Online signature authentication method and device, computer device, computer readable storage medium and computer program product

Abstract

The application relates to an online signature authentication method and device, computer equipment, a computer readable storage medium and a computer program product, and relates to the technical field of artificial intelligence. The method comprises the following steps: obtaining a signature image of an online signature and time sequence information of the online signature based on a signature sequence of the online signature to be authenticated; the time sequence information is used to represent the time sequence characteristics of the writing track of the online signature; performing static feature extraction processing on the signature image to obtain static features of the online signature, and performing dynamic feature extraction processing on the time sequence information to obtain dynamic features of the online signature; performing enhancement processing on the dynamic features according to the static features to obtain enhanced dynamic features of the online signature; determining a signature distance between the online signature and a template signature of a target user corresponding to the online signature according to the enhanced dynamic features of the online signature, and performing signature authentication on the online signature based on the signature distance. The method can improve the authentication accuracy of the online signature.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to an online signature authentication method, apparatus, computer device, computer-readable storage medium, and computer program product. Background Technology

[0002] Signature authentication is the process of comparing a signature claiming to belong to a user with that user's genuine template signature to determine its authenticity.

[0003] In related technologies, online signature authentication can generally be divided into authentication based on global representation learning and authentication based on local representation learning. However, authentication based on global representation learning weakens the temporal details of online signatures, while authentication based on local representation learning ignores the global structural information that is particularly crucial for signature authentication. Therefore, both methods based on global and local representation learning result in low accuracy of online signature authentication. Summary of the Invention

[0004] Therefore, it is necessary to provide an online signature authentication method, apparatus, computer equipment, computer-readable storage medium, and computer program product that can improve the authentication accuracy of online signatures, addressing the aforementioned technical problem of low authentication accuracy of online signatures.

[0005] Firstly, this application provides an online signature authentication method, including:

[0006] Based on the signature sequence of the online signature to be authenticated, the signature image of the online signature and the temporal information of the online signature are obtained; the temporal information is used to characterize the temporal features of the writing trajectory of the online signature.

[0007] Using a pre-trained signature authentication model, static feature extraction is performed on the signature image to obtain the static features of the online signature; and dynamic feature extraction is performed on the time-series information to obtain the dynamic features of the online signature.

[0008] Based on the signature authentication model and the static features, the dynamic features are enhanced to obtain the enhanced dynamic features of the online signature;

[0009] Based on the signature authentication model and the enhanced dynamic features of the online signature, the signature distance between the online signature and the template signature of the target user corresponding to the online signature is determined, and the online signature is authenticated based on the signature distance.

[0010] In one embodiment, the dynamic feature includes a plurality of dynamic feature elements;

[0011] The step of enhancing the dynamic features based on the signature authentication model and the static features to obtain the enhanced dynamic features of the online signature includes:

[0012] Based on the signature authentication model and the static features, determine the static temporal attention weights corresponding to each dynamic feature element in the dynamic features;

[0013] Based on the signature authentication model and the static temporal attention weights corresponding to each dynamic feature element, the dynamic feature elements are enhanced to obtain the enhanced dynamic features of the online signature.

[0014] In one embodiment, determining the static temporal attention weights corresponding to each dynamic feature element in the dynamic features based on the signature authentication model and the static features includes:

[0015] Using the signature authentication model, each dynamic feature element in the static feature and the dynamic feature is concatenated to obtain the concatenated dynamic feature element corresponding to each dynamic feature element.

[0016] Based on the signature authentication model and each concatenated dynamic feature element, the static temporal attention weights corresponding to each dynamic feature element are obtained.

[0017] In one embodiment, the enhancement processing of each dynamic feature element based on the signature authentication model and the static temporal attention weights corresponding to each dynamic feature element to obtain the enhanced dynamic features of the online signature includes:

[0018] By fusing the signature authentication model with each dynamic feature element and the static temporal attention weight corresponding to each dynamic feature element, the first enhanced dynamic feature element corresponding to each dynamic feature element is obtained.

[0019] By using the signature authentication model, each dynamic feature element and the first enhanced dynamic feature element corresponding to each dynamic feature element are concatenated to obtain the second enhanced dynamic feature element corresponding to each dynamic feature element.

[0020] By combining the second enhanced dynamic feature elements through the signature authentication model, the enhanced dynamic features of the online signature are obtained.

[0021] In one embodiment, the signature sequence of the online signature includes the position information and writing pressure of the online signature at each sampling time point;

[0022] The process of obtaining the signature image and timing information of the online signature based on the signature sequence of the online signature to be authenticated includes:

[0023] Based on the position information and writing pressure of the online signature at each sampling time point, a signature image of the online signature is drawn, and a time function of the online signature corresponding to preset motion parameters is constructed.

[0024] The time function is determined as the timing information of the online signature.

[0025] In one embodiment, determining the signature distance between the online signature and the template signature of the target user corresponding to the online signature, based on the signature authentication model and the enhanced dynamic features of the online signature, includes:

[0026] When there are multiple template signatures, the intra-class distance between multiple template signatures is determined based on the enhanced dynamic features of each template signature, and the intra-class distance is used as a reference distance. In addition, the feature distance between the enhanced dynamic features of the online signature and the enhanced dynamic features of each template signature is determined respectively.

[0027] Based on the feature distances and the reference distance, the signature distance between the online signature and the multiple template signatures is determined.

