Electronic signature verification method and device, storage medium and electronic equipment

Abstract

The invention discloses an electronic signature verification method and device, a storage medium and electronic equipment. Relates to the technical field of artificial intelligence. The method comprises the following steps: acquiring a to-be-verified target signature video generated by a target object in an interactive application; performing feature extraction on the target signature video through a target feature extraction model to obtain first feature information; second feature information is obtained from a database based on a label corresponding to the target object, and the second feature information is obtained based on a real signature video of the target object; and performing electronic signature verification according to the first feature information and the second feature information to obtain a verification result. Through the electronic signature verification method and device, the problem that the verification cost of the electronic signature is relatively high because the electronic signature pen of a special hardware device needs to be used for capturing the dynamic writing track in the signature process of the user to judge the authenticity of the signature of the user in the related technology is solved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and more specifically, to a method and device for verifying an electronic signature, a storage medium, and an electronic device. Background Art

[0002] Financial institutions have long faced challenges with offline signatures, including cumbersome procedures, lengthy review cycles, and security risk management. With the recent push to develop the digital economy and achieve the organic integration of the digital and real economies, electronic handwritten signatures, as a key enabler of digital transformation, are gaining widespread adoption across multiple industries and fields, offering convenient, efficient, and secure authentication. Continuous breakthroughs in next-generation information technologies, such as 5G, cloud computing, blockchain, artificial intelligence, and the Internet of Things, will further fuel the future development of the electronic handwritten signature industry.

[0003] In the field of electronic signature technology, especially in the customer signature verification process of financial institutions, current technical solutions face significant challenges and limitations. Mainstream electronic signature verification technology relies on an electronic signature pen with special hardware devices to capture the dynamic writing trajectory of the user during the signing process. The authenticity of the user's signature is determined by comparing the similarity between the user's real-time signature and the customer's reserved signature writing trajectory. This method greatly increases the difficulty of signature forgery due to the use of writing trajectory information. However, signature verification methods that rely on special hardware greatly increase implementation costs and also limit the scope of application scenarios. For example, the use of a designated electronic signature pen to complete signature collection and verification not only increases equipment investment, but also restricts the seamless integration of online business, making it difficult for customers to effectively verify electronic signatures in non-branch environments, seriously affecting the flexibility and convenience of business processes.

[0004] Currently, no effective solution has been proposed to the problem that electronic signature pens in related technologies require the use of special hardware devices to capture the dynamic writing trajectory of the user during the signing process to determine the authenticity of the user's signature, resulting in relatively high costs for electronic signature verification. Summary of the Invention

[0005] The main purpose of this application is to provide an electronic signature verification method and device, storage medium and electronic device to solve the problem in the related art that an electronic signature pen needs to use a special hardware device to capture the dynamic writing trajectory of the user's signature process to determine the authenticity of the user's signature, resulting in a relatively high cost for electronic signature verification.

[0006] To achieve the above objectives, according to one aspect of the present application, a method for verifying an electronic signature is provided. The method comprises: obtaining a target signature video to be verified, generated by a target subject in an interactive application; performing feature extraction on the target signature video using a target feature extraction model to obtain first feature information; obtaining second feature information from a database based on a tag corresponding to the target subject, wherein the second feature information is obtained based on the target subject's real signature video; and performing electronic signature verification based on the first and second feature information to obtain a verification result.

[0007] Furthermore, performing feature extraction on the target signature video through a target feature extraction model to obtain first feature information includes: performing feature extraction of a first scale on the target signature video through a first feature extraction layer in the target feature extraction model to obtain first initial feature information; performing feature extraction of a second scale on the target signature video through a second feature extraction layer in the target feature extraction model to obtain second initial feature information; and performing feature fusion based on the first initial feature information and the second initial feature information to obtain the first feature information.

[0008] Furthermore, performing feature extraction of a first scale on the target signature video through the first feature extraction layer in the target feature extraction model to obtain first initial feature information includes: performing feature extraction on the target signature video through a three-dimensional convolutional layer of multiple scales in the first feature extraction layer to obtain feature information of each scale; and processing the feature information of each scale to obtain the first initial feature information.

[0009] Furthermore, feature extraction is performed on the target signature video through the three-dimensional convolution layers of multiple scales in the first feature extraction layer to obtain feature information of each scale, including: feature extraction is performed on the target signature video through the first three-dimensional convolution layer of the three-dimensional convolution layers of the multiple scales to obtain feature information corresponding to the first three-dimensional convolution layer; feature extraction is performed on the target signature video and the feature information corresponding to the first three-dimensional convolution layer through the second three-dimensional convolution layer of the three-dimensional convolution layers of the multiple scales to obtain feature information corresponding to the second three-dimensional convolution layer; the step of performing feature extraction through the third three-dimensional convolution layer of the three-dimensional convolution layers of the multiple scales is repeated until the feature information corresponding to the last three-dimensional convolution layer is obtained, and feature information of each scale is obtained based on the currently obtained feature information.

[0010] Furthermore, processing the feature information of each scale to obtain the first initial feature information includes: performing random splicing processing on the feature information of each scale in pairs to obtain multiple spliced ​​feature information; performing random convolution calculation on the multiple spliced ​​feature information and the feature information of each scale in pairs to obtain multiple calculated feature information; and performing splicing processing on the multiple calculated feature information to obtain the first initial feature information.

[0011] Furthermore, the target feature extraction model is trained by the following steps: obtaining weight parameters of a pre-trained two-dimensional feature extraction model; constructing an initial feature extraction model based on the weight parameters; performing three-dimensional expansion processing on the initial feature extraction model to obtain a three-dimensional initial feature extraction model; and training the three-dimensional initial feature extraction model through a training sample set to obtain the target feature extraction model.

