signature that needs to be analyzed for stability, The signature sample set is the set of handwritten signatures with which the stability calculation of the signature to be tested is performed. Through the signature similarity evaluation, in particular, a signature segment matching calculation and a partial matching segment difference value calculation are used to complete each signature in the signature sample set. The signature similarity evaluation between the sample and the signature to be tested, and then according to the signature similarity evaluation, the number of signature samples similar to the signature to be tested in the signature sample set is obtained, and compared with the preset threshold to determine whether the signature to be tested is stable, Finally, the stability evaluation of the signature to be tested is completed. Methods include:

get signature;

signature similarity evaluation;

Stability assessment;

Obtain signatures, including:

The signature to be tested and the signature sample set from the same writer are collected through the handwriting input device. The signature to be tested is the handwritten signature that needs to be analyzed for stability, and the signature sample set is the handwritten signature for stability calculation with the signature to be tested. Set, let S represent a signature to be tested by user U, set B={S ₁ , S ₂ , S ₃ ,..., S _N } represents the signature sample set of user U, where N≥1, that is:

Obtain the signature, get a signature S of user U and a signature sample set B of user U;

Signature similarity assessment, specifically including:

When the user's signature to be tested is obtained, it is necessary to combine the signature to be tested with each signature sample in the signature sample set to form a corresponding signature pair, and then perform signature similarity evaluation on each set of signature pairs, especially using A signature segment matching calculation and local matching segment difference value calculation, which realizes the similar calculation of the local matching segments of each group of signature pairs, and realizes the similar calculation of the signature pairs of each group of signature pairs according to the obtained similar calculation results of the partial matching segments, Finally, according to the signature pair similarity calculation results of each group of signature pairs, the similarity evaluation between the signature to be tested and each signature sample is completed;

One of signature segment matching calculation and partial matching segment difference value calculation specifically refers to: firstly, each group of signature pairs is divided into signature partial segments, so as to obtain the partial segment sequence and signature of the signature to be tested in each group of signature pairs The local segment sequence of the sample; and then use the local segment matching algorithm for the partial segment sequence of the signature to be tested and the partial segment sequence of the signature sample in each group of signature pairs to obtain the partial segment sequence of the signature to be tested and the partial segment sequence of the signature sample The corresponding relationship sequence with "one-to-one" characteristics, that is, the sequence of partial matching segments; finally, calculate the difference value of local matching segments for each pair of partial matching segments in the sequence of partial matching segments, especially for each pair of partial matching segments The local segment static features of the signature to be tested, the local segment dynamic features of the test signature, the local segment static features of the signature sample, and the local segment dynamic features of the signature sample are extracted, and the signature to be tested in each pair of extracted partial matching segments Local segment static features, local segment dynamic features of the signature to be tested, local segment static features of the signature sample, and local segment dynamic features of the signature sample are used to calculate the difference value of the local matching segment;

Among them, the local segment refers to the local signature curve formed between two adjacent key points in a signature; the key point refers to the signature local extreme points such as signature speed extreme points, signature pressure extreme points, etc.;

The partial matching segment is composed of two corresponding partial segments belonging to the partial segment sequence of the signature to be tested and the partial segment sequence of the signature sample;

Among them, the local segment static features refer to local segment static function features including signature two-dimensional font information and local segment static parameter features such as local segment curvature and local segment arc length;

Among them, the dynamic characteristics of the local segment refer to the dynamic function characteristics of the local segment including the local segment velocity and acceleration, and the local segment dynamic parameter characteristics such as the average writing speed of the local segment, the average writing acceleration of the local segment, and the average writing pressure of the local segment;

Stability assessment, specifically including:

After the signature similarity evaluation between the signature to be tested and each signature sample in the signature sample set is completed, the number of signature samples in the signature sample set that are similar to the signature to be tested is obtained, and the number of signature samples in the signature sample set that is similar to the signature to be tested is determined The relationship between the set threshold and the stability evaluation of the signature to be tested is completed. Let Z _en represent the number of signature samples in the signature sample set that are similar to the signature to be tested, and T ₃ is a preset value based on the stability requirements. Threshold, that is:

The stability evaluation of the signature to be tested is obtained by judging the relationship between the number Z _en of signature samples similar to the signature to be tested in the signature sample set and the preset threshold T ₃ according to the stability requirements.

As a further improvement to the prior art, the handwriting input device provided by the method of the present invention includes: a handwriting input module, a signal acquisition module, a signal processing module, and a display module, wherein:

Handwriting input module: it provides the user with a handwriting area, the user can write freely in the handwriting area, or write according to the shape given by the computer system method; at the same time, the computer system method receives the user's handwriting input information through the handwriting input module ;

Signal acquisition module: When the user writes in the handwriting input module, handwriting input information will be generated. The signal acquisition module connected to the handwriting input module can perceive and collect the handwriting information generated during the handwriting process. For example, it can collect handwriting information. Position information, force information, inclination information, etc.;

Signal processing module: After collecting the handwritten information generated during the handwriting process, the handwriting processing module needs to analyze and process the collected handwritten information to realize corresponding functions according to the requirements of the system, and output the processed results to the display module;

Display module: When the signal processing module finishes processing the collected handwritten letters, it will output the processing result to the display module, and the display module will simultaneously drive the corresponding display components to complete the display function of the processing result according to the processing result.