[0028] Secondly, this application also provides an online signature authentication device, comprising:

[0029] The information acquisition module is used to obtain the signature image of the online signature and the temporal information of the online signature based on the signature sequence of the online signature to be authenticated; the temporal information is used to characterize the temporal features of the writing trajectory of the online signature.

[0030] The feature extraction and processing module is used to perform static feature extraction processing on the signature image through a pre-trained signature authentication model to obtain the static features of the online signature, and to perform dynamic feature extraction processing on the time-series information to obtain the dynamic features of the online signature.

[0031] An enhancement processing module is used to enhance the dynamic features based on the signature authentication model and the static features to obtain the enhanced dynamic features of the online signature.

[0032] The signature authentication module is used to determine the signature distance between the online signature and the template signature of the target user corresponding to the online signature based on the signature authentication model and the enhanced dynamic features of the online signature, and to perform signature authentication on the online signature based on the signature distance.

[0033] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0034] Based on the signature sequence of the online signature to be authenticated, the signature image of the online signature and the temporal information of the online signature are obtained; the temporal information is used to characterize the temporal features of the writing trajectory of the online signature.

[0035] Using a pre-trained signature authentication model, static feature extraction is performed on the signature image to obtain the static features of the online signature; and dynamic feature extraction is performed on the time-series information to obtain the dynamic features of the online signature.

[0036] Based on the signature authentication model and the static features, the dynamic features are enhanced to obtain the enhanced dynamic features of the online signature;

[0037] Based on the signature authentication model and the enhanced dynamic features of the online signature, the signature distance between the online signature and the template signature of the target user corresponding to the online signature is determined, and the online signature is authenticated based on the signature distance.

[0038] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0039] Based on the signature sequence of the online signature to be authenticated, the signature image of the online signature and the temporal information of the online signature are obtained; the temporal information is used to characterize the temporal features of the writing trajectory of the online signature.

[0040] Using a pre-trained signature authentication model, static feature extraction is performed on the signature image to obtain the static features of the online signature; and dynamic feature extraction is performed on the time-series information to obtain the dynamic features of the online signature.

[0041] Based on the signature authentication model and the static features, the dynamic features are enhanced to obtain the enhanced dynamic features of the online signature;

[0042] Based on the signature authentication model and the enhanced dynamic features of the online signature, the signature distance between the online signature and the template signature of the target user corresponding to the online signature is determined, and the online signature is authenticated based on the signature distance.

[0043] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0044] Based on the signature sequence of the online signature to be authenticated, the signature image of the online signature and the temporal information of the online signature are obtained; the temporal information is used to characterize the temporal features of the writing trajectory of the online signature.

[0045] Using a pre-trained signature authentication model, static feature extraction is performed on the signature image to obtain the static features of the online signature; and dynamic feature extraction is performed on the time-series information to obtain the dynamic features of the online signature.

[0046] Based on the signature authentication model and the static features, the dynamic features are enhanced to obtain the enhanced dynamic features of the online signature;

[0047] Based on the signature authentication model and the enhanced dynamic features of the online signature, the signature distance between the online signature and the template signature of the target user corresponding to the online signature is determined, and the online signature is authenticated based on the signature distance.

[0048] The aforementioned online signature authentication method, apparatus, computer device, computer-readable storage medium, and computer program product first obtain the signature image and temporal information of the online signature based on the signature sequence of the online signature to be authenticated; the temporal information is used to characterize the temporal features of the writing trajectory of the online signature; then, through a pre-trained signature authentication model, static feature extraction processing is performed on the signature image to obtain the static features of the online signature, and dynamic feature extraction processing is performed on the temporal information to obtain the dynamic features of the online signature; next, based on the signature authentication model and the static features, the dynamic features are enhanced to obtain the enhanced dynamic features of the online signature; finally, based on the signature authentication model and the enhanced dynamic features of the online signature, the signature distance between the online signature and the template signature of the target user corresponding to the online signature is determined, and signature authentication of the online signature is performed based on the signature distance. In this way, static features representing the global structure of online signatures can be obtained through the signature image, and dynamic features representing the temporal details of online signatures can be obtained through the temporal information. Based on the static features representing the global structure, the dynamic features representing the temporal details can be enhanced to obtain enhanced dynamic features that can represent both the global structure and temporal details of online signatures. Therefore, signature authentication based on enhanced dynamic features considers both the global structure and temporal details of online signatures, thereby improving the accuracy of online signature authentication. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is a schematic diagram of a signature authentication model in one embodiment;

[0051] Figure 2 This is a flowchart illustrating an online signature authentication method in one embodiment;

[0052] Figure 3 This is a flowchart illustrating the steps of enhancing dynamic features based on a signature authentication model and static features to obtain enhanced dynamic features for online signatures in one embodiment.

[0053] Figure 4 This is a flowchart illustrating the steps of determining the static temporal attention weights corresponding to each dynamic feature element in a dynamic feature based on a signature authentication model and static features in one embodiment.

[0054] Figure 5 This is a flowchart illustrating the steps of enhancing the dynamic features of an online signature based on a signature authentication model and the static temporal attention weights corresponding to each dynamic feature element.

[0055] Figure 6 This is a flowchart illustrating the steps of enhancing dynamic features based on a signature authentication model and static features to obtain enhanced dynamic features for online signatures, as shown in another embodiment.

[0056] Figure 7 This is a flowchart illustrating the online signature authentication method in another embodiment;

[0057] Figure 8 This is a structural block diagram of an online signature authentication device in one embodiment;

[0058] Figure 9 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0060] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0061] The online signature authentication method provided in this application embodiment can be achieved through, for example... Figure 1 The signature authentication model shown is implemented.