[0012] Furthermore, before obtaining the second feature information from the database based on the label corresponding to the target object, the method also includes: obtaining real signature videos of multiple objects, and batch processing the real signature videos to obtain multiple batches of real signature videos; performing feature extraction on the real signature videos of each batch through the target feature extraction model to obtain real feature information; constructing key-value pairs based on the real feature information of the multiple objects and the labels corresponding to the multiple objects, and storing the key-value pairs in the database.

[0013] To achieve the above-mentioned objectives, according to another aspect of the present application, a device for verifying an electronic signature is provided. The device comprises: a first acquisition unit for acquiring a target signature video to be verified, generated by a target subject in an interactive application; a first extraction unit for extracting features from the target signature video using a target feature extraction model to obtain first feature information; a second acquisition unit for acquiring second feature information from a database based on a tag corresponding to the target subject, wherein the second feature information is obtained based on the target subject's real signature video; and a verification unit for performing electronic signature verification based on the first feature information and the second feature information to obtain a verification result.

[0014] Furthermore, the first extraction unit includes: a first extraction subunit, used to perform first-scale feature extraction on the target signature video through the first feature extraction layer in the target feature extraction model to obtain first initial feature information; a second extraction subunit, used to perform second-scale feature extraction on the target signature video through the second feature extraction layer in the target feature extraction model to obtain second initial feature information; and a fusion subunit, used to perform feature fusion based on the first initial feature information and the second initial feature information to obtain the first feature information.

[0015] Furthermore, the first extraction subunit includes: an extraction module, used to extract features of the target signature video through three-dimensional convolution layers of multiple scales in the first feature extraction layer to obtain feature information of each scale; and a processing module, used to process the feature information of each scale to obtain the first initial feature information.

[0016] Furthermore, the extraction module includes: a first extraction submodule, used to perform feature extraction on the target signature video through the first three-dimensional convolution layer of the three-dimensional convolution layers of the multiple scales, and obtain feature information corresponding to the first three-dimensional convolution layer; a second extraction submodule, used to perform feature extraction on the target signature video and the feature information corresponding to the first three-dimensional convolution layer through the second three-dimensional convolution layer of the three-dimensional convolution layers of the multiple scales, and obtain feature information corresponding to the second three-dimensional convolution layer; an execution submodule, used to repeatedly perform the step of performing feature extraction through the third three-dimensional convolution layer of the three-dimensional convolution layers of the multiple scales until the feature information corresponding to the last three-dimensional convolution layer is obtained, and obtain feature information of each scale based on the currently obtained feature information.

[0017] Furthermore, the processing module includes: a first splicing submodule, used to perform pairwise random splicing processing on the feature information of each scale to obtain multiple spliced ​​feature information; a calculation submodule, used to perform pairwise random convolution calculation on the multiple spliced ​​feature information and the feature information of each scale to obtain multiple calculated feature information; a second splicing submodule, used to perform splicing processing on the multiple calculated feature information to obtain the first initial feature information.

[0018] Furthermore, the target feature extraction model is trained using the following devices: a third acquisition unit, used to obtain the weight parameters of a pre-trained two-dimensional feature extraction model; a construction unit, used to construct an initial feature extraction model based on the weight parameters; a processing unit, used to perform three-dimensional expansion processing on the initial feature extraction model to obtain a three-dimensional initial feature extraction model; a training unit, used to train the three-dimensional initial feature extraction model through a training sample set to obtain the target feature extraction model.

[0019] Furthermore, the device also includes: a fourth acquisition unit, used to obtain real signature videos of multiple objects before obtaining the second feature information from the database based on the label corresponding to the target object, and batch-process the real signature videos to obtain multiple batches of real signature videos; a second extraction unit, used to extract features from the real signature videos of each batch through the target feature extraction model to obtain real feature information; a construction unit, used to construct key-value pairs based on the real feature information of the multiple objects and the labels corresponding to the multiple objects, and store the key-value pairs in the database.

[0020] According to another aspect of an embodiment of the present invention, an electronic device is provided, including: a memory storing an executable program; and a processor for running the program, wherein the program executes any one of the above-mentioned electronic signature verification methods when running.

[0021] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided, which stores a program, wherein when the program is running, the device where the storage medium is located is controlled to execute any of the above-mentioned electronic signature verification methods.

[0022] In an embodiment of the present application, the following steps are adopted: obtaining a target signature video to be verified generated by a target object in an interactive application; performing feature extraction on the target signature video through a target feature extraction model to obtain first feature information; obtaining second feature information from a database based on a label corresponding to the target object, wherein the second feature information is obtained based on the real signature video of the target object; performing electronic signature verification based on the first feature information and the second feature information to obtain a verification result, which solves the problem in the related art that an electronic signature pen with a special hardware device is required to capture the dynamic writing trajectory of the user during the signing process to determine the authenticity of the user's signature, resulting in a relatively high cost for electronic signature verification.

[0023] In this solution, signature videos to be verified are collected through interactive applications with the target object. A pre-trained target feature extraction model is then used to analyze and extract features from the collected signature videos to obtain first feature information. Simultaneously, based on the specific target object identifier, previously stored second feature information is retrieved from a database. Finally, the first feature information collected in real time is compared with the second feature information in the database to obtain the aforementioned verification result. By simplifying hardware requirements and optimizing the feature extraction and comparison processes, this approach not only reduces the overall cost of electronic signature verification and improves operational convenience, but also achieves the technical effect of improving signature verification efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0025] Figure 1 A hardware structure block diagram of a computer terminal for implementing a method for verifying an electronic signature is shown;

[0026] Figure 2 is a flowchart of a method for verifying an electronic signature according to an embodiment of the present application;

[0027] Figure 3is a schematic diagram of a target feature extraction model provided according to an embodiment of the present application;

[0028] Figure 4 is a schematic diagram of a first feature extraction layer provided according to an embodiment of the present application;

[0029] Figure 5 This is a flowchart of the business processing provided in accordance with the embodiment of the present application;

[0030] Figure 6 is a schematic diagram of an electronic signature verification device provided according to an embodiment of the present application;

[0031] Figure 7 This is a structural block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0032] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0033] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in a sequence other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0034] It should be noted that the collected information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for display, data for analysis, etc.) involved in this application are information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data comply with relevant laws, regulations and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation portals for users to choose to authorize or refuse. For example, an interface is set up between this system and relevant users or institutions to provide users with corresponding operation portals for users to choose to agree or refuse the automated decision-making results; if the user chooses to refuse, the expert decision-making process will be entered.

