Electronic signature information verification method

Abstract

The present application relates to the technical field of network security, and especially relates to a kind of electronic signature information verification method, the method comprises: obtaining the writing data of dotter pen, wherein the writing data includes: writing strength, handwriting coordinates and the coordinate occurrence time corresponding to each handwriting coordinate;After the writing data is transmitted to the server by connecting the dotter pen with the server, the writing data is compared with the preset multiple handwriting analysis models, to determine the signature similarity evaluation result corresponding to each handwriting analysis model in the writing data;The signature similarity evaluation result corresponding to each handwriting analysis model is weighted and accumulated to obtain the final score of the writing data;According to the final score, the validity of the signature corresponding to the writing data is judged.The present application improves the security and authentication efficiency of electronic signature information verification.

Description

Technical Field

[0001] This invention relates to the field of network security technology, and in particular to a method for verifying electronic signature information. Background Technology

[0002] With the rapid development of society and economy and the continuous progress of technology, signatures are playing an increasingly important role in business and daily life, and have great practical significance for society. At the same time, users have put forward higher requirements for the security and convenience of signature authentication.

[0003] Existing signature methods generally include manual signature, stamp, and digital certificate signature. However, these three methods have significant drawbacks. On the one hand, it is difficult to distinguish the authenticity and legitimacy of a signature using manual signature and stamp. That is, using manual signature or stamp for contract or business authentication is prone to forgery, making it difficult to protect the legitimate rights and interests of both parties. On the other hand, digital certificate signature is very inconvenient to use in daily life, and the authentication cost of digital certificate signature is high. That is, digital certificates require the computer to install relevant drivers and provide a real-name authentication certificate such as a USB key. For example, signing a loan agreement between ordinary people requires the use of a computer and software developed specifically for signing the loan agreement for authentication, which makes the signing operation very inconvenient.

[0004] In summary, existing electronic signature verification methods suffer from low signature security and cumbersome signature authentication processes. Summary of the Invention

[0005] The main objective of this invention is to provide an electronic signature information verification method, which aims to improve the security and authentication efficiency of electronic signature information verification.

[0006] To achieve the above objectives, the present invention provides an electronic signature information verification method, the electronic signature information verification method comprising:

[0007] The writing data of the dot matrix pen is obtained, wherein the writing data includes: writing pressure, pen stroke coordinates, and the time of occurrence of each pen stroke coordinate;

[0008] After the writing data is transmitted to the server via the dot matrix pen, the writing data is compared with multiple preset handwriting analysis models to determine the signature similarity evaluation result of the writing data in each of the handwriting analysis models.

[0009] The signature similarity evaluation results corresponding to each of the handwriting analysis models are weighted and accumulated to obtain the final score of the writing data;

[0010] The validity of the signature corresponding to the written data is determined based on the final score.

[0011] Optionally, the dot matrix pen is a writing pen equipped with an infrared recognition sensor camera and a pressure sensor;

[0012] When the dot matrix pen writes on a preset signature paper, the writing pressure, the coordinates of each stroke, and the time of occurrence of each stroke coordinate are obtained in real time. The signature paper is paper printed with a dot matrix pattern.

[0013] Optionally, the server is equipped with built-in software for receiving the writing data transmitted by the dot matrix pen;

[0014] The server also has built-in raw data samples and note analysis algorithms corresponding to each handwriting analysis model. The raw data samples include a set of raw data feature values, which are handwriting habit data that are compared and analyzed with the writing data.

[0015] After the server receives the signature comparison instruction, it compares the writing data uploaded by the dot matrix pen with the note analysis algorithm corresponding to each handwriting analysis model to obtain the signature similarity evaluation results.

[0016] The signature similarity evaluation results are weighted and accumulated to obtain the final score of the written data, and a legality threshold is set.

[0017] The validity of the signature corresponding to the written data is verified based on the final score and the legality threshold.

[0018] Optionally, the handwriting analysis model includes: an overall similarity model, a force / time analysis model, a handwriting corner analysis model, a stroke analysis model, a continuous stroke analysis model, and a writing continuity analysis model;

[0019] The note analysis algorithm includes: the overall similarity analysis algorithm corresponding to the overall similarity model, the force / time analysis algorithm corresponding to the force / time analysis model, the handwriting angle analysis algorithm corresponding to the handwriting angle analysis model, the pen tip analysis algorithm corresponding to the pen tip analysis model, the continuous stroke analysis algorithm corresponding to the continuous stroke analysis model, and the writing continuity analysis algorithm corresponding to the writing continuity analysis model.

[0020] Optionally, the original data samples are set in the original data feature value data model;

[0021] The original data feature value data model is used to collect pre-set handwritten signatures and handwritten texts by primitive people, and to generate the original data feature values ​​based on the handwritten signatures and handwritten texts, wherein there are multiple handwritten signatures and handwritten texts.

[0022] Optionally, the signature similarity assessment results include: overall similarity score, pressure / time similarity score, handwriting corner similarity score, pen stroke similarity score, cursive similarity score, and writing continuity similarity score;

[0023] The steps of the overall similarity analysis algorithm corresponding to the overall similarity analysis model include:

[0024] Obtain the latest signature information collected by the dot matrix pen, and determine the writing coordinate parameters corresponding to the signature information;

[0025] After scaling the overall signature appearance corresponding to the writing coordinate parameters to the maximum overlap with the comparison signature appearance corresponding to the original data feature values, the scaled overall signature appearance is compared with the comparison signature appearance to output the overall similarity result score.

