System and method for generating electronic signature handwriting database

Abstract

The invention discloses a system and method for generating an electronic signature handwriting database, and the system comprises a handwriting database main library, a handwriting database sub-library, and an authoritative backup library, the handwriting database main library carries out the data collection and storage of collected personal handwriting, and an algorithm model carries out the training, recognition, and identification tasks of the stored handwriting data; the chirography database sub-database is a chirography data sub-database matrix of a series of chirography data sub-databases, a data synchronization mechanism is adopted between the chirography data main database and the chirography data sub-databases, and chirography data information is exchanged between the chirography database main database and the authoritative backup database through data backup and update. The data isolated island of the handwriting data is broken, the data flow of the handwriting data is realized and accelerated, and a complete handwriting database construction scheme for the digital handwriting is constructed.

Description

Technical Field

[0001] The present invention relates to the field of big data and artificial intelligence technology, and in particular to a system and method for generating an electronic signature handwriting database. Background Art

[0002] Electronic signature digital handwriting is the electronic handwriting formed by the user writing according to his or her own writing habits on a device that can collect electronic digital signatures. The collection devices are diverse, including but not limited to mobile phones, tablets, and signature pads. The unique writing habits and unique writing methods of personal digital handwriting can reflect personal biometrics. Biometrics can be divided into physiological characteristics and behavioral characteristics. Physiological characteristics include face, fingerprints, irises, palm prints, vein patterns, etc., while behavioral characteristics include gait and handwriting, of which digital handwriting belongs to behavioral characteristics. Since handwriting behavioral characteristics are inherent characteristics formed by people's long-term writing habits, whether it is static handwriting characteristics such as writing contours and appearance, or dynamic behavioral characteristics of the ups and downs of the writing process, they are unique, difficult to copy, and easy to use. They are widely cited in application scenarios that emphasize personal identity, such as government affairs, law, banking and finance.

[0003] Therefore, it is obvious to establish a handwriting database infrastructure based on electronic signature digital handwriting to ensure the continuous updating of handwriting data, continuous iteration of handwriting algorithms, and continuous expansion of handwriting services. However, there is currently almost no complete handwriting database construction plan for digital handwriting. There are many problems with existing handwriting-related services or functions: handwriting data is isolated, and it is difficult to integrate and unify data between various business scenarios and collection devices; the handwriting data processing methods are single and scattered, and no standard operating procedures have been formed; the handwriting data flow is stagnant, and data efficiency has failed to stimulate the potential for sustained growth. In order to solve the current problem, the handwriting database infrastructure construction plan needs to solve practical problems such as handwriting data collection and storage, handwriting data processing iteration, and regular use of handwriting data. Summary of the invention

[0004] The present invention aims to solve the problems of handwriting data islands in the prior art, difficulty in integrating and unifying handwriting data of various business systems, and insufficient utilization of data efficiency, and solves the problems of handwriting data collection and storage, handwriting data processing iteration, and regular use of handwriting data.

[0005] Based on the first aspect of the present application, a personal handwriting database generation system is proposed, including a handwriting database main library, a handwriting database sub-library, and an authoritative backup library. The handwriting data sub-library includes a handwriting data sub-library matrix established according to different business system data sources, and the handwriting database sub-library type is determined according to different service module calling methods. The handwriting data main library adopts different data synchronization mechanisms according to different types of sub-libraries to synchronize handwriting data and related information. The handwriting database main library and the authoritative backup library exchange handwriting data and related information through data backup and update.

[0006] Further preferably, the data collection module standardizes all handwriting data obtained from the main database and stores them in a specified format; the algorithm module constructs a personal handwriting model in the sub-database according to different business system standards.

[0007] For further optimization, the handwriting data main library calls the data collection module to store the handwriting data from all sources in a specified format, and calls the algorithm module to process the handwriting data; the sub-library builds, iterates, and uses personal handwriting data in a standardized or customized manner according to the requirements of different business sub-nodes; the backup library is responsible for backing up the information of the handwriting data main library.