[0035] Example 1

[0036] According to an embodiment of the present application, an embodiment of a method for verifying an electronic signature is also provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0037] The method embodiment provided in the first embodiment of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1 The hardware structure block diagram of a computer terminal (or mobile device) for implementing the electronic signature verification method is shown in FIG. Figure 1 As shown, the computer terminal 10 (or mobile device) may include one or more (illustrated as 102a, 102b, ..., 102n in the figure) processors 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. It will be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components than shown, or with Figure 1 Different configurations shown.

[0038] It should be noted that the one or more processors 102 and / or other data processing circuits described above may generally be referred to herein as "data processing circuitry". The data processing circuitry may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuitry may be a single independent processing module, or may be incorporated in whole or in part into any of the other components of the computer terminal 10 (or mobile device). As described in the embodiments of the present application, the data processing circuitry serves as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0039] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the electronic signature verification method in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, implementing the above-mentioned electronic signature verification method. The memory 104 may include a high-speed random access memory and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely located relative to the processor 102, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0040] The transmission device 106 is configured to receive or transmit data via a network. A specific example of the aforementioned network may include a wireless network provided by the communications provider of the computer terminal 10. In one embodiment, the transmission device 106 includes a network interface controller (NIC), which can be connected to other network devices via a base station to enable communication with the Internet. In another embodiment, the transmission device 106 may be a radio frequency (RF) module, which is configured to communicate with the Internet wirelessly.

[0041] The display may be a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 (or mobile device).

[0042] Under the above operating environment, this application provides Figure 2 The electronic signature verification method shown. Figure 2 Flowchart of a method for verifying an electronic signature according to Embodiment 1 of the present application. The method includes:

[0043] Step S201: obtaining a target signature video to be verified generated by a target object in an interactive application.

[0044] Optionally, when a user (i.e., the target subject) signs a financial transaction through an interactive application (e.g., a financial institution's app or an online financial institution interface), the interactive application will record the user's signature process to generate a current target signature video. The target signature video will capture all dynamic details of the user's signature, including the starting point, movement trajectory, speed, direction, and pauses of the stroke. Unlike traditional electronic signature verification methods based on a single image frame, signature data in video form can provide richer and more comprehensive information, helping to improve the accuracy and security of verification.

[0045] Step S202: extract features from the target signature video using a target feature extraction model to obtain first feature information.

[0046] Optionally, the target feature extraction model aims to extract information reflecting signature style and behavioral characteristics from the signature videos generated by users in the interactive application. The target feature extraction model can be a convolutional neural network that automatically identifies and extracts spatiotemporal features from video frames, including but not limited to: stroke shape and direction, stroke strength and continuity, and user writing habits such as pauses and overlapping strokes.

[0047] In an optional embodiment, the target signature video is segmented into a series of continuous image frames. These image frames are preprocessed, such as size normalization, grayscale conversion, or color space conversion, to meet the model's input requirements. The preprocessed frames are input into the target feature extraction model, which learns and extracts the dynamic writing features contained in the video through multi-layer convolution, pooling, and full connection operations. These features may be encoded as high-dimensional vectors for easy calculation and comparison. The target feature extraction model also considers the context of the entire video and integrates the features of each image frame to obtain a more comprehensive signature feature representation, namely, the first feature information mentioned above.

[0048] Step S203: obtaining second feature information from a database based on a tag corresponding to the target object, wherein the second feature information is obtained based on a real signature video of the target object.

[0049] Optionally, the database stores a large amount of pre-entered second feature information. For example, when a user first registers or sets up an electronic signature, they will be required to submit a video of their actual electronic handwritten signature (i.e., the aforementioned real signature video). The real signature video will be fed into the target feature extraction model, from which dynamic features related to the individual's writing habits will be extracted. These feature data (i.e., second feature information) will be associated with the user's identity information (i.e., the aforementioned tag) and stored in the database.

[0050] In an optional embodiment, when verifying a target subject's signature, the target subject corresponding to the video to be verified is first identified. This can be determined by reading the user ID contained in the video or by the user's login status in the interactive application. This identifier is then used to search the database for matching second feature information.

[0051] Step S204: perform electronic signature verification based on the first characteristic information and the second characteristic information to obtain a verification result.

[0052] Optionally, after obtaining the first and second feature information, the similarity or distance between the first and second feature information can be calculated to assess whether they match. For example, the cosine similarity between the first and second feature information can be calculated to obtain a result. The result can be compared with a preset threshold. If the similarity between the two sets of feature information exceeds the threshold, the signature is considered valid, i.e., verification succeeds. Conversely, if the similarity is below the threshold, the signature may be deemed invalid, i.e., verification fails.

[0053] For example, if the high degree of similarity between the first and second feature information exceeds a set threshold, the verification result will be marked as a valid signature. If the similarity is lower than the threshold, or there are obvious mismatches, the verification result will be marked as an invalid signature, and the user may be required to provide additional verification information. By accurately comparing the first and second feature information, the verification process can effectively distinguish between legitimate signatures and potential forged signatures, enhancing the security and credibility of electronic signatures. Compared to traditional methods that rely solely on static images, this method based on dynamic video features can provide more verification dimensions, reduce the error rate, and improve overall verification accuracy. It also makes the signing process more difficult to imitate or deceive.