[0026] The steps of the force / time analysis algorithm corresponding to the force / time analysis model include:

[0027] Determine the writing intensity / timetable parameter corresponding to the signature information, compare the writing intensity / timetable parameter with the writing intensity / timetable parameter corresponding to the original data feature value, and output the intensity / time similarity result score;

[0028] The steps of the handwriting angle analysis algorithm corresponding to the handwriting angle analysis model include:

[0029] Determine the handwriting coordinate change information corresponding to the signature information, and determine each writing corner subdivision type based on the handwriting coordinate change information, wherein the writing corner subdivision type includes horizontal bend, vertical hook and horizontal bend hook;

[0030] The writing corner sub-type is compared with the corner writing habit type corresponding to the original data feature value to output the handwriting corner similarity result score;

[0031] The steps of the brushstroke analysis algorithm corresponding to the brushstroke analysis model include:

[0032] The fine classification of each stroke is determined by the coordinate / force change of the signature information, wherein the fine classification of the stroke includes long vertical stroke, long downward stroke, lifting hook and pausing stroke;

[0033] The fine classification of the pen stroke writing style is compared with the pen stroke writing habit type corresponding to the feature value of the original data to output the pen stroke similarity result score;

[0034] The steps of the stroke analysis algorithm corresponding to the stroke analysis model include:

[0035] The fine classification of each stroke is determined by the coordinate / time / force changes of the signature information. The fine classification of strokes includes, but is not limited to, horizontal strokes connected to left-falling strokes, horizontal strokes connected to vertical strokes, hooks connected to dots, and dots connected to horizontal strokes.

[0036] The fine classification of cursive writing styles is compared with the cursive writing habit types corresponding to the feature values ​​of the original data to output the cursive similarity score.

[0037] The steps of the writing continuity analysis algorithm corresponding to the writing continuity analysis model include:

[0038] Determine the coordinates / time data corresponding to the signature information, and make a continuity judgment on the time occupied by each stroke in the overall signature of the signature information based on the coordinates / time data, so as to determine the continuity data of the signature information;

[0039] The coherence data is compared with the writing habit coherence data corresponding to the feature values ​​of the original data to output the writing continuity similarity result score.

[0040] Optionally, the step of performing a weighted cumulative operation on the signature similarity evaluation results corresponding to each of the handwriting analysis models to obtain the final score of the writing data includes:

[0041] The overall similarity score, the force / time similarity score, the handwriting angle similarity score, the pen stroke similarity score, the cursive similarity score, and the writing continuity similarity score are statistically weighted according to preset weighting parameters to obtain the final score of the writing data.

[0042] Optionally, each of the handwriting analysis models is equipped with a convolutional neural network algorithm, which serves as a tool for training the original values ​​and comparing the differences between the samples and the original values.

[0043] The convolutional neural network algorithm is used to collect the writing data multiple times to determine the similarity feature value corresponding to each of the writing data, and to determine the signature similarity evaluation result corresponding to each of the handwriting analysis models according to the convolutional neural network algorithm.

[0044] Optionally, the step of determining the validity of the signature corresponding to the written data based on the final score includes:

[0045] The final score is compared with a preset total score threshold to obtain the signature result corresponding to the written data; wherein the signature result includes: confirming that the signature corresponding to the written data is genuine, confirming that the signature corresponding to the written data is not genuine, and being unable to determine.

[0046] When it is determined that the signature corresponding to the written data cannot be determined, the determination of the validity of the signature corresponding to the written data is abandoned based on the conclusion that it cannot be determined, or manual intervention is selected based on the conclusion that it cannot be determined, and the undetermined data is iterated back to the convolutional neural network algorithm to optimize the feature data of the original data sample.

[0047] In this invention, the application first acquires writing data from a dot matrix pen, including writing pressure, handwriting coordinates, and the time of occurrence of each handwriting coordinate. Then, the writing data is transmitted to the server via a connection between the dot matrix pen and the server. After receiving the writing data, the server compares it with multiple preset handwriting analysis models to determine the signature similarity evaluation results of the writing data in each handwriting analysis model. Then, the signature similarity evaluation results are weighted and accumulated to obtain the final score of the writing data. Finally, the validity of the signature corresponding to the writing data is determined based on the final score.