[0008] Further optimization, the type of handwriting database sub-library is distinguished according to different business needs and different data synchronization mechanisms are adopted. When the function service calling method uses the Internet public network for docking, it is determined to be a "Type A" sub-library, and its data synchronization service method adopts real-time synchronization; when the function service calling method is that the algorithm model is deployed on the Internet public network, and the remaining function modules are deployed in the demander's intranet environment, it is a "Type B" sub-library, and its data synchronization service method adopts periodic synchronization; when the function service calling method is that all function module service calls are deployed in the sub-library intranet environment, it is a "Type C" sub-library, and its data synchronization service adopts on-demand synchronization.

[0009] After further optimization, the data collection module is responsible for pre-processing and unified collection of multi-type data from multiple sources, multiple scenarios and multiple devices. The pre-processing includes: data item repair, historical data cleaning, business docking data quality verification, abnormal data marking, data classification and grading of the collected handwriting data, so as to obtain preliminary qualified handwriting data with complete basic data format, complete business data information and compatible data from various sources.

[0010] Further optimization, the storage of handwriting data from all sources in a prescribed format includes: the data collection module pre-processes the original data of personal biometric information, personal identity information, and collection device information collected from different sources to unify the full amount of handwriting information in a standard format; the handwriting algorithm model extracts, calculates, learns, and adjusts the handwriting data features to generate personal model data representing the personal handwriting writing habits; when the personal handwriting model compares and identifies handwritten handwriting, the sum of the algorithm model associated data forms the comparison model data; and the data governance process data used as intermediate data in the process of status identification of handwriting data.

[0011] Further optimization, the real-time synchronization includes: when the sub-library requests a service call from the main library, it synchronizes the relevant data required for the requested service to the main library in real time. The relevant data includes but is not limited to handwriting basic data, personal model data, and business information data. When the main library responds to the service request, it also synchronizes the corresponding service data to the sub-library in real time; the periodic synchronization is to periodically synchronize the handwriting related information required by the algorithm model only in the relevant links of the algorithm model and when it needs to be synchronized with the main library information; the on-demand synchronization is that when the sub-library contains the algorithm synchronization service and the algorithm synchronization service is enabled, the algorithm required by the sub-library will be synchronized with the corresponding algorithm of the main library, and the handwriting data involved in the service will be synchronized at the same time.

[0012] Further optimization is carried out to process the handwriting data into preliminary qualified handwriting data, which includes: basic handwriting data information, additional handwriting data information, and data collection business information.

[0013] After further optimization, the algorithm model implements training, recognition, and identification tasks on the stored handwriting data; according to the handwriting recognition goals and algorithm implementation path, the specific model selection, model framework selection, hidden layer design, activation function design, and loss function design of the personal handwriting model are determined.

[0014] For further optimization, the data update and backup between the main library and the backup library adopts the data update backup service. When the backup library only provides the backup function, the desensitized and encrypted personal handwriting feature data in the personal model data in the main library is backed up to the backup library. The main library regularly synchronizes the latest user handwriting feature information to the backup library, and the backup library regularly synchronizes the new sources collected by the main library; when the backup library not only provides the backup function but also provides the handwriting recognition service, in addition to the interactive user handwriting feature information, the main library also regularly synchronizes the latest handwriting recognition algorithm with the backup library, and the backup library regularly flows back the recognition request data.

[0015] Based on the second aspect of the present application, a handwriting database generation method is proposed, including establishing a handwriting data sub-library matrix according to different business system data sources and constructing different types of handwriting data sub-libraries based on different functional service calling methods, determining the handwriting database sub-library types and different data synchronization mechanisms between the handwriting data main library and different types of sub-libraries according to different business needs, and exchanging handwriting data information between the handwriting database main library and the authoritative backup library through data backup and update.

[0016] For further optimization, the data collection module standardizes the handwriting data of all sub-databases obtained from the main database and stores them in the prescribed format; the algorithm model calls the corresponding network model according to the business sub-node standard, determines the structure, number of layers and sequential construction of hidden layers including convolutional layer, pooling layer and fully connected layer, and calls the activation function design and loss function design to build a personal handwriting model in the sub-database.