[0054] In summary, the target object's interactive application collects signature videos to be verified. A pre-trained target feature extraction model is then used to analyze and extract features from the collected signature videos to obtain first feature information. Simultaneously, based on the specific target object identifier, previously stored second feature information is retrieved from a database. Finally, the real-time collected first feature information is compared with the second feature information in the database to obtain the aforementioned verification result. By simplifying hardware requirements and optimizing the feature extraction and comparison processes, the overall cost of electronic signature verification is reduced and operational convenience is improved, thereby achieving the technical effect of improving signature verification efficiency.

[0055] Optionally, in the electronic signature verification method provided in the embodiment of the present application, feature extraction is performed on the target signature video through a target feature extraction model to obtain first feature information, including: performing feature extraction of a first scale on the target signature video through a first feature extraction layer in the target feature extraction model to obtain first initial feature information; performing feature extraction of a second scale on the target signature video through a second feature extraction layer in the target feature extraction model to obtain second initial feature information; and performing feature fusion based on the first initial feature information and the second initial feature information to obtain first feature information.

[0056] In an optional embodiment, the target feature extraction model is deployed with a dual-path feature extraction layer (i.e., the first feature extraction layer and the second feature extraction layer mentioned above). The first feature extraction layer can use a smaller convolution kernel (such as 3x3x3) for feature extraction to obtain the above-mentioned first initial feature information; these smaller convolution kernels can efficiently capture the details of a single stroke or local area in the video. Compared with the first layer, the task of the second feature extraction layer is to capture deeper and more abstract features, especially those related to the dynamics of writing behavior, the relationship between strokes, and the overall style of the signature. The second feature extraction layer uses a larger convolution kernel (such as 5x5x5 or larger) for feature extraction to obtain the above-mentioned second initial feature information. These larger convolution kernels and deep network structures can analyze a wider range of spatiotemporal relationships and understand the inherent logic and dynamic behavior of the signature in more detail.

[0057] After respectively obtaining the first initial feature information and the second initial feature information, feature fusion is performed on the first initial feature information and the second initial feature information. For example, the first feature information and the second initial feature information are fused by simple concatenation, weighted sum, attention mechanism, or a specially designed fusion network (sub-network), etc. to obtain the above-mentioned first feature information.

[0058] Through first-scale feature extraction, the model quickly captures basic visual features in the video and establishes a preliminary outline of the signature. Second-scale feature extraction further explores the dynamic properties and stylistic characteristics of the signature, enhancing the model's ability to understand the complexity of the signature. The final feature fusion step ensures that the model not only sees the appearance of the signature but also understands its essence, achieving higher accuracy and robustness during the verification process.

[0059] Optionally, in the electronic signature verification method provided in an embodiment of the present application, performing feature extraction of a first scale on the target signature video through the first feature extraction layer in the target feature extraction model to obtain first initial feature information includes: performing feature extraction on the target signature video through a three-dimensional convolution layer of multiple scales in the first feature extraction layer to obtain feature information of each scale; and processing the feature information of each scale to obtain the first initial feature information.

[0060] In an optional embodiment, the first feature extraction layer consists of multiple scaled 3D convolutional layers, each responsible for extracting features at a specific scale from the target signature video. These scales may vary in time and space. For example, some 3D convolutional layers may focus on local changes over a short period of time, while others may capture global information over a long period of time.

[0061] For example, the first feature extraction layer may include three-dimensional convolutional layers of 3x3x3, 5x5x5, and 7x7x7 scales. It should be noted that the second feature extraction layer may include convolution kernels of 11x11x11, 15x15x15, and 21x21x21. Feature extraction is performed on the target signature video using three-dimensional convolutional layers of multiple scales in the first feature extraction layer to obtain feature information at each scale. This feature information at each scale is then further processed, such as through pooling, normalization, the use of non-linear transformations such as activation functions like ReLU, and possible dimensionality reduction operations, to ensure that the format and dimensionality of the feature information are suitable for subsequent fusion and analysis. The processed feature information needs to be fused to form the first initial feature information. Fusion can employ a variety of strategies, including but not limited to feature concatenation, weighted summation, using attention mechanisms to emphasize features at certain scales, or integrating multi-scale information through a specially designed fusion sub-network.

[0062] The resulting first initial feature information is a powerful representation that combines features from different scales and fully captures the visual and dynamic characteristics of the target signature video. This multi-scale approach helps the model capture subtle nuances in signatures and maintains good generalization even with varying writing speeds, pen strength, or style.

[0063] By analyzing signatures at multiple scales, the model can extract more diverse features from signature videos, improving its ability to discern signature authenticity. Extracting features at different scales enhances the model's ability to withstand varying writing conditions, ensuring the verification process remains stable even when signatures are blurred or subject to background noise.

[0064] Optionally, in the electronic signature verification method provided in an embodiment of the present application, feature extraction is performed on the target signature video through the three-dimensional convolution layers of multiple scales in the first feature extraction layer to obtain feature information of each scale, including: feature extraction is performed on the target signature video through the first three-dimensional convolution layer of the three-dimensional convolution layers of multiple scales to obtain feature information corresponding to the first three-dimensional convolution layer; feature extraction is performed on the target signature video and the feature information corresponding to the first three-dimensional convolution layer through the second three-dimensional convolution layer of the three-dimensional convolution layers of multiple scales to obtain feature information corresponding to the second three-dimensional convolution layer; the step of performing feature extraction through the third three-dimensional convolution layer of the three-dimensional convolution layers of multiple scales is repeated until the feature information corresponding to the last three-dimensional convolution layer is obtained, and the feature information of each scale is obtained based on the currently obtained feature information.

[0065] In an optional embodiment, the first three-dimensional convolution layer performs primary feature extraction on the target signature video, for example, using a 3x3x3 convolution kernel to capture local texture, color or shape features within the video frame to obtain feature information corresponding to the first three-dimensional convolution layer mentioned above.