[0048] Unlike traditional signature methods, this application compares the acquired handwriting data with multiple pre-set handwriting analysis models on the server to obtain the signature similarity assessment results of the handwriting data in each handwriting analysis model. Then, the signature similarity assessment results are weighted and accumulated to obtain the final score of the handwriting data. The validity of the signature corresponding to the handwriting data is then determined based on the final score. This allows for convenient and quick detection of the authenticity of the signed paper containing the handwriting data, and effectively avoids the cumbersome digital certificate authentication process and the difficulty in distinguishing the authenticity of handwritten signatures, thereby improving the security and convenience of signature authentication. Attached Figure Description

[0049] Figure 1 This is a flowchart illustrating the first embodiment of the electronic signature information verification method of the present invention;

[0050] Figure 2 This is a schematic diagram illustrating the electronic signature information verification method according to an embodiment of the present invention;

[0051] Figure 3 This is a schematic diagram of the overall similarity analysis algorithm involved in the embodiment of the present invention;

[0052] Figure 4 This is a schematic diagram of the force / time analysis model involved in the embodiment of the present invention;

[0053] Figure 5 This is a schematic diagram of the handwriting angle analysis algorithm involved in the embodiment of the present invention;

[0054] Figure 6 This is a schematic diagram of the brushstroke analysis algorithm involved in the embodiment of the present invention;

[0055] Figure 7 This is a schematic diagram of the writing continuity analysis algorithm involved in the embodiment of the present invention;

[0056] Figure 8 This is a schematic diagram of the scoring process for the overall similarity analysis model.

[0057] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0058] This invention provides a method for verifying electronic signature information, referring to... Figure 1 As shown, Figure 1 This is a flowchart illustrating the first embodiment of the electronic signature information verification method of the present invention.

[0059] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application.

[0060] In this embodiment, the electronic signature information verification method of the present invention is applied to a terminal device for validating electronic signature information, and is specifically executed by the control center in the terminal device.

[0061] The electronic signature information verification method of the present invention includes:

[0062] Step S10: Obtain writing data of the dot matrix pen, wherein the writing data includes: writing pressure, pen coordinates and the time of occurrence of each pen coordinate;

[0063] In this embodiment, the control center first acquires the writing data of the dot matrix pen, wherein the writing data includes: writing pressure, pen stroke coordinates, and the time of occurrence of each of the pen stroke coordinates.

[0064] It should be noted that a dot matrix pen can be understood as an input device, a writing tool equipped with an infrared recognition sensor (such as a high-speed camera) and a pressure sensor. The working principle of a dot matrix pen can be understood as follows: the high-speed camera at the front of the pen captures the movement trajectory data of the pen tip in real time. At the same time, the pressure sensor of the pen acquires pressure data, and stores the pressure sensing data and movement trajectory data in the memory of the pen. In other words, when the pen tip is pressed down, the pressure sensor is triggered, and the built-in high-speed camera of the pen takes a picture of the dot matrix that the pen tip passes through on the signing paper. Then, the dot matrix coordinates, handwriting sequence, pressure data, writing time and other information are stored in the memory of the pen. That is, when the dot matrix pen writes on the signing paper, it can acquire writing data in real time, including writing pressure, the coordinates of each handwriting and the time of occurrence of each handwriting coordinate.

[0065] Writing data can be understood as the writing pressure, writing time, and stroke coordinates of the dot matrix pen tip on the signing paper. In other words, since a character has multiple strokes, each stroke includes the stroke coordinates (x, y) of the dot matrix pen on the signing paper, the time of occurrence of each stroke coordinate, and the writing pressure. That is, writing data can be understood as a triplet of information consisting of {coordinates (x, y), time, pressure}.

[0066] In this embodiment, the present application obtains the writing data on the signing paper by using a dot matrix pen, which effectively avoids the cumbersome operation process required by existing digital certificate signature authentication to obtain the electronic signature data on the signing paper, and effectively improves the verification efficiency and authentication convenience of electronic signatures.

[0067] Step S20: After the writing data is transmitted to the server via the dot matrix pen, the writing data is compared with multiple preset handwriting analysis models to determine the signature similarity evaluation result of the writing data in each of the handwriting analysis models.

[0068] In this embodiment, refer to Figure 2 , Figure 2 This is a schematic diagram of electronic signature information verification involved in an embodiment of the present invention. A dot matrix pen is connected to a user terminal via a mobile phone or other data transmission device, and the writing data is uploaded to a server. After confirming that the writing data is on the server of the user terminal, the writing data is compared with the original data feature values ​​in each handwriting analysis model to determine the signature similarity evaluation result corresponding to the writing data in each handwriting analysis model. Then, the signature similarity evaluation results are weighted and accumulated to obtain the final score of the writing data. Finally, the validity of the signature corresponding to the writing data is determined based on the final score.

[0069] It should be noted that the preset handwriting analysis models can be understood as algorithmic models that process writing data based on convolutional neural network algorithms. These models can include six types of analysis algorithms: overall similarity model, pressure / time analysis model, handwriting angle analysis model, pen stroke analysis model, cursive analysis model, and writing continuity analysis model. In addition, the application of each handwriting analysis algorithm model can be customized according to the usage scenario of the writing data and the actual needs of the user.

[0070] Additionally, it should be noted that the usage scenarios can include signatures on bank checks, lease agreements, business contracts, and IOUs / receipts.

[0071] For example, based on the signing situation of business contracts corresponding to the usage scenario, it can be determined that the overall similarity analysis model accounts for 10%, the stroke strength / time analysis model accounts for 25%, the handwriting corner analysis model accounts for 25%, the pen tip analysis model accounts for 10%, and the signature continuity analysis model accounts for 30%. In other words, the larger the weight ratio of a certain handwriting analysis model, the more decisive the signature similarity evaluation result corresponding to that handwriting analysis algorithm plays in determining whether the paper on which the writing data is located is valid.