The handwriting input device is either a computing device or an instrument externally connected with a handwriting input device, or a PDA, a handwriting input mobile phone, a Tablet PC, or a server.

The handwriting input device captures handwritten signature data, which also includes extracting data from existing signature databases.

The beneficial effects of the present invention are:

Compared with the prior art, the present invention collects the signature to be tested and the signature sample set through the handwriting input device, and performs signature similarity evaluation on the signature to be tested and each signature sample in the signature sample set. The provided handwritten signature stability evaluation method effectively solves the problem of extracting signature details neglected by the existing handwritten signature stability evaluation methods, and the method steps of the present invention are signature acquisition, signature similarity evaluation, and stability evaluation.

In the signature similarity evaluation, the local matching segment similarity calculation and the signature pair similarity calculation are performed after the signature pair is obtained, wherein the partial matching segment similarity calculation includes the signature segmentation matching calculation and the partial matching segment difference value calculation to realize the signature similarity evaluation. The signature segmentation matching calculation realizes the segmentation matching processing of the signature to be tested and each signature sample in the signature sample set to obtain a sequence of partial matching segments; the calculation of the difference value of the partial matching segments realizes the feature extraction of the partial matching segment, and according to the extracted partial matching segment The feature completes the calculation of the difference value of the local matching segment;

That is, the signature segment matching calculation is performed on the signature to be tested and each signature sample in the signature sample set, and the partial matching segment sequence corresponding to each signature sample in the test signature and signature sample set is obtained, which is a manifestation of meticulousness; the test signature and The local matching segment sequence corresponding to each signature sample in the signature sample set is extracted from the static and dynamic features of the local matching segment, and the difference value calculation of the partial matching segment is completed according to the extracted static and dynamic features of the partial matching segment, so as to realize the The stability analysis of the test signature is a comprehensive embodiment.

In the stability assessment, is it judged whether the number of similar signatures is less than the threshold? If the number of similar signatures is less than the threshold, the signature to be tested is unstable; if the number of similar signatures is greater than the threshold, the signature to be tested is stable, and the stability evaluation of the signature to be tested is realized.

That is, according to the signature similarity evaluation results, the number of signature samples similar to the signature to be tested in the signature sample set is obtained, and compared with the preset threshold to determine whether the signature to be tested is stable, and finally complete the stability evaluation of the signature to be tested.

The method of the present invention also provides a handwriting input device, including a handwriting input module, a signal acquisition module, a signal processing module, and a display module, for realizing the acquisition and calculation of a handwritten signature, but the method of the present invention is not limited to the handwriting input device , any computing device or instrument with handwriting input function or externally connected with handwriting input equipment, such as PDA, handwriting input mobile phone, Tablet PC, server, etc., can use the method of the present invention to evaluate the stability of handwritten signatures.

Description of drawings:

Fig. 1 is a block diagram of handwritten signature stability processing of the present invention;

Fig. 2 is the flowchart of signature similarity evaluation of the present invention;

Fig. 3 is the flow chart of similar calculation of local matching segment of the present invention;

Fig. 4 is the flow chart of similarity calculation of signature pair of the present invention;

Fig. 5 is a flowchart of the stability evaluation of the present invention.

Detailed ways:

Embodiments of the present invention will be further explained below in conjunction with the accompanying drawings:

FIG. 1 is a block diagram of handwritten signature stability processing in the present invention; specifically, it includes a handwritten input device 1 and a handwritten signature stability evaluation method flow.

Handwriting input device 1 specifically comprises: handwriting input module 11, signal acquisition module 12, signal processing module 13, display module 14, wherein:

The handwriting input module 11 is that the handwriting input module provides a handwriting area to the user, and the user can write freely in the handwriting area, or write according to the shape given by the computer system; at the same time, the computer system receives the user's handwriting through the handwriting input module 11. Enter information.

The signal acquisition module 12 is that when the user writes in the handwriting input module 11, handwriting input information can be generated, and the signal acquisition module 12 connected with the handwriting input module 11 can sense and collect the handwriting information produced in the handwriting process, for example, can collect Position information, force information, inclination information, etc. in handwritten information.

Signal processing module 13 is after finishing collecting the handwriting information that produces in the handwriting process, handwriting processing module 13 needs to analyze and process the handwriting information of collection, to realize corresponding function according to the demand of computer system, and the result of processing is exported to Display module 14.