[0066] Based on the feature information output by the first 3D convolutional layer, the second 3D convolutional layer (e.g., a 5x5x5 convolution kernel) further explores wider areas and longer time series in the video to identify features such as connections between strokes and changes in writing speed. It should be noted that the second 3D convolutional layer receives not only the original video data but also the feature information from the first 3D convolutional layer as input. This allows it to not only continue to extract features from the video itself but also further analyze the results of the previous layer to capture more complex and in-depth features.

[0067] The above process is repeated until all the predetermined 3D convolutional layers have been traversed. Each new 3D convolutional layer extracts deeper features based on the previous layer, gradually building up the various dimensional features of the signature behavior.

[0068] Through a multi-level feature extraction process, the model can better understand the details in the signature video. The use of multi-scale three-dimensional convolutional layers enables the model to perform feature recognition at different levels, enhancing its sensitivity to signature details and the robustness of the verification process.

[0069] Optionally, in the electronic signature verification method provided in an embodiment of the present application, processing the feature information of each scale to obtain the first initial feature information includes: performing pairwise random splicing processing on the feature information of each scale to obtain multiple spliced ​​feature information; performing pairwise random convolution calculation on the multiple spliced ​​feature information and the feature information of each scale to obtain multiple calculated feature information; and performing splicing processing on the multiple calculated feature information to obtain the first initial feature information.

[0070] In an optional embodiment, feature information at each scale obtained from the first feature extraction layer is randomly selected and concatenated pairwise. This random concatenation helps the model learn multiple possible combinations of feature information at different scales, improving its generalization capabilities when faced with unknown or highly variable signature samples. Pairwise random convolution calculations are performed on the concatenated feature information and the feature information at each individual scale. This random convolution calculation can further delve into the feature information, uncovering more details and patterns hidden in the signature video. After completing the pairwise random convolution calculations, the multiple calculated feature information is concatenated to obtain the first initial feature information.

[0071] Through random splicing and random convolution, the feature information of electronic signature videos is effectively captured and integrated from multiple scales, which not only improves the dimension and complexity of the feature information, but also enhances the expressiveness and adaptability of the model in signature verification tasks, thereby achieving the technical effect of improving the accuracy of electronic signature verification.

[0072] Optionally, in the electronic signature verification method provided in the embodiment of the present application, the target feature extraction model is trained using the following steps: obtaining the weight parameters of a pre-trained two-dimensional feature extraction model; constructing an initial feature extraction model based on the weight parameters; performing three-dimensional expansion processing on the initial feature extraction model to obtain a three-dimensional initial feature extraction model; and training the three-dimensional initial feature extraction model using a training sample set to obtain a target feature extraction model.

[0073] In an optional embodiment, weight parameters are obtained from an existing deep learning model (i.e., the pre-trained 2D feature extraction model described above). For example, the weight parameters are obtained from a 2D convolutional neural network trained on a large amount of image data. The weight parameters of the pre-trained 2D feature extraction model have been learned and optimized over a long period of time and can effectively extract and understand features in the image, providing a good starting point for subsequent 3D models.

[0074] Then, the weight parameters obtained above are used to build an initial feature extraction model. The initial feature extraction model is two-dimensional and structurally matches the pre-trained two-dimensional feature extraction model, ensuring that these weight parameters can be directly loaded without changing its basic architecture. After building the two-dimensional initial feature extraction model, it is expanded into three dimensions, that is, it is converted into a three-dimensional model capable of processing video data. For example, the convolution kernel in the two-dimensional model (such as 3x3) is expanded to three dimensions (such as 3x3x3) to add the time dimension to the network, which can capture the relationship between video frames.

[0075] Finally, the three-dimensional initial feature extraction model is trained using a training sample set containing electronic signature videos and corresponding real clip labels to obtain the final target feature extraction model.

[0076] Through 3D extension processing, a 3D feature extraction model suitable for electronic signature videos can be quickly built. This approach avoids the extensive time and computing resources required to train the model from scratch, while also ensuring that the model can fully understand and utilize the rich spatiotemporal features in the video data.

[0077] Optionally, in the electronic signature verification method provided in the embodiment of the present application, before obtaining the second feature information from the database based on the label corresponding to the target object, the method also includes: obtaining real signature videos of multiple objects, and batch processing the real signature videos to obtain multiple batches of real signature videos; performing feature extraction on the real signature videos of each batch through a target feature extraction model to obtain real feature information; constructing key-value pairs based on the real feature information of multiple objects and the labels corresponding to the multiple objects, and storing the key-value pairs in the database.

[0078] In an optional embodiment, an interactive signature application or device allows users (i.e., the multiple subjects described above) to complete their handwritten signatures on a screen or other digital interface while simultaneously recording the entire signing process. The collected video is then processed in batches. Batching optimizes the allocation of computing resources and avoids excessive resource usage caused by processing large amounts of data all at once.

[0079] The target feature extraction model then extracts features from each batch of real signature videos. For example, preprocessing operations such as normalization, scaling, and cropping are performed on the video data to ensure consistency and high quality of the input data. The model then uses pre-configured components such as multi-scale 3D convolutional layers to extract feature information from the videos, including but not limited to handwriting dynamics, writing speed, and pressure changes. This information is then integrated to form the true feature information for each signature video.

[0080] Finally, the extracted real-world feature information is combined with the corresponding object (i.e., user)'s identity information (object tag) to construct a key-value pair. The key is typically a customer number or other unique identifier, and the value is the extracted feature information. This key-value pair is stored in a database. When verifying a user's electronic signature, the tag can be used to quickly locate their feature information, enabling fast query and comparison.

[0081] In an optional embodiment, the schematic diagram of the target feature extraction model is as follows Figure 3 As shown, the system includes a dual-path feature extraction layer (i.e., a first feature extraction layer and a second feature extraction layer), a feature fusion layer, and a classification head. The dual-path feature extraction layer extracts features from the signature video to be verified. The feature fusion layer then fuses the features extracted by the dual-path feature extraction layer. The classification head then compares and analyzes the fused features with the features corresponding to the real signature video to obtain the signature verification result.