[0062] Reference Figure 1 The signature authentication model is a neural network model composed of multiple neural networks. These neural networks include a backbone network for dynamic feature extraction, a static temporal attention enhancement network for enhancement processing, and an authentication network for signature authentication. The backbone network is a convolutional recurrent neural network, the static temporal attention enhancement network includes a deep convolutional neural network for static feature extraction and a long short-term memory layer for calculating static temporal attention weights, and the authentication network is a neural network based on the dynamic time warping algorithm.

[0063] In one embodiment, such as Figure 2 As shown, an online signature authentication method is provided. This embodiment applies this method to devices equipped with, for example, [the following]. Figure 1 The server shown is an example of a signature authentication model. It can be understood that this method can also be applied to servers equipped with similar signature authentication models. Figure 1 The terminal of the signature authentication model shown can also be applied to terminals that include servers and terminals and are equipped with, for example, […]. Figure 1 The system illustrating the signature authentication model is implemented through the interaction between a server and a terminal. The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The terminal can be, but is not limited to, various personal computers, laptops, smartphones, tablets, etc. In this embodiment, the method includes the following steps:

[0064] Step S202: Based on the signature sequence of the online signature to be authenticated, obtain the signature image of the online signature and the timing information of the online signature.

[0065] Signature authentication is the process of comparing a signature claiming to belong to a user with the user's genuine template signature to determine its authenticity.

[0066] Online signatures refer to signatures written by the signer on a signing device (such as a smartphone or tablet). It should be noted that for online signatures, the position information and writing pressure at various sampling time points during the writing process are usually collected as the signature sequence of the online signature.

[0067] Among them, the signature sequence of online signatures is the position information and writing pressure of the online signature at each sampling time point in its writing process.

[0068] Among them, timing information is used to characterize the timing features of the writing trajectory of online signature during its writing process; in practical applications, timing information can be represented by time functions, such as the writing trajectory function, writing speed function, writing pressure function, and writing angle function of online signature.

[0069] Specifically, the signing device starts from the moment the signer begins writing on the signing device, and collects the position information and writing pressure of the strokes written by the signer on the signing device according to the preset sampling time points, so as to obtain the online signature sequence and upload the signature sequence to the server.

[0070] The server receives the signature sequence of the online signature uploaded by the signing device, and draws the signature image of the online signature based on the signature sequence. It also constructs a time function that characterizes the temporal features of the writing trajectory of the online signature during its writing process, thereby obtaining the temporal information of the online signature.

[0071] Step S204: Using the pre-trained signature authentication model, static feature extraction is performed on the signature image to obtain the static features of the online signature, and dynamic feature extraction is performed on the time sequence information to obtain the dynamic features of the online signature.

[0072] Among them, the signature authentication model is as follows: Figure 1 The example shown is a neural network model composed of multiple neural networks.

[0073] Among them, static features are used to characterize the global structural information of online signatures, that is, the global features of online signatures; dynamic features are used to characterize the temporal details of online signatures, that is, the local features of online signatures.

[0074] Specifically, see Figure 1 The server inputs the online signature image and the online signature timing information into the signature authentication model. The deep convolutional neural network in the static temporal attention enhancement network of the signature authentication model then extracts static features from the signature image to obtain the static features of the online signature. Furthermore, by using the backbone network in the signature authentication model, dynamic features of the time-series information are extracted to obtain the dynamic features of the online signature. .

[0075] In practical applications, the deep convolutional neural network is the VGG16 network (Visual Geometry Group 16 layers, a convolutional neural network proposed by the Visual Geometry Group at Oxford University).

[0076] In practical applications, the backbone network consists of two one-dimensional convolutional layers and two long short-term memory layers, with each convolutional layer followed by a ReLU (Rectified Linear Unit) activation function. The structure of the backbone network is shown in Table 1.

[0077] Table 1 Backbone Network Structure

[0078]

[0079] In practical applications, dynamic features are extracted based on the backbone network shown in Table 1. It has 128 dimensions.

[0080] Step S206: Based on the signature authentication model and static features, enhance the dynamic features to obtain the enhanced dynamic features of the online signature.

[0081] Among them, the enhanced dynamic features, compared with the dynamic features, enhance the parts that are beneficial to handwriting recognition in the signature authentication task, while suppressing the parts that have little impact on the signature authentication task.

[0082] Specifically, see Figure 1 The server enhances the long short-term memory layer in the static temporal attention enhancement network of the signature authentication model. Based on the static features representing global structural information, the dynamic features representing temporal details are enhanced, so that the dynamic features learn the static features. This enhances the parts of the dynamic features that are beneficial to handwriting recognition in the signature authentication task and suppresses the parts that have little impact on the signature authentication task, resulting in enhanced dynamic features that can represent both global structural information and temporal details.

[0083] Step S208: Based on the signature authentication model and the enhanced dynamic features of the online signature, determine the signature distance between the online signature and the template signature of the target user corresponding to the online signature, and perform signature authentication on the online signature based on the signature distance.

[0084] The target user is the user to whom the signer claims the online signature to be authenticated belongs.

[0085] The template signature is the actual signature entered by the target user into the server.