[0072] Step 30: Determine the validity of the signature corresponding to the written data based on the final score.

[0073] In this embodiment, the control center determines the validity of the signature corresponding to the written data based on the final score.

[0074] In this embodiment, the present application performs weighted cumulative processing on the writing data using multiple preset handwriting analysis algorithm models to determine the signature similarity evaluation result corresponding to each handwriting analysis model. Then, a weighted cumulative operation is performed based on each signature similarity evaluation result to obtain the final score of the writing data. The validity of the signature paper containing the writing data is then detected based on the final score. The present invention obtains the final score of the writing data by performing a weighted cumulative operation, and then determines the validity of the signature corresponding to the writing data based on the final score, thereby further improving the accuracy of signature authentication on the signature paper containing the writing data and effectively ensuring the legality of the signature paper containing the writing data.

[0075] In another embodiment, this application can also determine the validity of the signature on the signature paper containing the written data based on the signature change analysis algorithm of the dot matrix pen and multiple handwriting analysis algorithms.

[0076] The algorithm obtains the signature change result score corresponding to the signature change analysis algorithm. Then, it performs a weighted summation operation on the signature similarity evaluation results and the signature change result score to obtain the final score of the written data. Finally, it determines the validity of the signature corresponding to the written data based on the final score.

[0077] It should be noted that the analysis of changes to signed documents mainly involves additions, rewritings, or alterations to other clauses made several seconds after the initial signature. Because dot-matrix pens record the writing on the same sheet of paper over time, this analysis of changes to signed documents is divided into time periods, allowing for the reproduction of each subsequent addition to the signed document.

[0078] In summary, in this invention, the application first acquires writing data from a dot-matrix pen, including writing pressure, handwriting coordinates, and the time of occurrence of each handwriting coordinate. Then, the writing data is transmitted to a server via a connection between the dot-matrix pen and the server. Upon receiving the writing data, the server compares it with multiple preset handwriting analysis models to determine the signature similarity evaluation results for each model. The signature similarity evaluation results are then weighted and accumulated to obtain the final score for the writing data. Finally, the validity of the signature corresponding to the writing data is determined based on the final score.

[0079] Unlike traditional signature methods, this application compares the acquired handwriting data with multiple pre-set handwriting analysis models on the server to obtain the signature similarity assessment results of the handwriting data in each handwriting analysis model. Then, the signature similarity assessment results are weighted and accumulated to obtain the final score of the handwriting data. The validity of the signature corresponding to the handwriting data is then determined based on the final score. This allows for convenient and quick detection of the authenticity of the signed paper containing the handwriting data, and effectively avoids the cumbersome digital certificate authentication process and the difficulty in distinguishing the authenticity of handwritten signatures, thereby improving the security and convenience of signature authentication.

[0080] Furthermore, based on the first embodiment of the electronic signature information verification of the present invention, a second embodiment of the electronic signature information verification of the present invention is proposed.

[0081] In this embodiment, the dot matrix pen is a writing pen equipped with an infrared recognition sensor camera and a pressure sensor;

[0082] When the dot matrix pen writes on a preset signature paper, the writing pressure, the coordinates of each stroke, and the time of occurrence of each stroke coordinate are obtained in real time. The signature paper is paper printed with a dot matrix pattern.

[0083] It should be noted that signature paper can be understood as paper printed with a dot matrix pattern that is invisible to the naked eye. In other words, the dot matrix pattern is a pattern composed of very tiny dots arranged according to a special algorithm. That is, the function of signature paper can be understood as providing coordinate parameter information to the dot matrix pen to ensure that the dot matrix pen can accurately record the signer's writing data when writing on the signature paper.

[0084] Furthermore, in some feasible embodiments, the server is equipped with built-in software for receiving the writing data transmitted by the dot matrix pen;

[0085] The server also has built-in raw data samples and note analysis algorithms corresponding to each handwriting analysis model. The raw data samples include a set of raw data feature values, which are handwriting habit data that are compared and analyzed with the writing data.

[0086] After the server receives the signature comparison instruction, it compares the writing data uploaded by the dot matrix pen with the note analysis algorithm corresponding to each handwriting analysis model to obtain the signature similarity evaluation results.

[0087] The signature similarity evaluation results are weighted and accumulated to obtain the final score of the written data, and a legality threshold is set.

[0088] The validity of the signature corresponding to the written data is verified based on the final score and the legality threshold.

[0089] Furthermore, in some other feasible embodiments, the handwriting analysis model includes: an overall similarity model, a force / time analysis model, a handwriting corner analysis model, a stroke analysis model, a continuous stroke analysis model, and a writing continuity analysis model;

[0090] The note analysis algorithm includes: the overall similarity analysis algorithm corresponding to the overall similarity model, the force / time analysis algorithm corresponding to the force / time analysis model, the handwriting angle analysis algorithm corresponding to the handwriting angle analysis model, the pen tip analysis algorithm corresponding to the pen tip analysis model, the continuous stroke analysis algorithm corresponding to the continuous stroke analysis model, and the writing continuity analysis algorithm corresponding to the writing continuity analysis model.