[0082] In an optional embodiment, the schematic diagram of the first feature extraction layer is as follows: Figure 4 As shown, the first feature extraction layer includes: a three-dimensional convolution layer and a feature fusion layer. The first three-dimensional convolution layer performs primary feature extraction on the target signature video. For example, a 3x3x3 convolution kernel is used to capture local texture, color or shape features within the video frame to obtain the feature information corresponding to the first three-dimensional convolution layer mentioned above. Based on the feature information output by the first three-dimensional convolution layer, the second three-dimensional convolution layer (such as a 5x5x5 convolution kernel) further explores a wider area and a longer time series in the video to identify features such as the connection between strokes and changes in writing speed. The last three-dimensional convolution layer performs feature extraction again based on the features output by the second three-dimensional convolution layer and the original video, thereby obtaining the feature information output by each layer. Finally, the feature fusion layer is used to fuse the feature information output by each layer to obtain the final fused feature.

[0083] In an optional embodiment, taking the installment payment financial service as an example, the business process of incorporating electronic handwritten signature into the business process is as follows: Figure 5 As shown, the installment payment contract is confirmed through the APP, the user signs, and obtains the signature video. The identity is verified through the signature video. After confirmation, the contract is approved. After approval, a receipt is generated and returned to the user.

[0084] The electronic signature verification method provided in the embodiment of the present application obtains a target signature video to be verified generated by a target object in an interactive application; performs feature extraction on the target signature video through a target feature extraction model to obtain first feature information; obtains second feature information from a database based on a label corresponding to the target object, wherein the second feature information is obtained based on the real signature video of the target object; and performs electronic signature verification based on the first feature information and the second feature information to obtain a verification result, which solves the problem in the related art that an electronic signature pen with a special hardware device is required to capture the dynamic writing trajectory of the user during the signing process to determine the authenticity of the user's signature, resulting in a relatively high cost for electronic signature verification.

[0085] In this solution, signature videos to be verified are collected through interactive applications with the target object. A pre-trained target feature extraction model is then used to analyze and extract features from the collected signature videos to obtain first feature information. Simultaneously, based on the specific target object identifier, previously stored second feature information is retrieved from a database. Finally, the first feature information collected in real time is compared with the second feature information in the database to obtain the aforementioned verification result. By simplifying hardware requirements and optimizing the feature extraction and comparison processes, this approach not only reduces the overall cost of electronic signature verification and improves operational convenience, but also achieves the technical effect of improving signature verification efficiency.

[0086] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0087] Example 2

[0088] The present application also provides an electronic signature verification device. It should be noted that the electronic signature verification device of the present application can be used to execute the electronic signature verification method provided in the present application. The following describes the electronic signature verification device provided in the present application.

[0089] According to an embodiment of the present application, a device for implementing the above-mentioned electronic signature verification method is also provided, such as Figure 6 As shown, the device includes: a first acquisition unit 601, a first extraction unit 602, a second acquisition unit 603 and a verification unit 604.

[0090] The first acquisition unit 601 is configured to acquire a target signature video to be verified generated by a target object in an interactive application;

[0091] A first extraction unit 602 is configured to extract features from a target signature video using a target feature extraction model to obtain first feature information;

[0092] A second acquiring unit 603 is configured to acquire second feature information from a database based on a tag corresponding to the target object, wherein the second feature information is obtained based on a real signature video of the target object;

[0093] The verification unit 604 is configured to perform electronic signature verification based on the first characteristic information and the second characteristic information to obtain a verification result.

[0094] The electronic signature verification device provided in the embodiment of the present application obtains a target signature video to be verified generated by a target object in an interactive application through a first acquisition unit 601; a first extraction unit 602 performs feature extraction on the target signature video through a target feature extraction model to obtain first feature information; a second acquisition unit 603 obtains second feature information from a database based on a label corresponding to the target object, wherein the second feature information is obtained based on the real signature video of the target object; a verification unit 604 verifies the electronic signature based on the first feature information and the second feature information to obtain a verification result, which solves the problem in the related art that an electronic signature pen with a special hardware device is required to capture the dynamic writing trajectory of the user during the signing process to determine the authenticity of the user's signature, resulting in a relatively high cost for electronic signature verification.

[0095] In this solution, signature videos to be verified are collected through interactive applications with the target object. A pre-trained target feature extraction model is then used to analyze and extract features from the collected signature videos to obtain first feature information. Simultaneously, based on the specific target object identifier, previously stored second feature information is retrieved from a database. Finally, the first feature information collected in real time is compared with the second feature information in the database to obtain the aforementioned verification result. By simplifying hardware requirements and optimizing the feature extraction and comparison processes, this approach not only reduces the overall cost of electronic signature verification and improves operational convenience, but also achieves the technical effect of improving signature verification efficiency.

[0096] Optionally, in the electronic signature verification device provided in the embodiment of the present application, the first extraction unit includes: a first extraction subunit, used to perform first-scale feature extraction on the target signature video through the first feature extraction layer in the target feature extraction model to obtain first initial feature information; a second extraction subunit, used to perform second-scale feature extraction on the target signature video through the second feature extraction layer in the target feature extraction model to obtain second initial feature information; and a fusion subunit, used to perform feature fusion based on the first initial feature information and the second initial feature information to obtain first feature information.

[0097] Optionally, in the electronic signature verification device provided in an embodiment of the present application, the first extraction subunit includes: an extraction module, used to extract features of the target signature video through a three-dimensional convolution layer of multiple scales in the first feature extraction layer to obtain feature information of each scale; and a processing module, used to process the feature information of each scale to obtain first initial feature information.