[0086] Specifically, see Figure 1The server, through the authentication network in the signature authentication model, calculates the signature distance between the online signature and the corresponding template signature based on the enhanced dynamic features of the online signature and the enhanced dynamic features of the template signature of the target user corresponding to the online signature. Then, the server calculates the signature distance based on the signature distance and a preset signature distance threshold. Online signature authentication is performed. If the signature distance is less than a preset signature distance threshold... If the online signature is determined to be the genuine signature of the target user, and the signature distance is greater than or equal to a preset signature distance threshold, then the online signature is considered to be genuine. If so, the online signature is determined to be a forged signature of the corresponding target user.

[0087] In practical applications, the server can predetermine the enhanced dynamic characteristics of the template signatures of each stored user.

[0088] In practical applications, the process of obtaining the enhanced dynamic features of template signatures is similar to that of online signatures, and will not be described in detail here.

[0089] The online signature authentication method described above involves the following steps: First, the server obtains the signature image and temporal information of the online signature based on the signature sequence of the online signature to be authenticated. The temporal information is used to characterize the temporal features of the writing trajectory of the online signature. Then, the server uses a pre-trained signature authentication model to perform static feature extraction processing on the signature image to obtain the static features of the online signature, and performs dynamic feature extraction processing on the temporal information to obtain the dynamic features of the online signature. Next, the server enhances the dynamic features based on the signature authentication model and the static features to obtain the enhanced dynamic features of the online signature. Finally, the server determines the signature distance between the online signature and the template signature of the target user corresponding to the online signature based on the signature authentication model and the enhanced dynamic features of the online signature, and performs signature authentication on the online signature based on the signature distance. In this way, through the signature image, the server can obtain static features representing the global structural information of the online signature, and through the temporal information, the server can obtain dynamic features representing the temporal details of the online signature. Based on the static features representing the global structural information, the server can enhance the dynamic features representing the temporal details to obtain enhanced dynamic features that can represent both the global structural information and the temporal details of the online signature. Therefore, the signature authentication based on enhanced dynamic features considers both the global structural information and the temporal details of the online signature, thereby improving the authentication accuracy of the online signature.

[0090] In one exemplary embodiment, the dynamic feature includes a plurality of dynamic feature elements.

[0091] like Figure 3As shown, step S206 above, which enhances the dynamic features based on the signature authentication model and static features to obtain the enhanced dynamic features of the online signature, specifically includes the following steps:

[0092] Step S302: Based on the signature authentication model and static features, determine the static temporal attention weights corresponding to each dynamic feature element in the dynamic features.

[0093] Step S304: Based on the signature authentication model and the static temporal attention weights corresponding to each dynamic feature element, enhance each dynamic feature element to obtain the enhanced dynamic features of the online signature.

[0094] Specifically, see Figure 1 For each dynamic feature element in the dynamic features, the server calculates the corresponding static temporal attention weight based on the static features through the long short-term memory layer in the static temporal attention enhancement network in the signature authentication model. Then, the server enhances the dynamic feature element based on the static temporal attention corresponding to the dynamic feature element through the static temporal attention enhancement network, so that the dynamic features learn the static features and obtain enhanced dynamic features that can represent both global structural information and temporal details.

[0095] In practical applications, the long short-term memory layer in a static temporal attention enhancement network is immediately followed by an activation function.

[0096] For example, the static characteristics of online signatures are: The dynamic characteristics of online signatures are ,in, For dynamic feature elements In dynamic features The corresponding timestamp; the server enhances the processing to make the dynamic features... Learning static features Enhanced dynamic features are obtained. .

[0097] In this embodiment, the server enhances the dynamic features representing temporal details by using static features that represent global structural information. This enables the dynamic features to learn from the static features, resulting in enhanced dynamic features that can represent both global structural information and temporal details. Online signature authentication is performed based on these enhanced dynamic features, taking into account both the global structural information and the temporal details of the online signature, thereby improving the authentication accuracy of the online signature.

[0098] In one exemplary embodiment, such as Figure 4As shown, step S302 above, based on the signature authentication model and static features, determines the static temporal attention weights corresponding to each dynamic feature element in the dynamic features, specifically including the following steps:

[0099] Step S402: Using the signature authentication model, concatenate each dynamic feature element in the static and dynamic features respectively to obtain the concatenated dynamic feature element corresponding to each dynamic feature element.

[0100] Step S404: Based on the signature authentication model and each concatenated dynamic feature element, obtain the static temporal attention weights corresponding to each dynamic feature element.

[0101] Among them, the static temporal attention weight is used to characterize the importance of the corresponding dynamic feature element in the signature authentication task. It is easy to understand that the larger the static temporal attention weight of the dynamic feature element, the more beneficial it is to handwriting recognition in the signature authentication task, while the smaller the static temporal attention weight of the dynamic feature element, the less impact it has on the signature authentication task.

[0102] Specifically, for each dynamic feature element in the dynamic features, the server concatenates the static features and the dynamic feature element through the static temporal attention enhancement network in the signature authentication model to obtain the concatenated dynamic feature element corresponding to the dynamic feature element; then, based on the long short-term memory layer and the activation function immediately following it in the static temporal attention enhancement network, and the concatenated dynamic feature element corresponding to each dynamic feature element, the server obtains the static temporal attention weight corresponding to each dynamic feature element.