[0091] Furthermore, in some feasible embodiments, the original data samples are set in the original data feature value data model;

[0092] The original data eigenvalue data model is used to collect the written signatures and written texts preset by the original person, and generate the original data eigenvalues according to the written signatures and the written texts, where the number of the written signatures and the written texts is multiple.

[0093] Further, in some other feasible embodiments, the steps of the overall similarity analysis algorithm corresponding to the overall similarity analysis model include:

[0094] Step A10: Obtain the signature information newly collected by the dot matrix pen, and determine the writing coordinate parameters corresponding to the signature information;

[0095] In this embodiment, the control center first obtains the signature information newly collected by the dot matrix pen through the dot matrix pen, and then determines the writing coordinate parameters corresponding to the signature information. Among them, the signature information includes writing data. In other words, the signature information includes writing strength, handwriting coordinates, the coordinate occurrence time corresponding to each handwriting coordinate, and the change of handwriting coordinates, etc.

[0096] Step A20: After scaling the overall signature appearance corresponding to the writing coordinate parameters to the maximum overlap degree with the comparison signature appearance corresponding to the original data eigenvalue, compare the scaled overall signature appearance with the comparison signature appearance to output the overall similarity result score;

[0097] In this embodiment, the control center scales the overall signature appearance corresponding to the writing coordinate parameters to the maximum overlap degree with the comparison signature appearance corresponding to the original data eigenvalue, and then superimposes the scaled overall signature appearance on the comparison signature appearance for comparison to determine the overall similarity result score.

[0098] It should be noted that the overall signature appearance refers to the shape of each character. Refer to Figure 3 as shown. Figure 3 is a schematic diagram of the overall similarity analysis algorithm involved in the solution of the embodiment of the present invention. A represents the overall signature appearance corresponding to each of the two characters "Zhang San", and B represents the comparison signature appearance corresponding to each of the two characters "Zhang San" of the original data eigenvalue. For example, in this embodiment, the control center first intercepts the shape of "Zhang" in A, and then scales the proportion of "Zhang" intercepted from A to be the same or similar to "Zhang" in "Zhang San" in B (the maximum overlap degree) through scaling operation, and then superimposes the scaled A "Zhang" on B "Zhang" to obtain Figure 3 the C "Zhang" shown in Figure 3The result shown in the figure is the overall similarity score of A. Similarly, the overall similarity score of A can be obtained. Then, based on the overall similarity score of A and A, the overall similarity score of A is obtained.

[0099] The overall similarity score can be understood as the probability value of the signature appearance being legal, and is generally expressed as 0%-100%.

[0100] For example, refer to Figure 8 As shown, Figure 8 This is a schematic diagram of the scoring process for the overall similarity analysis model. When using the overall similarity model for scoring, the overall signature appearance will be substituted into the model. Figure 8 The weighted curve shown (i.e., comparing the appearance of the signature) indicates that when the similarity reaches 85%, the overall similarity score of the written data will be appropriately increased; however, when the similarity reaches 98%, the overall similarity score will decrease; and when it approaches 100%, the overall similarity score is suppressed to a very low level, thereby reducing the risk of using a computer or other technical means to copy the overall signature appearance exactly as it is, and thus effectively improving the correctness of electronic signature authentication.

[0101] In another embodiment, the steps of the force / time analysis algorithm corresponding to the force / time analysis model include:

[0102] Step B10: Determine the writing intensity / timetable parameter corresponding to the signature information, compare the writing intensity / timetable parameter with the writing intensity / timetable parameter corresponding to the original data feature value, and output the intensity / time similarity result score;

[0103] In this embodiment, refer to Figure 4 , Figure 4 This is a schematic diagram of the force / time analysis model involved in an embodiment of the present invention. In this embodiment, since the force and habits of the same person are consistent in the same writing, the control center first determines the writing force / time parameters corresponding to the signature information, and then compares the writing force / time parameters with the writing force / time parameters corresponding to the original data feature values ​​to output a force / time similarity score. In other words, within the same duration of stroke writing, by comparing the writing force / time parameters with the writing force / time parameters corresponding to the original data feature values, a force / time similarity score can be obtained.

[0104] It should be noted that the probability value of the intensity / time similarity result score can be expressed in the range of 0%-100%.

[0105] In another embodiment, the steps of the handwriting corner analysis algorithm corresponding to the handwriting corner analysis model include:

[0106] Step C10: Determine the handwriting coordinate change information corresponding to the signature information, and determine each type of writing corner subdivision according to the handwriting coordinate change information, where the types of writing corner subdivision include horizontal fold, vertical hook, and horizontal fold hook;

[0107] In this embodiment, the control center first determines the handwriting coordinate change information corresponding to the signature information, and then determines each type of writing corner subdivision according to the handwriting coordinate change information. The types of writing corner subdivision include, but are not limited to, horizontal fold, vertical hook, and horizontal fold hook. For example, refer to Figure 5 , Figure 5 is a schematic diagram of the handwriting corner analysis algorithm involved in the solution of the embodiment of the present invention. The first stroke of the character "Zhang" is composed of 1-8 triples (coordinates (x, y), strength, time). Points 1, 2, 3, and 4 are used as the coordinate information before the stroke corner, and points 5, 6, 7, and 8 are used as the coordinate information after the stroke corner. Then, the handwriting coordinate change information is determined according to the coordinate information before the stroke corner and the coordinate information after the stroke corner. Then, according to the writing corner corresponding to the handwriting coordinate change information, it is judged which type of writing corner subdivision it belongs to among the preset corner types. In other words, first obtain the horizontal vector according to points 1, 2, 3, and 4; then obtain the vector after the corner from points 4, 5, 6, 7, and 8, and then the angle and direction of this corner can be obtained. It can also be understood that the strength before the corner is analyzed according to points 1-4, and the strength change after the corner is obtained from points 4-8, so as to determine the type of writing corner subdivision.