The display module 14 is that after the signal processing module 13 finishes processing the handwritten letter collected, the result of the processing will be output to the display module 14, and the display module 14 will simultaneously drive the corresponding display elements to complete the processing result according to the processing result. Display function.

The handwritten signature stability assessment method flow specifically includes:

Step 100: Start; specifically, it means starting to execute the handwriting stability evaluation operation flow.

Step 200: Obtain the signature; specifically, the signature to be tested and the signature sample set from the same writer are obtained through collection by the handwriting input device 1, the signature to be tested is the handwritten signature that needs to be analyzed for stability, and the signature sample set is the The set of handwritten signatures for stability analysis of the signature to be tested, let S represent a signature to be tested by user U, and the set B={S ₁ , S ₂ , S ₃ ,...,S _N } represents the signature sample set of user U , where N≥1, namely:

Obtain the signature, and obtain a signature S of user U and a signature sample set B of user U.

Step 300: Signature similarity evaluation; specifically, when the user's signature to be tested is obtained through the handwriting input device 1, it is necessary to combine the signature to be tested with each signature sample in the signature sample set to form a corresponding signature pair, Then carry out the signature similarity evaluation for each group of signature pairs, especially adopt a kind of signature segmentation matching calculation and local matching segment difference value calculation to realize the partial matching segment similarity calculation of each group of signature pairs, and according to the obtained partial matching segment The similarity calculation results realize the signature pair similarity calculation of each group of signature pairs, and finally, complete the signature similarity evaluation between the signature to be tested and each signature sample according to the signature pair similarity calculation results of each group of signature pairs.

Step 400: Stability assessment; specifically, after the signature similarity assessment between the signature to be tested and each signature sample in the signature sample set is completed, the number of signature samples in the signature sample set that are similar to the signature to be tested is obtained, and by judging the number of signature samples in the signature sample set The relationship between the number of signature samples similar to the signature to be tested and the set threshold, so as to complete the stability assessment of the signature to be tested, let _Zen represent the number of signature samples similar to the signature to be tested in the signature sample set, T ₃ is the preset threshold according to the requirements for stability, namely:

The stability evaluation of the signature to be tested is obtained by judging the number Z _en of signature samples in the signature sample set that are similar to the signature to be tested, and according to the relationship between the preset threshold T ₃ required for stability.

Step 500: end; specifically, it means that the handwriting stability assessment operation flow is completed.

Fig. 2 is the flowchart of signature similarity evaluation of the present invention; Its steps specifically include:

Step 310: Start; specifically, it means starting to execute the signature similarity evaluation operation flow.

Step 320: Obtain a signature pair; specifically, when the user's signature to be tested is obtained through the handwriting input device 1, it is necessary to combine the signature to be tested with each signature sample in the signature sample set to form a signature pair. After a signature sample is extracted from the signature sample set to form a signature pair with the signature to be tested, the number corresponding to the signature sample is deleted from the signature sample set number.

In order to better understand the above steps and explain how to obtain the signature pair process, let S represent a signature to be tested by user U, and set B={S ₁ , S ₂ , S ₃ ,...,S _N } represents user U signature sample set, where 1, 2, 3, ..., N represents the number corresponding to each signature sample in the signature sample set of user U, where N≥1, then:

Take any signature sample S _i from the signature sample set B of user U and a signature S to be tested by user U to form a signature pair, where the value range of i is 1, 2, 3, ..., N, and the The number i corresponding to the signature sample S _i is deleted from the number of the signature sample set B of the user U.

Step 330: Similar calculation of partial matching segments; specifically, after obtaining signature pairs, a set of signature pairs is obtained, and then signature segment matching calculation is performed on the signatures to be tested and the signature samples in the set of signature pairs to obtain the set of signatures Align the partial matching segment sequence corresponding to the signature to be tested and the signature sample, and calculate the local matching segment difference value for the partial matching segment sequence, and finally compare the obtained partial matching segment difference value calculation result with the preset threshold to complete The similarity calculation of the local matching segment in the group signature pair.

Wherein, the similarity calculation of the local matching segment also includes the calculation of the difference value of the local matching segment, which realizes the calculation of the difference value of the local matching segment corresponding to the partial matching segment of the group of signatures, especially for the signature to be tested in each pair of partial matching segments. The local segment static features, the local segment dynamic features of the signature to be tested, the local segment static features of the signature sample, and the local segment dynamic features of the signature sample are extracted, and the local segment static features of the signature to be tested in each pair of extracted partial matching segments are extracted. feature, the local segment dynamic feature of the signature to be tested, the local segment static feature of the signature sample, and the local segment dynamic feature of the signature sample to calculate the difference value of the local matching segment.

Step 340: Signature pair similarity calculation; specifically, after completing the partial matching section similarity calculation in a group of signature pairs, the total length of the partial matching section similarity between the signature to be tested and the signature sample in the group signature pair is obtained, and then The obtained partial matching segment similarity total length is calculated with the partial segment sequence length of the signature to be tested in the group signature pair, and finally the signature pair similarity calculation of the group signature pair is completed according to the calculation result.