[0098] Optionally, in the electronic signature verification device provided in the embodiment of the present application, the extraction module includes: a first extraction submodule, used to perform feature extraction on the target signature video through the first three-dimensional convolution layer of multiple-scale three-dimensional convolution layers, and obtain feature information corresponding to the first three-dimensional convolution layer; a second extraction submodule, used to perform feature extraction on the target signature video and the feature information corresponding to the first three-dimensional convolution layer through the second three-dimensional convolution layer of multiple-scale three-dimensional convolution layers, and obtain feature information corresponding to the second three-dimensional convolution layer; an execution submodule, used to repeatedly execute the step of performing feature extraction through the third three-dimensional convolution layer of multiple-scale three-dimensional convolution layers until the feature information corresponding to the last three-dimensional convolution layer is obtained, and obtain feature information of each scale based on the currently obtained feature information.

[0099] Optionally, in the electronic signature verification device provided in the embodiment of the present application, the processing module includes: a first splicing sub-module, used to perform pairwise random splicing processing on the feature information of each scale to obtain multiple spliced ​​feature information; a calculation sub-module, used to perform pairwise random convolution calculation on the multiple spliced ​​feature information and the feature information of each scale to obtain multiple calculated feature information; a second splicing sub-module, used to perform splicing processing on the multiple calculated feature information to obtain first initial feature information.

[0100] Optionally, in the electronic signature verification device provided in the embodiment of the present application, the target feature extraction model is trained using the following devices: a third acquisition unit, used to obtain the weight parameters of a pre-trained two-dimensional feature extraction model; a construction unit, used to construct an initial feature extraction model based on the weight parameters; a processing unit, used to perform three-dimensional expansion processing on the initial feature extraction model to obtain a three-dimensional initial feature extraction model; and a training unit, used to train the three-dimensional initial feature extraction model through a training sample set to obtain a target feature extraction model.

[0101] Optionally, in the electronic signature verification device provided in the embodiment of the present application, the device also includes: a fourth acquisition unit, used to obtain real signature videos of multiple objects before obtaining the second feature information from the database based on the label corresponding to the target object, and batch-process the real signature videos to obtain multiple batches of real signature videos; a second extraction unit, used to extract features from the real signature videos of each batch through a target feature extraction model to obtain real feature information; a construction unit, used to construct key-value pairs based on the real feature information of multiple objects and the labels corresponding to the multiple objects, and store the key-value pairs in the database.

[0102] It should be noted that the first acquisition unit 601, the first extraction unit 602, the second acquisition unit 603, and the verification unit 604 described above correspond to steps S201 to S204 in the first embodiment. The examples and application scenarios implemented by the four units and the corresponding steps are the same, but are not limited to the contents disclosed in the first embodiment. It should be noted that the above modules or units can be hardware components or software components stored in a memory (e.g., memory 104) and processed by one or more processors (e.g., processors 102a, 102b, ..., 102n). The above units can also be part of a device and can be run in the computer terminal 10 provided in the first embodiment.

[0103] Example 3

[0104] An embodiment of the present application may provide an electronic device, Figure 7 This is a structural block diagram of an electronic device according to an embodiment of the present application. Figure 7 As shown, the electronic device may include: one or more ( Figure 7 Only one is shown) processor 702, memory 704, storage controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0105] Among them, the memory can be used to store software programs and modules, such as program instructions / modules corresponding to the methods and devices in the embodiments of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, implementing the above-mentioned method. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the terminal via a network. Examples of the above-mentioned network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network and a combination thereof.

[0106] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: obtain the target signature video to be verified generated by the target object in the interactive application; extract features of the target signature video through the target feature extraction model to obtain first feature information; obtain second feature information from the database based on the label corresponding to the target object, wherein the second feature information is obtained based on the real signature video of the target object; perform electronic signature verification based on the first feature information and the second feature information to obtain a verification result.

[0107] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: extracting features of the target signature video through the target feature extraction model to obtain the first feature information, including: performing first-scale feature extraction on the target signature video through the first feature extraction layer in the target feature extraction model to obtain the first initial feature information; performing second-scale feature extraction on the target signature video through the second feature extraction layer in the target feature extraction model to obtain the second initial feature information; and performing feature fusion based on the first initial feature information and the second initial feature information to obtain the first feature information.

[0108] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: performing feature extraction of the target signature video at a first scale through the first feature extraction layer in the target feature extraction model to obtain the first initial feature information, including: performing feature extraction of the target signature video through the three-dimensional convolution layer of multiple scales in the first feature extraction layer to obtain feature information of each scale; processing the feature information of each scale to obtain the first initial feature information.

[0109] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: extracting features of the target signature video through the three-dimensional convolution layers of multiple scales in the first feature extraction layer to obtain feature information of each scale, including: extracting features of the target signature video through the first three-dimensional convolution layer of the three-dimensional convolution layers of multiple scales to obtain feature information corresponding to the first three-dimensional convolution layer; extracting features of the target signature video and the feature information corresponding to the first three-dimensional convolution layer through the second three-dimensional convolution layer of the three-dimensional convolution layers of multiple scales to obtain feature information corresponding to the second three-dimensional convolution layer; repeating the step of extracting features through the third three-dimensional convolution layer of the three-dimensional convolution layers of multiple scales until the feature information corresponding to the last three-dimensional convolution layer is obtained, and obtaining feature information of each scale based on the currently obtained feature information.

[0110] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: processing the feature information of each scale to obtain the first initial feature information, including: performing random splicing processing on the feature information of each scale in pairs to obtain multiple spliced ​​feature information; performing random convolution calculation on the multiple spliced ​​feature information and the feature information of each scale in pairs to obtain multiple calculated feature information; and performing splicing processing on the multiple calculated feature information to obtain the first initial feature information.