[0103] In practical applications, the server will incorporate dynamic features. and static features As input to the static temporal attention enhancement network, static features are concatenated through the static temporal attention enhancement network. and dynamic features Each dynamic feature element in the network is used to obtain the corresponding concatenated dynamic feature elements. For example, a static temporal attention enhancement network concatenates static features. and dynamic feature elements The corresponding splicing dynamic feature elements are obtained. splicing static features and dynamic feature elements The corresponding splicing dynamic feature elements are obtained. ... splicing static features and dynamic feature elements The corresponding splicing dynamic feature elements are obtained. Then, the static temporal attention enhancement network concatenates the various dynamic feature elements. As input to the Long Short-Term Memory (LSTM) layer in a static temporal attention enhancement network, each dynamic feature element is calculated through the LSM layer and the activation function that follows it. The corresponding static temporal attention weights, for example, dynamic feature elements. Corresponding static temporal attention weights Dynamic feature elements Corresponding static temporal attention weights ..., dynamic feature elements Corresponding static temporal attention weights .

[0104] In this embodiment, the server uses a static temporal attention enhancement network to calculate the static temporal attention weights corresponding to each dynamic feature element in the dynamic features based on static features. This facilitates subsequent enhancement of each dynamic feature element based on the static temporal attention weights, enabling the dynamic features to learn from the static features. This enhances the parts of each dynamic feature element that are beneficial to handwriting recognition in the signature authentication task, while suppressing the parts that have little impact on the signature authentication task.

[0105] In one exemplary embodiment, such as Figure 5 As shown, step S304 above, based on the signature authentication model and the static temporal attention weights corresponding to each dynamic feature element, performs enhancement processing on each dynamic feature element to obtain the enhanced dynamic features of the online signature, specifically including the following steps:

[0106] Step S502: Through the signature authentication model, each dynamic feature element and the corresponding static temporal attention weight are fused to obtain the first enhanced dynamic feature element corresponding to each dynamic feature element.

[0107] Step S504: Using the signature authentication model, each dynamic feature element and the first enhanced dynamic feature element corresponding to each dynamic feature element are concatenated to obtain the second enhanced dynamic feature element corresponding to each dynamic feature element.

[0108] Step S506: Through the signature authentication model, combine each of the second enhanced dynamic feature elements to obtain the enhanced dynamic features of the online signature.

[0109] The fusion of each dynamic feature element and the corresponding static temporal attention weight refers to performing a dot product on each dynamic feature element and the corresponding static temporal attention weight.

[0110] Among them, splicing each dynamic feature element and the first enhanced dynamic feature element corresponding to each dynamic feature element refers to the residual connection of each dynamic feature element and the first enhanced dynamic feature element corresponding to each dynamic feature element.

[0111] Specifically, the server uses the static temporal attention enhancement network in the signature authentication model to perform dot product on each dynamic feature element and its corresponding static temporal attention weight, respectively, to obtain the first enhanced dynamic feature element corresponding to each dynamic feature element; for example, dot product of dynamic feature elements. Corresponding static temporal attention weights The corresponding first enhanced dynamic feature element is obtained. Dot product of dynamic feature elements Corresponding static temporal attention weights The corresponding first enhanced dynamic feature element is obtained. ..., dot product of dynamic feature elements Corresponding static temporal attention weights The corresponding first enhanced dynamic feature element is obtained. .

[0112] Then, the server uses a static temporal attention enhancement network to residually connect each dynamic feature element and its corresponding first enhanced dynamic feature element to obtain the second enhanced dynamic feature element for each dynamic feature element; for example, residually connecting dynamic feature elements... and the corresponding first enhanced dynamic feature element The corresponding second enhanced dynamic feature element is obtained. Residual connection dynamic feature elements and the corresponding first enhanced dynamic feature element The corresponding second enhanced dynamic feature element is obtained. ..., residual connection dynamic feature elements and the corresponding first enhanced dynamic feature element The corresponding second enhanced dynamic feature element is obtained. .

[0113] Finally, the server uses a static temporal attention enhancement network to combine the various second-enhanced dynamic feature elements to obtain the enhanced dynamic features of the online signature. .

[0114] In this embodiment, the server enhances each dynamic feature element through a static temporal attention enhancement network and static temporal attention weights corresponding to each dynamic feature element. This allows the dynamic features to learn static features, thereby enhancing the parts of each dynamic feature element that are beneficial to handwriting recognition in the signature authentication task and suppressing the parts that have little impact on the signature authentication task. This results in enhanced dynamic features that can represent both global structural information and temporal details.

[0115] In one exemplary embodiment, such as Figure 6The diagram shows another flowchart of step S206 above, in which dynamic features are enhanced based on the signature authentication model and static features to obtain enhanced dynamic features for online signatures.

[0116] In one exemplary embodiment, the online signature sequence includes the location information of the online signature at each sampling time point and the writing pressure.

[0117] The location information can be represented by the horizontal and vertical coordinates in a planar coordinate system.

[0118] Step S202 above, based on the signature sequence of the online signature to be authenticated, obtains the signature image and timing information of the online signature, specifically including the following: drawing the signature image of the online signature based on the position information and writing pressure of the online signature at each sampling time point, and constructing a time function of the online signature corresponding to preset motion parameters; and determining the time function as the timing information of the online signature.

[0119] The preset motion parameters include at least the horizontal coordinate of the planar coordinate system. The vertical coordinate of a plane coordinate system Writing pressure Horizontal velocity in a planar coordinate system vertical velocity in a planar coordinate system ,angle and its functions and ,speed First-order difference ,angle First-order difference Logarithmic radius of curvature Centripetal acceleration and total acceleration .