Among them, the similar total length of the local matching segments in the group signature pair refers to: the partial segment sequence and signature sample of the test signature in the group signature pair are obtained after signature segment matching and calculation The partial sequence of partial segments, and the partial matching segment sequence corresponding to this group of signature pairs, the similarity between the partial matching segment sequences is judged by the similarity calculation of the partial matching segment, and the group signature corresponding to the similarity of the partial matching segment is judged Calculate the segment arc length of the local segment of the signature to be tested, and add up the partial segment arc length of the group of signatures corresponding to the similarity of the local matching segment for each judgment result, and obtain the partial match of the group of signatures segments of similar total length.

Wherein, the length of the partial segment sequence of the signature to be tested in the group signature pair refers to: the partial segment sequence of the signature to be tested in the group signature pair is obtained after signature segmentation matching calculation of the signature to be tested in the group signature pair. The local segment arc length corresponding to each partial segment sequence of the signature to be tested in the signature pair, and the local segment arc lengths corresponding to all the partial segment sequences of the signature to be tested in the signature pair are accumulated to obtain the signature pair to be tested Measure the length of the partial segment sequence of the signature.

Step 350: Determine whether the signature sample set is empty? If yes, go to step 360 to end the signature similarity evaluation, if not, go to step 320 and continue to obtain the signature pair; specifically, every time a signature sample is extracted from the signature sample set to form a signature pair with the signature to be tested, and a local After the matching segment similarity calculation and signature pair similarity calculation, the number corresponding to the signature sample is deleted from the signature sample set number, and when the signature sample set number is empty, it means that each signature sample in the signature sample set has been completed and pending Signature similarity evaluation of the test signature, the signature similarity evaluation process ends at this time, otherwise, continue to obtain the signature pair and perform the signature similarity evaluation process.

Step 360: end; specifically, it means that the signature similarity evaluation operation flow is completed.

Fig. 3 is the flow chart of similar calculation of local matching segment of the present invention; Its steps specifically include:

Step 331: Start; specifically, acquire a group of signature pairs, and start to execute the operation process of partial matching segment similarity calculation.

Step 332: Preprocessing; specifically, during the signature collection process, due to the interference of noise and the shaking of the hand during signature, the signature data will be affected, so the signature to be tested and the signature in the group of signature pairs need to be Sample raw signature data for preprocessing.

Step 333: Signature segment matching calculation; specifically, after completing the preprocessing of the original signature data of the signature to be tested and the signature sample in the group of signature pairs, use the signature segment matching calculation to perform segment matching processing on the group of signature pairs , to obtain the local segment sequence of the signature to be tested and the partial segment sequence of the signature sample in the group signature pair, and then apply the local segment matching algorithm to the partial segment sequence of the signature to be tested in the group signature pair and the partial segment sequence of the signature sample, Obtain the corresponding relationship sequence with "one-to-one" characteristic between the partial segment sequence of the signature to be tested and the partial segment sequence of the signature sample, that is, the partial matching segment sequence; and the adopted signature segmentation matching calculation has the following two characteristics:

a. The signature segmentation matching calculation adopted can realize the consistent key point extraction of the signature to be tested and the signature sample in the set of signature pairs;

b. The adopted signature segment matching calculation can realize that the length of the partial segment sequence of the signature to be tested in the set of signature pairs is the same as the length of the partial segment sequence of the signature sample.

Among them, the local segment refers to the local signature curve formed between two adjacent key points in a signature; the key point refers to the signature local extreme points such as signature speed extreme points, signature pressure extreme points, etc.;

The partial matching segment is composed of two corresponding partial segments belonging to the partial segment sequence of the signature to be tested and the partial segment sequence of the signature sample;

In order to better understand the above steps, explain how to obtain a partial matching segment sequence through signature segment matching calculation, let S represent a signature to be tested by user U, set B={S ₁ , S ₂ , S ₃ ,... , S _N } represents the signature sample set of user U, where 1, 2, 3, ..., N represents the number corresponding to each signature sample in the signature sample set, where N≥1, P _i ={S, S _i } Indicates a set of signature pairs obtained, where the value of i is between 1, 2, 3, ..., N;

Let A _C be the partial segment sequence corresponding to the signature S to be tested in the signature pair P _i , Ac={ac(1), ac(2),..., ac(K ₁ )}, B _C is the signature pair P _i The partial segment sequence corresponding to the signature sample S _i in the middle, Bc={bc(1), bc(2),..., bc(K ₂ )}, C is the partial segment corresponding to the signature S to be tested in the signature pair P _i The partial matching segment sequence composed of the sequence A _C and the partial segment sequence B _C corresponding to the signature sample S _i in the signature pair P _i , C={c(1), c(2),...,c(K)} ;