[0111] The processor can call the information and application programs stored in the memory through the transmission device to execute the following steps: the target feature extraction model is trained using the following steps: obtaining the weight parameters of the pre-trained two-dimensional feature extraction model; constructing an initial feature extraction model based on the weight parameters; performing three-dimensional expansion processing on the initial feature extraction model to obtain a three-dimensional initial feature extraction model; training the three-dimensional initial feature extraction model through a training sample set to obtain a target feature extraction model.

[0112] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: before obtaining the second feature information from the database based on the label corresponding to the target object, the method also includes: obtaining real signature videos of multiple objects, and batch processing the real signature videos to obtain multiple batches of real signature videos; extracting features from the real signature videos of each batch through the target feature extraction model to obtain real feature information; constructing key-value pairs based on the real feature information of the multiple objects and the labels corresponding to the multiple objects, and storing the key-value pairs in the database.

[0113] It can be understood by those skilled in the art that Figure 7 The structure shown is for illustration only, and the electronic device may also be a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a PDA, a mobile Internet device (MID), a PAD, or other terminal devices. Figure 7 It does not limit the structure of the above electronic device. For example, the electronic device may also include Figure 7 More or fewer components (such as network interfaces, display devices, etc.) shown in, or with Figure 7 Different configurations shown.

[0114] A person skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer-readable storage medium, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0115] Example 4

[0116] The embodiment of the present application further provides a computer-readable storage medium. Optionally, in this embodiment, the storage medium can be used to store the program code executed by the electronic signature verification method provided in the first embodiment.

[0117] Optionally, in this embodiment, the storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group.

[0118] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing the steps of the method for verifying an electronic signature.

[0119] The serial numbers of the above embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0120] In the above embodiments of the present application, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0121] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0122] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0123] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0124] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0125] The above is only a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for verifying an electronic signature, characterized in that: include: Obtain a target signature video to be verified generated by a target object in an interactive application; Extracting features from the target signature video using a target feature extraction model to obtain first feature information; Acquire second feature information from a database based on a tag corresponding to the target object, wherein the second feature information is obtained based on a real signature video of the target object; An electronic signature is verified based on the first characteristic information and the second characteristic information to obtain a verification result.

2. The method according to claim 1, characterized in that The target signature video is subjected to feature extraction by a target feature extraction model, and the first feature information obtained includes: Performing feature extraction of a first scale on the target signature video through a first feature extraction layer in the target feature extraction model to obtain first initial feature information; Performing feature extraction of a second scale on the target signature video using a second feature extraction layer in the target feature extraction model to obtain second initial feature information; Feature fusion is performed based on the first initial feature information and the second initial feature information to obtain the first feature information.

3. The method according to claim 2, characterized in that Performing first-scale feature extraction on the target signature video through the first feature extraction layer in the target feature extraction model to obtain first initial feature information includes: Performing feature extraction on the target signature video through the three-dimensional convolutional layers of multiple scales in the first feature extraction layer to obtain feature information at each scale; The feature information at each scale is processed to obtain the first initial feature information.

4. The method according to claim 3, characterized in that The target signature video is subjected to feature extraction by using the three-dimensional convolutional layers of multiple scales in the first feature extraction layer, and feature information of each scale is obtained, including: Performing feature extraction on the target signature video through the first three-dimensional convolutional layer of the multiple-scale three-dimensional convolutional layers to obtain feature information corresponding to the first three-dimensional convolutional layer; Extracting features from the target signature video and feature information corresponding to the first three-dimensional convolutional layer through a second three-dimensional convolutional layer of the multiple-scale three-dimensional convolutional layers to obtain feature information corresponding to the second three-dimensional convolutional layer; Repeat the step of performing feature extraction on the third three-dimensional convolutional layer of the three-dimensional convolutional layers of the multiple scales until feature information corresponding to the last three-dimensional convolutional layer is obtained, and obtain feature information of each scale based on the currently obtained feature information.

5. The method according to claim 3, characterized in that Processing the feature information of each scale to obtain the first initial feature information includes: Perform random splicing on each scale’s feature information to obtain multiple spliced ​​feature information; Perform pairwise random convolution calculation on multiple spliced ​​feature information and feature information of each scale to obtain multiple calculated feature information; The plurality of calculated feature information are concatenated to obtain the first initial feature information.

6. The method according to claim 1, wherein The target feature extraction model is trained using the following steps: Get the weight parameters of the pre-trained two-dimensional feature extraction model; Constructing an initial feature extraction model based on the weight parameters; Performing three-dimensional expansion processing on the initial feature extraction model to obtain a three-dimensional initial feature extraction model; The three-dimensional initial feature extraction model is trained using a training sample set to obtain the target feature extraction model.

7. The method according to claim 1, characterized in that Before acquiring the second feature information from the database based on the tag corresponding to the target object, the method further includes: Acquire real signature videos of multiple subjects, and process the real signature videos in batches to obtain multiple batches of real signature videos; Extract features from each batch of real signature videos using the target feature extraction model to obtain real feature information; Key-value pairs are constructed according to the real feature information of the multiple objects and the labels corresponding to the multiple objects, and the key-value pairs are stored in the database.

8. An electronic signature verification device, characterized in that: include: A first acquiring unit is configured to acquire a target signature video to be verified generated by a target object in an interactive application; A first extraction unit is configured to extract features from the target signature video using a target feature extraction model to obtain first feature information; A second acquiring unit is configured to acquire second feature information from a database based on a tag corresponding to the target object, wherein the second feature information is obtained based on a real signature video of the target object; The verification unit is used to perform electronic signature verification based on the first characteristic information and the second characteristic information to obtain a verification result.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored executable program, wherein when the executable program is run, the device where the computer-readable storage medium is located is controlled to execute the electronic signature verification method according to any one of claims 1 to 8.

10. An electronic device, characterized in that: include: a memory storing an executable program; A processor, configured to run the program, wherein the program, when running, executes the electronic signature verification method according to any one of claims 1 to 8.

Images