[0120] Specifically, the signature sequence of the online signature includes the horizontal coordinate of the online signature in the plane coordinate system at each sampling time point. The vertical coordinate of a plane coordinate system Writing pressure The server uses the horizontal coordinate of the plane coordinate system at each sampling time point based on the online signature. The vertical coordinate of a plane coordinate system Writing pressure Draw the signature image of the online signature, and calculate the horizontal coordinate of the online signature in the plane coordinate system. The vertical coordinate of a plane coordinate system Writing pressure Horizontal velocity in a planar coordinate system vertical velocity in a planar coordinate system ,angle and its functions and ,speed First-order difference ,angle First-order difference Logarithmic radius of curvature Centripetal acceleration and total acceleration There are twelve time functions in total; then, the server determines the timing information of the online signature based on the above twelve time functions.

[0121] In practical applications, the server uses Python's Matplotlib plotting library to plot the signature sequence as an offline signature image with a stroke width of 2.

[0122] In this embodiment, based on the signature sequence of the online signature to be authenticated, the server can obtain a signature image that represents global information of the online signature, and temporal information that represents local information of the online signature.

[0123] In an exemplary embodiment, the signature distance between the online signature and the template signature of the target user corresponding to the online signature is determined based on the signature authentication model and the enhanced dynamic features of the online signature. Specifically, this includes the following: when there are multiple template signatures, the intra-class distance between the multiple template signatures is determined based on the enhanced dynamic features of each template signature, and the intra-class distance is used as a reference distance; and the feature distance between the enhanced dynamic features of the online signature and the enhanced dynamic features of each template signature is determined respectively; and the signature distance between the online signature and the multiple template signatures is determined based on the feature distances and the reference distance.

[0124] Specifically, each user enters at least one template signature into the server. When a user has multiple template signatures, the server, through the authentication network, first determines the intra-class distance between the multiple template signatures based on the enhanced dynamic features of each template signature. This intra-class distance is then used as a reference distance for calculating the signature distance. Simultaneously, the server, through the authentication network, determines the feature distance between the enhanced dynamic features of the online signature and the enhanced dynamic features of each template signature for each template signature. Then, the server, through the authentication network, calculates the signature distance between the online signature and the overall multiple template signatures based on the feature distances between the enhanced dynamic features of the online signature and the enhanced dynamic features of each template signature, as well as the reference distance. The server considers any calculated signature distance less than a preset signature distance threshold. In this case, the online signature is determined to be the genuine signature of the corresponding target user.

[0125] In practical applications, the server calculates the signature distance based on Formula 1:

[0126] (Formula 1)

[0127] in, Enhanced dynamic features for online signatures, This refers to the signature distance corresponding to the online signature. For the number of template signatures, for The first template signature in the Enhanced dynamic features of template signatures; The reference distance is the intra-class distance between multiple template signatures; Enhanced dynamic features for online signatures and the first Feature distance between enhanced dynamic features of template signatures Calculated using Formula 2:

[0128] (Formula 2)

[0129] in, To enhance dynamic features and enhanced dynamic features Feature distance between them; To enhance dynamic features and enhanced dynamic features The DTW (Dynamic Time Warping) distance between them.

[0130] In this embodiment, when there are multiple template signatures, the server can calculate the signature distance between the online signature and the overall template signature based on the enhanced dynamic features of each template signature and the enhanced dynamic features of the online signature. This allows the server to determine the authenticity of the online signature based on the signature distance, thereby achieving signature authentication of the online signature.

[0131] In an exemplary embodiment, the signature distance between the online signature and the template signature of the target user corresponding to the online signature is determined based on the signature authentication model and the enhanced dynamic features of the online signature. Specifically, the method further includes the following: when there is only one template signature, the feature distance between the enhanced dynamic features of the online signature and the enhanced dynamic features of the template signature is determined; and the signature distance between the online signature and the template signature is determined based on the feature distance and a preset reference distance.

[0132] Specifically, referring to Formula 1, for the case where there is only one template signature, the server pre-sets a reference distance. The server first determines the feature distance between the enhanced dynamic features of the online signature and the enhanced dynamic features of the template signature, and then determines the signature distance between the online signature and the template signature based on the feature distance and the preset reference distance.

[0133] In practical applications, the preset reference distance is 1, meaning that when there is only one template signature, the reference distance is [value missing]. .

[0134] In an exemplary embodiment, the loss function of the authentication network in the pre-trained signature authentication model is as follows:

[0135] The server samples each user according to a triplet (anchor, positive sample, negative sample). For the first... users The server samples an enhanced feature sequence from a real signature. As an anchor point, sampling Enhanced feature sequences of real signatures As positive samples, sampling An enhanced sequence of randomly forged signatures or skillfully forged signatures. As negative samples, each user has a total of A triplet.

[0136] No. The triplet loss function for each user is shown in Equation 3:

[0137] (Formula 3)

[0138] in, It represents a non-negative margin.

[0139] The overall loss function of the authentication network is shown in Equation 4:

[0140] (Formula 4)

[0141] In one exemplary embodiment, such as Figure 7 As shown, another online signature authentication method is provided. Taking the application of this method to a server as an example, the steps include:

[0142] Step S702: Based on the signature sequence of the online signature to be authenticated, obtain the signature image of the online signature and the timing information of the online signature.

[0143] Step S704: Using the pre-trained signature authentication model, static feature extraction processing is performed on the signature image to obtain the static features of the online signature, and dynamic feature extraction processing is performed on the time sequence information to obtain the dynamic features of the online signature.

[0144] Step S706: Using the signature authentication model, concatenate each dynamic feature element in the static and dynamic features respectively to obtain the concatenated dynamic feature element corresponding to each dynamic feature element.