Where K ₁ is the length of the partial segment sequence A _C corresponding to the signature S to be tested in the signature pair P _i , 1, 2, ..., K ₁ is the partial segment sequence A _C corresponding to the signature S to be tested in the signature pair P _i The segment number, ac(m) is expressed as the mth partial segment of the partial segment sequence A _C corresponding to the signature S to be tested in the signature pair P _i , and the value range of m is 1, 2, ..., K ₁ ;

Where K ₂ is the length of the local segment sequence B _C corresponding to the signature sample S _i in the signature pair P _i , 1, 2, ..., K ₂ is the local segment sequence B _C corresponding to the signature sample _{S i in} the signature pair P i The segment number of , bc(n) represents the nth partial segment of the partial segment sequence B _C corresponding to the signature sample S _i in the signature pair P _i , and the value range of n is 1, 2, ..., K ₂ ;

Where K is the length of the partial matching segment sequence C corresponding to the signature pair _Pi , 1, 2, ..., K is the segment number of the partial matching segment sequence _C corresponding to the signature pair Pi, and c(K) is expressed as The kth pair of partial matching segments of the partial matching segment sequence C corresponding to the signature pair P _i , where the value of k is in the range of 1, 2, ..., K;

K ₁ =K ₂ =K;

That is, according to the local segment sequence A _C corresponding to the signature S to be tested in the signature pair P i and the local segment sequence B _C corresponding to the signature sample _{S i in} the signature pair P _i , through the local segment matching algorithm, the signature pair P _i to be tested is obtained The corresponding sequence _{of partial segment sequence AC corresponding to signature S and partial segment sequence B C corresponding} to signature sample S _i in signature pair P _i with "one-to-one" characteristic is the partial matching segment sequence C.

Step 334: Calculate the difference value of the partial matching segment; specifically, after completing the signature segment matching calculation of the signature to be tested and the signature sample in a group of signature pairs, the partial matching segment sequence corresponding to the group of signature pairs is obtained, and from the Select a pair of partial matching segments from the partial matching segment sequence corresponding to the group signature pair, extract the local matching segment features corresponding to the pair of partial matching segments, specifically including the local segment static features and local segment dynamic features, and according to the extracted local segment Static features and partial segment dynamic features complete the calculation of the difference value of the partial matching segment. In addition, each time a pair of partial matching segments is selected from the partial matching segment sequence, the segment number corresponding to the pair of partial matching segments is changed from the segment number of the partial matching segment sequence Deleted in .

Among them, the local segment static features refer to local segment static function features including signature two-dimensional font information and local segment static parameter features such as local segment curvature and local segment arc length;

Among them, the dynamic characteristics of the local segment refer to the dynamic function characteristics of the local segment including the local segment velocity and acceleration, and the local segment dynamic parameter characteristics such as the average writing speed of the local segment, the average writing acceleration of the local segment, and the average writing pressure of the local segment;

Among them, the calculation of the difference value of the local matching segment can be completed by using the DTW algorithm, the DTW improved algorithm, or any algorithm that can perform time series matching and calculate the difference value for the local segment static function feature and the local segment dynamic function feature. Static parameter features and local segment dynamic parameter features can be completed by using algorithms such as calculating the absolute value of the difference between eigenvalues.

Step 335: Determine whether the difference value of the local matching segment is smaller than the threshold T ₁ ? If yes, then jump to step 336 to update the similar total length of the partial matching segment, if not, then directly jump to step 337 to judge whether the partial matching segment sequence is empty; specifically, when completing the partial matching segment of a pair of partial matching segments After the difference value is calculated, the difference value of the local matching segment corresponding to the pair of local matching segments is obtained, and the relationship between the difference value of the local matching segment corresponding to the pair of local matching segments and the set threshold T ₁ is judged, and the pair of local matching segments is obtained. Matches the similarity of the segment.

Step 336: Update the total similarity length of partial matching segments; specifically, according to the relationship between judging the difference value of partial matching segments and the set threshold T ₁ , obtain the similarity status of a pair of partial matching segments in a group of signature pairs, and update The partial matching segments in the set of signature pairs have similar total lengths.

Among them, the similar total length of the local matching segments in the group signature pair refers to: the partial segment sequence and signature sample of the test signature in the group signature pair are obtained after signature segment matching and calculation The partial sequence of partial segments, and the partial matching segment sequence corresponding to this group of signature pairs, the similarity between the partial matching segment sequences is judged by the similarity calculation of the partial matching segment, and the group signature corresponding to the similarity of the partial matching segment is judged Calculate the segment arc length of the local segment of the signature to be tested, and add up the partial segment arc length of the group of signatures corresponding to the similarity of the local matching segment for each judgment result, and obtain the partial match of the group of signatures segments of similar total length.