[0108] Step C20: Compare the type of writing corner subdivision with the type of corner writing habit corresponding to the original data feature value to output the handwriting corner similarity result score;

[0109] In this embodiment, since each user has his own habits and characteristics when writing corners, including different corner angles, the control center compares the type of writing corner subdivision with the type of corner writing habit corresponding to the original data feature value to output the handwriting corner similarity result score.

[0110] It should be noted that the probability value representation form of the handwriting corner similarity result score can be expressed as 0%-100%.

[0111] In another embodiment, the steps of the pen tip analysis algorithm corresponding to the pen tip analysis model include:

[0112] Step D10: Determine each type of pen tip writing method subdivision through the coordinate / strength change of the signature information, where the types of pen tip writing method subdivision include long vertical stroke, long oblique stroke, hook stroke, and pause stroke;

[0113] In this embodiment, refer to Figure 6 , Figure 6 This is a schematic diagram of the pen stroke analysis algorithm involved in an embodiment of the present invention. A dot-matrix pen records the process of writing pressure decreasing from strong to weak. The value between two strokes that are weaker than a certain degree at this point is the pen stroke. The control center can determine the sub-categories of each pen stroke style by analyzing the coordinate / pressure changes in the signature information. These sub-categories include, but are not limited to, long vertical strokes, long downward strokes, hooks, and pauses.

[0114] Step D20: Compare the fine classification of the pen stroke writing style with the pen stroke writing habit type corresponding to the feature value of the original data to output the pen stroke similarity result score;

[0115] In this embodiment, the fine classification of pen strokes is compared with the pen stroke writing habit type corresponding to the feature value of the original data to output a pen stroke similarity score.

[0116] It should be noted that the probability value of the pen stroke similarity score can be expressed as 0%-100%.

[0117] In another embodiment, the steps of the stroke analysis algorithm corresponding to the stroke analysis include:

[0118] Step E10: Determine the subcategories of each connected stroke style based on the coordinate / time / force changes of the signature information. The subcategories of connected stroke styles include, but are not limited to, horizontal strokes connected to left-falling strokes, horizontal strokes connected to vertical strokes, hooks connected to dots, and dots connected to horizontal strokes.

[0119] In this embodiment, the control center can determine the subcategories of each cursive writing style by the coordinate / time / force changes of the signature information. The subcategories of cursive writing styles include, but are not limited to, horizontal strokes connected to left-falling strokes, horizontal strokes connected to vertical strokes, hooks connected to dots, and dots connected to horizontal strokes.

[0120] Step E20: Compare the fine classification of cursive writing with the cursive writing habit type corresponding to the feature value of the original data to output the cursive similarity result score;

[0121] In this embodiment, the control center compares the fine classification of cursive writing with the cursive writing habit type corresponding to the feature value of the original data to output the cursive similarity result score.

[0122] In some feasible embodiments, the control center first obtains the main connected strokes corresponding to the signature information through the connected stroke analysis algorithm, that is, the connected stroke types included in the fine classification of connected strokes. Then, it calls the pen tip analysis algorithm to determine the changes in pen tip strength in the main connected stroke. After determining the changes in pen tip strength in the signature connected stroke, the overall similarity analysis algorithm is used to compare the appearance of the main connected stroke and the connected strokes corresponding to the main connected stroke to analyze the appearance, and then the connected stroke similarity result score can be obtained.

[0123] It should be noted that the probability value representation of the connected stroke similarity result score can be expressed as 0% - 100%.

[0124] In another embodiment, the steps of the writing continuity algorithm corresponding to the writing continuity include:

[0125] Step F10: Determine the coordinate / time data corresponding to the signature information, and perform a coherence judgment on the time occupied by each stroke in the overall signature of the signature information according to the coordinate / time data, so as to determine the coherence data of the signature information;

[0126] In this embodiment, the control center first determines the coordinate / time data corresponding to the signature information, then determines the time occupied by each stroke in the overall signature of the signature information according to the coordinate / time data, and then performs a coherence judgment on the time occupied by each stroke. In other words, a coherence judgment is performed on the time occupied by each stroke of the entire signature.

[0127] For example, referring to Figure 7 , Figure 7 is a schematic diagram of the writing continuity analysis algorithm involved in the solution of the embodiment of the present invention. The total time for completing the entire signature of "good" is statistically calculated, and then the time occupied by each stroke is determined according to the total time of "good" and the duration between each breakpoint of "good", that is, the coherence data of the signature information.

[0128] Step F20: Compare the coherence data with the writing habit coherence data corresponding to the original data feature value to output the writing continuity similarity result score.