[0145] Step S708: Based on the signature authentication model and each concatenated dynamic feature element, obtain the static temporal attention weights corresponding to each dynamic feature element.

[0146] Step S710: Through the signature authentication model, each dynamic feature element and the corresponding static temporal attention weight are fused to obtain the first enhanced dynamic feature element corresponding to each dynamic feature element.

[0147] Step S712: Using the signature authentication model, each dynamic feature element and the first enhanced dynamic feature element corresponding to each dynamic feature element are concatenated to obtain the second enhanced dynamic feature element corresponding to each dynamic feature element.

[0148] Step S714: By combining each of the second enhanced dynamic feature elements through the signature authentication model, the enhanced dynamic features of the online signature are obtained.

[0149] Step S716: When there are multiple template signatures for the target user corresponding to the online signature, the intra-class distance between the multiple template signatures is determined based on the enhanced dynamic features of each template signature, and the intra-class distance is used as a reference distance. In addition, the feature distance between the enhanced dynamic features of the online signature and the enhanced dynamic features of each template signature is determined respectively.

[0150] Step S718: Determine the signature distance between the online signature and multiple template signatures based on each feature distance and reference distance.

[0151] Specifically, the signature distance is calculated based on the DTW algorithm.

[0152] In this embodiment, a static temporal attention mechanism is used to introduce global structural information of the static signature graph into the online signature authentication method based on the DTW algorithm, which is oriented towards local representation learning. This allows for the learning of signature representations that take into account both global and local information, thus overcoming the shortcomings of existing methods that only focus on local signature features and improving the accuracy of signature authentication.

[0153] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0154] Based on the same inventive concept, this application also provides an online signature authentication apparatus for implementing the online signature authentication method described above. The solution provided by this apparatus is similar to the implementation described in the above method; therefore, the specific limitations in one or more online signature authentication apparatus embodiments provided below can be found in the limitations of the online signature authentication method described above, and will not be repeated here.

[0155] In one exemplary embodiment, such as Figure 8 As shown, an online signature authentication device is provided, comprising: an information acquisition module 802, a feature extraction module 804, an enhancement processing module 806, and a signature authentication module 808, wherein:

[0156] The information acquisition module 802 is used to obtain the signature image of the online signature and the timing information of the online signature based on the signature sequence of the online signature to be authenticated; the timing information is used to characterize the timing features of the writing trajectory of the online signature.

[0157] The feature extraction module 804 is used to perform static feature extraction processing on the signature image through a pre-trained signature authentication model to obtain the static features of the online signature, and to perform dynamic feature extraction processing on the time sequence information to obtain the dynamic features of the online signature.

[0158] The enhancement processing module 806 is used to enhance the dynamic features based on the signature authentication model and static features to obtain the enhanced dynamic features of the online signature.

[0159] The signature authentication module 808 is used to determine the signature distance between the online signature and the template signature of the target user corresponding to the online signature based on the signature authentication model and the enhanced dynamic features of the online signature, and to perform signature authentication on the online signature based on the signature distance.

[0160] In one exemplary embodiment, the dynamic feature includes a plurality of dynamic feature elements.

[0161] The enhancement processing module 806 is also used to determine the static temporal attention weights corresponding to each dynamic feature element in the dynamic features based on the signature authentication model and static features; and to perform enhancement processing on each dynamic feature element based on the signature authentication model and the static temporal attention weights corresponding to each dynamic feature element to obtain the enhanced dynamic features of the online signature.

[0162] In an exemplary embodiment, the enhancement processing module 806 is further configured to, through the signature authentication model, concatenate each dynamic feature element in the static features and dynamic features respectively to obtain the concatenated dynamic feature element corresponding to each dynamic feature element; and based on the signature authentication model and each concatenated dynamic feature element, obtain the static temporal attention weight corresponding to each dynamic feature element.

[0163] In an exemplary embodiment, the enhancement processing module 806 is further configured to: fuse each dynamic feature element and the corresponding static temporal attention weight of each dynamic feature element respectively through a signature authentication model to obtain a first enhanced dynamic feature element corresponding to each dynamic feature element; concatenate each dynamic feature element and the corresponding first enhanced dynamic feature element of each dynamic feature element respectively through a signature authentication model to obtain a second enhanced dynamic feature element corresponding to each dynamic feature element; and combine each second enhanced dynamic feature element through a signature authentication model to obtain an enhanced dynamic feature for online signature.

[0164] In one exemplary embodiment, the online signature sequence includes the location information and writing pressure of the online signature at each sampling time point.

[0165] The information acquisition module 802 is also used to draw the signature image of the online signature based on the position information and writing pressure of the online signature at each sampling time point, and to construct the time function of the online signature corresponding to the preset motion parameters; and to determine the time function as the timing information of the online signature.

[0166] In an exemplary embodiment, the signature authentication module 808 is further configured to, when there are multiple template signatures, determine the intra-class distance between the multiple template signatures based on the enhanced dynamic features of each template signature, and use the intra-class distance as a reference distance, and determine the feature distance between the enhanced dynamic features of the online signature and the enhanced dynamic features of each template signature respectively; and determine the signature distance between the online signature and the multiple template signatures based on the feature distance and the reference distance.

[0167] The modules in the aforementioned online signature authentication device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0168] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 9 As shown, this computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores user template signatures and their enhanced dynamic features. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When executed by the processor, the computer program implements an online signature authentication method.

[0169] Those skilled in the art will understand that Figure 9 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0170] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0171] In one exemplary embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above-described method embodiments.