In order to better understand the above steps, it is explained that according to the relationship between judging the difference value of the local matching segment and the set threshold T ₁ , the total similarity length of the partial matching segment in a group of signature pairs is updated. Let S represent a user U Signatures to be tested, set B={S ₁ , S ₂ , S ₃ ,...,S _N } represents the signature sample set of user U, where 1, 2, 3,..., N represent each signature sample set The number corresponding to the signature sample, where N≥1, P _i ={S, S _i } indicates a set of signature pairs obtained, where the value of i is between 1, 2, 3, .., N;

T₁为根据对稳定性的要求预先设定的局部段相似阈值，其中x为局部匹配段差异值的维数值，

为第i维局部匹配段差异值，i取值在1，2，...，x之中，则有：Let Den represent the similar total length of the local matching segment between the signature S to be tested in the signature pair P _i and the signature sample S _i . In the initial case, Den=0, and Dif(k) is represented as the signature S to be tested in the signature pair _Pi . The difference value of the local matching segment obtained by calculating the difference value of the local matching segment for the kth pair of partial matching segment of the partial matching segment sequence corresponding to the signature sample S _i ,

T ₁ is the local segment similarity threshold preset according to the requirement for stability, Where x is the dimension value of the difference value of the local matching segment,

is the difference value of the ith-dimension local matching segment, and the value of i is among 1, 2, ..., x, then:

成立时，判定签名对P_i中待测签名S与签名样本S_i所对应的局部匹配段序列的第k对局部匹配段不相似，签名对P_i中待测签名S与签名样本S_i之间的局部匹配段相似总长度Den维持上次数值；对任意取值在1，2，...，x之中的i，均有

成立时，判定签名对P_i中待测签名S与签名样本S_i所对应的局部匹配段序列的第k对局部匹配段相似，签名对P_i中待测签名S与签名样本S_i之间的局部匹配段相似总长度Den进行更新操作：Den＝Den+Count；For any value i among 1, 2, ..., x, there exists

When it is established, it is determined that the k-th pair of partial matching segments of the partial matching segment sequence corresponding to the signature S in the signature pair P _i and the signature sample S _i are not similar, and the difference between the signature S in the signature pair P _i and the signature sample S _i The similar total length Den of the local matching segment between maintains the previous value; for any i that takes on the value of 1, 2, ..., x, there are

When it is established, it is determined that the k-th pair of partial matching segments of the partial matching segment sequence corresponding to the signature S to be tested in the signature pair P _i and the signature sample S _i are similar, and the relationship between the signature S to be tested in the signature pair P _i and the signature sample S _i The partial matching segment similar to the total length Den is updated: Den=Den+Count;

Among them, Count is the arc length of the partial segment of the signature S to be tested corresponding to the k-th pair of partial matching segments being judged to be similar.

Step 337: judge whether the sequence of partial matching segments is empty, if yes, jump to step 338, end the similarity calculation of the partial matching segments, if not, then directly jump to step 334, and continue the local matching segment difference value of the next pair of partial matching segments Calculation; Specifically, after extracting a pair of partial matching segments from the partial matching segment sequence corresponding to a group of signature pairs, the segment number corresponding to the pair of partial matching segments is deleted from the segment number of the partial matching segment sequence, then When the segment number of the partial matching segment sequence is empty, it means that each pair of partial matching segments in the partial matching segment sequence has completed the calculation of the difference value of the partial matching segment and obtained the corresponding similar total length of the partial matching segment. The partial matching segment similarity calculation process of the group signature pair is over, otherwise, it means that there are still unprocessed partial matching segments in the partial matching segment sequence, continue to obtain the next group of partial matching segments in the partial matching segment sequence, and perform partial matching segment difference value calculation.

Step 338: end; specifically, it means that the operation flow of similar calculation of partial matching segments is completed.

Fig. 4 is a schematic flow chart of the signature pair similarity calculation of the present invention; its steps specifically include:

Step 341: start; specifically, it means start to execute the signature pair similarity calculation operation flow.

Step 342: Calculate the similarity ratio R _O of the partial matching segment; specifically, after completing the similarity calculation of the partial matching segments of a group of signature pairs, obtain the total similarity length of the partial matching segments in the group of signature pairs, and calculate the partial matching segment length of the group of signature pairs The ratio of the segment similarity total length to the partial segment sequence length of the signature to be tested in the group of signature pairs is obtained to obtain the partial matching segment similarity ratio R _O in the group of signature pairs.

Wherein, the length of the partial segment sequence of the signature to be tested in the group signature pair refers to: the partial segment sequence of the signature to be tested in the group signature pair is obtained after signature segmentation matching calculation of the signature to be tested in the group signature pair. The local segment arc length corresponding to each partial segment sequence of the signature to be tested in the signature pair, and the local segment arc lengths corresponding to all the partial segment sequences of the signature to be tested in the signature pair are accumulated to obtain the signature pair to be tested Measure the partial segment length of the signature.