[0129] In this embodiment, the control center compares the coherence data with the writing habit coherence data corresponding to the original data feature value to output the writing continuity similarity result score.

[0130] It should be noted that the probability value representation of the writing continuity similarity result score can be expressed as 0% - 100%.

[0131] Furthermore, in some other feasible embodiments, the above step S30: Perform a weighted cumulative operation on the signature similarity evaluation results corresponding to each of the handwriting analysis models to obtain the final score of the writing data, and may further include the following implementation steps.

[0132] Step S301: Statistically weight the overall similarity result score, the force / time similarity result score, the handwriting angle similarity result score, the pen stroke similarity result score, the continuous stroke similarity result score, and the writing continuity similarity result score according to preset weighting parameters to obtain the final score of the writing data.

[0133] In this embodiment, the control center performs statistical weighting on the overall similarity result score, the force / time similarity result score, the handwriting angle similarity result score, the pen stroke similarity result score, the cursive similarity result score, and the writing continuity similarity result score according to preset weighting parameters to obtain the final score of the writing data.

[0134] Furthermore, in some feasible embodiments, each of the handwriting analysis models is equipped with a convolutional neural network algorithm, which serves as a tool for training the original values ​​and comparing the differences between the samples and the original values;

[0135] The convolutional neural network algorithm is used to collect the writing data multiple times to determine the similarity feature value corresponding to each of the writing data, and to determine the signature similarity evaluation result corresponding to each of the handwriting analysis models according to the convolutional neural network algorithm.

[0136] Furthermore, in some feasible embodiments, step S40 above, which determines the validity of the signature corresponding to the written data based on the final score, may also include the following implementation steps.

[0137] Step S401: Compare the final score with a preset total score threshold to obtain the signature result corresponding to the written data; wherein the signature result includes: confirming that the signature corresponding to the written data is genuine, confirming that the signature corresponding to the written data is not genuine, and being unable to determine.

[0138] In this embodiment, the control center compares the final score with a preset total score threshold to obtain the signature result corresponding to the written data; the signature result includes three situations: confirming that the signature corresponding to the written data is genuine, confirming that the signature corresponding to the written data is not genuine, and being unable to determine.

[0139] Step S402: When it is determined that the signature corresponding to the written data cannot be judged, the decision to abandon the judgment of the validity of the signature corresponding to the written data is abandoned based on the conclusion that it cannot be judged, or, based on the conclusion that it cannot be judged, manual intervention is selected for analysis, and the undetermined data is iterated back to the convolutional neural network algorithm to optimize the feature data of the original data sample.

[0140] In this embodiment, when it is determined that the signature corresponding to the written data cannot be determined, the decision to abandon the determination of the validity of the signature corresponding to the written data is made based on the conclusion that it cannot be determined. Alternatively, manual intervention is chosen based on the conclusion that it cannot be determined, and the undetermined data is iterated back to the convolutional neural network algorithm to optimize the feature data of the original data sample.

[0141] In summary, this application sets up six analysis algorithm models: overall similarity model, force / time analysis model, handwriting angle analysis model, stroke analysis model, continuous stroke analysis model, and writing continuity analysis model. These six analysis algorithm models can be randomly selected according to the usage scenario of the writing data. Then, the writing data is processed by each selected analysis algorithm model to obtain multiple signature similarity evaluation results. The signature similarity evaluation results are weighted and accumulated to obtain the final score of the writing data. The validity of the signature corresponding to the writing data is then determined based on the final score, thereby effectively improving the security and authentication efficiency of electronic signature verification.