[0172] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0173] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0174] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0175] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. An online signature authentication method, characterized in that, The method includes: Based on the signature sequence of the online signature to be authenticated, the signature image of the online signature and the temporal information of the online signature are obtained; the temporal information is used to characterize the temporal features of the writing trajectory of the online signature. The online signature is obtained by performing static feature extraction on the signature image using a pre-trained signature authentication model. The online signature's static features are then obtained by performing dynamic feature extraction on the time-series information. The dynamic features include multiple dynamic feature elements. Using the signature authentication model, each dynamic feature element in the static feature and the dynamic feature is concatenated to obtain the concatenated dynamic feature element corresponding to each dynamic feature element. Based on the signature authentication model and each of the concatenated dynamic feature elements, the static temporal attention weights corresponding to each of the dynamic feature elements are obtained; By fusing each of the dynamic feature elements and the static temporal attention weights corresponding to each of the dynamic feature elements through the signature authentication model, the first enhanced dynamic feature element corresponding to each of the dynamic feature elements is obtained. By using the signature authentication model, each dynamic feature element and the first enhanced dynamic feature element corresponding to each dynamic feature element are concatenated to obtain the second enhanced dynamic feature element corresponding to each dynamic feature element. By combining each of the second enhanced dynamic feature elements through the signature authentication model, the enhanced dynamic features of the online signature are obtained; Based on the signature authentication model and the enhanced dynamic features of the online signature, the signature distance between the online signature and the template signature of the target user corresponding to the online signature is determined, and the online signature is authenticated based on the signature distance.

2. The method according to claim 1, characterized in that, The online signature sequence includes the position information and writing pressure of the online signature at each sampling time point; The process of obtaining the signature image and timing information of the online signature based on the signature sequence of the online signature to be authenticated includes: Based on the position information and writing pressure of the online signature at each sampling time point, a signature image of the online signature is drawn, and a time function of the online signature corresponding to preset motion parameters is constructed. The time function is determined as the timing information of the online signature.

3. The method according to claim 1 or 2, characterized in that, The step of determining the signature distance between the online signature and the template signature of the target user corresponding to the online signature, based on the signature authentication model and the enhanced dynamic features of the online signature, includes: When there are multiple template signatures, the intra-class distance between multiple template signatures is determined based on the enhanced dynamic features of each template signature, and the intra-class distance is used as a reference distance. In addition, the feature distance between the enhanced dynamic features of the online signature and the enhanced dynamic features of each template signature is determined respectively. Based on the feature distances and the reference distance, the signature distance between the online signature and the multiple template signatures is determined.

4. The method according to claim 3, characterized in that, The step of determining the feature distance between the enhanced dynamic features of the online signature and the enhanced dynamic features of each template signature includes: For each template signature, obtain the dynamic time-warped distance between the enhanced dynamic features of the online signature and the enhanced dynamic features of the template signature; Based on the dynamic time warping distance between the enhanced dynamic features of the online signature and the enhanced dynamic features of the template signature, the feature distance between the enhanced dynamic features of the online signature and the enhanced dynamic features of the template signature is obtained.

5. The method according to claim 1, characterized in that, The step of fusing each of the dynamic feature elements and the corresponding static temporal attention weights to obtain the first enhanced dynamic feature element corresponding to each of the dynamic feature elements includes: For each dynamic feature element, the dynamic feature element and its corresponding static temporal attention weight are multiplied by a dot product to obtain the first enhanced dynamic feature element corresponding to the dynamic feature element.

6. The method according to claim 1, characterized in that, The step of concatenating each of the dynamic feature elements and the corresponding first enhanced dynamic feature elements to obtain the corresponding second enhanced dynamic feature elements includes: For each dynamic feature element, a residual connection is made between the dynamic feature element and the first enhanced dynamic feature element corresponding to the dynamic feature element to obtain the second enhanced dynamic feature element corresponding to the dynamic feature element.

7. An online signature authentication device, characterized in that, The device includes: The information acquisition module is used to obtain the signature image of the online signature and the temporal information of the online signature based on the signature sequence of the online signature to be authenticated; the temporal information is used to characterize the temporal features of the writing trajectory of the online signature. The feature extraction module is used to perform static feature extraction processing on the signature image using a pre-trained signature authentication model to obtain the static features of the online signature, and to perform dynamic feature extraction processing on the time-series information to obtain the dynamic features of the online signature; the dynamic features include multiple dynamic feature elements. The enhancement processing module is used to: concatenate each dynamic feature element from the static features and the dynamic features respectively through the signature authentication model to obtain concatenated dynamic feature elements corresponding to each dynamic feature element; obtain static temporal attention weights corresponding to each dynamic feature element based on the signature authentication model and each concatenated dynamic feature element; fuse each dynamic feature element and its corresponding static temporal attention weights respectively through the signature authentication model to obtain first enhanced dynamic feature elements corresponding to each dynamic feature element; concatenate each dynamic feature element and its corresponding first enhanced dynamic feature elements respectively through the signature authentication model to obtain second enhanced dynamic feature elements corresponding to each dynamic feature element; and combine each of the second enhanced dynamic feature elements through the signature authentication model to obtain the enhanced dynamic features of the online signature. The signature authentication module is used to determine the signature distance between the online signature and the template signature of the target user corresponding to the online signature based on the signature authentication model and the enhanced dynamic features of the online signature, and to perform signature authentication on the online signature based on the signature distance.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images