In order to better understand the above steps, let S represent a signature to be tested by user U, and the set B={S ₁ , S ₂ , S ₃ ,...,S _N } represents the signature sample set of user U, where 1 , 2, 3, ..., N represents the number corresponding to each signature sample in the signature sample set, where N≥1, P _i = {S, S _i } represents a set of signature pairs obtained, where the value of i is in Between 1, 2, 3, ..., N, Tl _S represents the partial segment sequence length of the signature S to be tested in the signature pair P _i , and _Li represents the signature S to be tested and the sample signature S in the signature pair P _i The total length of the partial matching segment between _i is similar, and R _O is the similarity ratio of the partial matching segment in the group signature pair P _i , which is:

${\mathrm{<span\; class="notranslate">\; R</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; L</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}{{\mathrm{<span\; class="notranslate">\; Tl</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}}$

That is, according to the similar total length L _i of the local matching segment between the signature _S to be tested and the sample signature S _i in the signature pair P i and the partial segment sequence length Tl _S of the signature S to be tested in the signature pair P _i , the group of signature pairs is obtained The local matching segment similarity ratio R _O in P _i .

Step 343: Determine whether the ratio R _O is smaller than the threshold T ₂ ? If yes, then jump to step 345, determine that the signature to be tested in the group signature pair is not similar to the signature sample, end the signature pair similarity calculation process, if not, then jump to step 344, determine that the signature to be tested in the group signature pair is not similar to the signature sample The signature samples are similar, and the statistics of similar signatures are updated; specifically, let S represent a signature to be tested by user U, and set B={S ₁ , S ₂ , S ₃ ,..., S _N } represents user U’s Signature sample set, where 1, 2, 3, ..., N represents the number corresponding to each signature sample in the signature sample set, where N≥1, P _i = {S, S _i } represents a set of signature pairs obtained , where the value of i is between 1, 2, 3, ..., N, R _O is the similarity ratio of the local matching segment in the group signature pair P _i , and T ₂ is the pre-set matching according to the stability requirements Signature similarity calculation threshold, then:

When R _O < T ₂ is established, it is determined that the signature S to be tested in the group signature pair P _i is not similar to the signature sample _S i, and when R _O > T ₂ is established, it is determined that the signature S to be tested in the group signature pair P _i Similar to the signature sample S _i .

Step 344: Update the statistical quantity of similar signatures; specifically, update the statistical quantity of similar signatures according to whether the judgment ratio R _O is less than the threshold T ₂ , let Q _en represent the statistical quantity of signature samples similar to the signature to be tested in the signature sample set, In the initial case, Q _en =0, R _O is the similarity ratio of local matching segments in a group of signature pairs, and T ₂ is the pre-set matching signature similarity calculation threshold according to the stability requirements, then:

When R _O < T ₂ holds true, that is, the signature to be tested in this group of signature pairs is not similar to the signature sample, and the statistical quantity Q _en of similar signatures maintains the previous value. When R _O > T ₂ holds, that is, the group of signature pairs The signature to be tested is similar to the signature sample, and the statistical quantity Q _en of the similar signature is added by 1.

Step 345: end; specifically, it means that the signature pair similarity calculation operation flow is completed.

Fig. 5 is the flowchart of stability assessment of the present invention; Its steps specifically include:

Step 410: start; specifically, it means starting to execute the stability assessment operation process;

Step 420: Determine whether the number of similar signatures is less than the threshold T ₃ ? If yes, skip to step 430 to determine that the signature to be tested is unstable; if not, skip to step 440 to determine that the signature to be tested is stable; After the similarity calculation, the number of signature samples similar to the signature to be tested in the signature sample set is obtained. By judging the relationship between the number of signature samples in the signature sample set similar to the signature to be tested and the set threshold, the test signature is completed. Stability evaluation, let Z _en represent the number of similar signatures obtained after the signature pair similarity calculation between the signature to be tested and the signature sample set, and T ₃ is the preset threshold according to the stability requirements, then:

When Zen≥T ₃ is established, it is determined that the signature to be tested is stable; when Zen<T ₃ is established, it is determined that the signature to be tested is unstable.

Step 430: The signature to be tested is unstable; specifically, when it is judged that the statistical quantity of similar signatures is less than a preset threshold according to the stability requirement, it indicates that the signature to be tested is unstable.

Step 440: The signature to be tested is stable; specifically, when it is judged that the statistical quantity of similar signatures is greater than a preset threshold according to the stability requirement, it means that the signature to be tested is stable.

Step 450: end; specifically, it means that the stability assessment operation flow is completed.

That is, in the embodiment of the present invention, when a user submits a signature to be tested through a handwriting input device, the stability assessment result of the user's signature to be tested is obtained through a handwritten signature stability assessment method provided by the present invention.