[0142] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0143] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0144] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a computer-readable storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0145] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for verifying electronic signature information, characterized in that, The electronic signature information verification method includes: The writing data of the dot matrix pen is obtained, wherein the writing data includes: writing pressure, pen stroke coordinates, and the time of occurrence of each pen stroke coordinate; After the writing data is transmitted to the server via the dot matrix pen, the writing data is compared with multiple preset handwriting analysis models to determine the signature similarity evaluation result of the writing data in each of the handwriting analysis models. The signature similarity evaluation results corresponding to each of the handwriting analysis models are weighted and accumulated to obtain the final score of the writing data; The validity of the signature corresponding to the writing data is determined based on the final score; the handwriting analysis model includes: overall similarity model, pressure / time analysis model, handwriting angle analysis model, pen stroke analysis model, cursive analysis model, and writing continuity analysis model; The note analysis algorithm includes: the overall similarity analysis algorithm corresponding to the overall similarity model, the force / time analysis algorithm corresponding to the force / time analysis model, the handwriting angle analysis algorithm corresponding to the handwriting angle analysis model, the pen tip analysis algorithm corresponding to the pen tip analysis model, the continuous stroke analysis algorithm corresponding to the continuous stroke analysis model, and the writing continuity analysis algorithm corresponding to the writing continuity analysis model. The signature similarity assessment results include: overall similarity score, pressure / time similarity score, handwriting corner similarity score, pen stroke similarity score, cursive similarity score, and writing continuity similarity score; The steps of the overall similarity analysis algorithm corresponding to the overall similarity analysis model include: Obtain the latest signature information collected by the dot matrix pen, and determine the writing coordinate parameters corresponding to the signature information; After scaling the overall signature appearance corresponding to the writing coordinate parameters to the maximum overlap with the signature appearance corresponding to the original data feature values, the scaled overall signature appearance is compared with the signature appearance to output the overall similarity result score. The steps of the force / time analysis algorithm corresponding to the force / time analysis model include: Determine the writing intensity / timetable parameter corresponding to the signature information, compare the writing intensity / timetable parameter with the writing intensity / timetable parameter corresponding to the original data feature value, and output the intensity / time similarity result score; The steps of the handwriting angle analysis algorithm corresponding to the handwriting angle analysis model include: Determine the handwriting coordinate change information corresponding to the signature information, and determine each writing corner subdivision type based on the handwriting coordinate change information, wherein the writing corner subdivision type includes horizontal bend, vertical hook and horizontal bend hook; The writing corner sub-type is compared with the corner writing habit type corresponding to the original data feature value to output the handwriting corner similarity result score; The steps of the brushstroke analysis algorithm corresponding to the brushstroke analysis model include: The fine classification of each stroke is determined by the coordinate / force change of the signature information, wherein the fine classification of the stroke includes long vertical stroke, long downward stroke, lifting hook and pausing stroke; The fine classification of the pen stroke writing style is compared with the pen stroke writing habit type corresponding to the feature value of the original data to output the pen stroke similarity result score; The steps of the stroke analysis algorithm corresponding to the stroke analysis model include: The fine classification of each stroke is determined by the coordinate / time / force changes of the signature information. The fine classification of strokes includes horizontal strokes connected to left-falling strokes, horizontal strokes connected to vertical strokes, hooks connected to dots, and dots connected to horizontal strokes. The fine classification of cursive writing styles is compared with the cursive writing habit types corresponding to the feature values ​​of the original data to output the cursive similarity score. The steps of the writing continuity analysis algorithm corresponding to the writing continuity analysis model include: Determine the coordinates / time data corresponding to the signature information, and make a continuity judgment on the time occupied by each stroke in the overall signature of the signature information based on the coordinates / time data, so as to determine the continuity data of the signature information; The coherence data is compared with the writing habit coherence data corresponding to the feature values ​​of the original data to output the writing continuity similarity result score.

2. The electronic signature information verification method as described in claim 1, characterized in that, The dot matrix pen is a writing pen equipped with an infrared recognition sensor camera and a pressure sensor; When the dot matrix pen writes on a preset signature paper, the writing pressure, the coordinates of each stroke, and the time of occurrence of each stroke coordinate are obtained in real time. The signature paper is paper printed with a dot matrix pattern.

3. The electronic signature information verification method as described in claim 1, characterized in that, The server is equipped with built-in software, which is used to receive the writing data transmitted by the dot matrix pen. The server also has built-in raw data samples and note analysis algorithms corresponding to each handwriting analysis model. The raw data samples include a set of raw data feature values, which are handwriting habit data that are compared and analyzed with the writing data. After the server receives the signature comparison instruction, it compares the writing data uploaded by the dot matrix pen with the note analysis algorithm corresponding to each handwriting analysis model to obtain the signature similarity evaluation results. The signature similarity evaluation results are weighted and accumulated to obtain the final score of the written data, and a legality threshold is set. The validity of the signature corresponding to the written data is verified based on the final score and the legality threshold.

4. The electronic signature information verification method as described in claim 2, characterized in that, The original data samples are set in the original data feature value data model; The original data feature value data model is used to collect pre-set handwritten signatures and handwritten texts by primitive people, and to generate the original data feature values ​​based on the handwritten signatures and handwritten texts, wherein there are multiple handwritten signatures and handwritten texts.

5. The electronic signature information verification method as described in claim 1, characterized in that, The step of performing a weighted summation operation on the signature similarity evaluation results corresponding to each of the handwriting analysis models to obtain the final score of the writing data includes: The overall similarity score, the force / time similarity score, the handwriting angle similarity score, the pen stroke similarity score, the cursive similarity score, and the writing continuity similarity score are statistically weighted according to preset weighting parameters to obtain the final score of the writing data.

6. The electronic signature information verification method as described in claim 5, characterized in that, Each of the handwriting analysis models is equipped with a convolutional neural network algorithm, which serves as a tool for training the original values ​​and comparing the differences between the samples and the original values. The convolutional neural network algorithm is used to collect the writing data multiple times to determine the similarity feature value corresponding to each of the writing data, and to determine the signature similarity evaluation result corresponding to each of the handwriting analysis models according to the convolutional neural network algorithm.

7. The electronic signature information verification method as described in claim 6, characterized in that, The step of determining the validity of the signature corresponding to the written data based on the final score includes: The final score is compared with a preset total score threshold to obtain the signature result corresponding to the written data; wherein, the signature result includes: confirming that the signature corresponding to the written data is genuine, confirming that the signature corresponding to the written data is not genuine, and being unable to determine. When it is determined that the signature corresponding to the written data cannot be determined, the determination of the validity of the signature corresponding to the written data is abandoned based on the conclusion that it cannot be determined, or manual intervention is selected based on the conclusion that it cannot be determined, and the undetermined data is iterated back to the convolutional neural network algorithm to optimize the feature data of the original data sample.

Images