Claims (4)

1. a handwritten signature stability assessment method based on computer system, collects handwritten signature data by handwriting input device (1), is characterized in that: The handwritten signature stability assessment method includes obtaining signature, signature similarity assessment and stability assessment, and adopts a signature segmentation matching calculation and partial matching segment difference value calculation, and the method steps are: A. Obtain signature: collect the signature to be tested and the signature sample set through the handwriting input device (1), where the signature to be tested is a handwritten signature that needs to be analyzed for stability, and the signature sample set is a handwritten signature that needs to be calculated for stability with the signature to be tested set; Let S represent a signature to be tested by user U, and set B={S ₁ , S ₂ , S ₃ ,..., S _N } represents the signature sample set of user U, where N≥1, that is: to obtain the signature, get A test signature S of user U and signature sample set B of user U; B. Signature similarity evaluation: get signature pair; Similar calculation of local matching segments; Signature pair similarity computation; According to the signature pair similarity calculation results of each group of signature pairs, it is judged whether the signature sample set is empty? Otherwise, go back to the step of obtaining the signature pair, if yes, complete the similarity evaluation between the signature to be tested and each signature sample; C. Stability assessment: Determine whether the number of similar signatures is less than the threshold T ₃ ? If the number of similar signatures is less than the threshold, the signature to be tested is unstable; if the number of similar signatures is greater than the threshold, the signature to be tested is stable, realizing the stability evaluation of the signature to be tested; In the B step, the operation process of the similar calculation of the local matching segment includes: Start, get a set of signature pairs, and start to execute the similar calculation operation process of the partial matching segment; Preprocessing, during the signature collection process, the signature data will be affected due to the interference of noise and the shaking of hands when signing, so preprocessing is performed on the original signature data of the signature to be tested and the signature sample in the group signature pair ; Signature segment matching calculation, using signature segment matching calculation to perform segment matching processing on the group of signature pairs, to obtain the partial segment sequence of the signature to be tested in the group signature pair and the partial segment sequence of the signature sample, and then the group signature pair The local segment sequence of the signature to be tested and the partial segment sequence of the signature sample adopt the local segment matching algorithm to obtain the corresponding relationship sequence with "one-to-one" characteristic between the partial segment sequence of the signature to be tested and the partial segment sequence of the signature sample, namely Partial match segment sequence; Calculate the difference value of the local matching segment. After completing the signature segment matching calculation of the signature to be tested and the signature sample in a group of signature pairs, the partial matching segment sequence corresponding to a group of signature pairs is obtained. From the local matching segment corresponding to this group of signature pairs Select a pair of local matching segments in the matching segment sequence, extract the local matching segment features corresponding to the local matching segments, including local segment static features and local segment dynamic features, and according to the extracted local segment static features and local segment dynamic features, After completing the calculation of the difference value of the partial matching segment, after selecting a pair of partial matching segments from the partial matching segment sequence, the segment number corresponding to the partial matching segment is deleted from the segment number of the partial matching segment sequence; Determine whether the difference value of the local matching segment is smaller than the threshold T ₁ ? If so, update the similar total length of the partial matching segment, if not, then judge whether the partial matching segment sequence is empty; Determine whether the sequence of partial matching segments is empty, if so, end the calculation of the similarity of the partial matching segments, if not, continue to calculate the difference value of the partial matching segments of the next pair of partial matching segments; End, the similar calculation operation process of the partial matching segment is completed; In step B, the operation process of the signed pair of similar calculations includes: Start, execute the signature pair similarity calculation operation flow; Calculate the similarity ratio R _O of the local matching segment. After completing the similarity calculation of the local matching segment of a group of signature pairs, the total similarity length of the local matching segment in the group of signature pairs is obtained, and the total similarity length of the local matching segment in the group of signature pairs is calculated. The partial segment sequence length ratio of the signature to be tested in the group signature pair is obtained to obtain the similarity ratio R _O of the local matching segment in the group signature pair; Determine whether the ratio R _O is smaller than the threshold T ₂ ? If yes, it is determined that the signature to be tested in the group of signature pairs is not similar to the signature sample, and the signature pair similarity calculation process is ended; if not, it is determined that the signature to be tested in the group of signature pairs is similar to the signature sample, and the statistics of similar signatures are updated quantity; At the end, the signature is completed for the similar calculation operation flow. 2. a kind of handwritten signature stability assessment method based on computer system according to claim 1, is characterized in that: described by handwritten input device (1) gathers handwritten signature data, wherein handwritten input device (1) comprises handwritten input module (11), a signal acquisition module (12), a signal processing module (13), and a display module (14). 3. a kind of handwritten signature stability assessment method based on computer system according to claim 2, it is characterized in that: described by handwriting input equipment (1) or be externally connected with the computing device or instrument of handwriting input equipment, or be PDA , handwriting input mobile phone, Tablet PC, server. 4. a kind of handwritten signature stability assessment method based on computer system according to claim 1, is characterized in that: in A step obtains signature, described by handwriting input device (1) gathers signature to be tested and signature sample set, Also includes extracting data from existing signature databases.

Images