[0017] For further optimization, when the function service module calling method uses the Internet public network for docking, the sub-database without functional service modules is a "Type A" sub-database, and its data synchronization service method uses real-time synchronization; when the function service module calling method is that the algorithm model is deployed in the main database, and the remaining functional modules are deployed in the sub-database, it is a "Type B" sub-database, and its data synchronization service method uses periodic synchronization; when the function service module calling method is that all functional service modules are deployed in the intranet environment of the sub-database corresponding to the business system, it is a "Type C" sub-database, and its data synchronization service uses on-demand synchronization.

[0018] Further optimization, the real-time synchronization includes: when the sub-library requests a service call from the main library, it synchronizes the requested service related data including handwriting basic data, personal model data, and business information data to the main library in real time. When the main library responds to the service request, it also synchronizes the corresponding service data to the sub-library in real time; the periodic synchronization is to synchronize the required handwriting basic data only when calling the algorithm model; the on-demand synchronization is to synchronize the sub-library algorithm and its handwriting data with the main library when the sub-library enables the algorithm synchronization service.

[0019] For further optimization, the data update and backup between the main library and the backup library adopts the data update backup service. When the backup library only provides the backup function, the desensitized and encrypted personal handwriting feature data in the personal model data in the main library is backed up to the backup library. The main library regularly synchronizes the latest user handwriting feature information to the backup library, and the backup library regularly synchronizes the new sources collected by the main library; when the backup library not only provides the backup function but also provides the handwriting recognition service, in addition to the interactive user handwriting feature information, the main library also regularly synchronizes the latest handwriting recognition algorithm with the backup library, and the backup library regularly flows back the recognition request data.

[0020] Further optimization, the storage of handwriting data from all sources in a prescribed format includes: the data collection module pre-processes the original data of personal biometric information, personal identity information, and collection device information collected from different sources to unify the full amount of handwriting information in a standard format; the handwriting algorithm model extracts, calculates, learns, and adjusts the handwriting data features to generate personal model data representing the personal handwriting writing habits; when the personal handwriting model compares and identifies handwritten handwriting, the sum of the algorithm model associated data forms the comparison model data; and the data governance process data used as intermediate data in the process of status identification of handwriting data.

[0021] The personal handwriting database generation system disclosed in the present invention breaks the "data island" of handwriting data through data collection, data storage, algorithm model, functional service core functions and the "main library-sub-library-backup library" construction plan based on the handwriting library construction plan, realizes and accelerates the data flow of handwriting data, stimulates the potential for continuous growth in the data efficiency of handwriting data, thereby ensuring the handwriting data development requirements of continuous updating of handwriting data, continuous iteration of handwriting algorithms, and continuous expansion of handwriting services, and constructs a complete handwriting database construction plan for digital handwriting. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A schematic diagram of the core architecture of the handwriting database in the exemplary embodiment of the present application; Figure 2 A schematic diagram of a data synchronization service in an exemplary embodiment of the application; Figure 3 A core data flow diagram of handwriting data in an exemplary embodiment of the application; Figure 4 This is a diagram of the data collection module architecture in an exemplary embodiment of the application; Figure 5 This is a diagram of the data storage module architecture in an exemplary embodiment of the application. DETAILED DESCRIPTION

[0023] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than 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 work are within the scope of protection of the present invention. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations.

[0024] The present invention constructs a handwriting database main library to collect and store collected personal handwriting data, and the algorithm model implements training, recognition and identification tasks on the stored handwriting data; the handwriting database sub-library is a handwriting data sub-library matrix of a series of handwriting data sub-libraries, and different data synchronization mechanisms are used between the handwriting data main library and the sub-libraries. The handwriting data information is exchanged between the handwriting database main library and the authoritative backup library through data backup and update.

[0025] like Figure 1The figure shows the core architecture of the handwriting database generation system in the exemplary embodiment of the present application. It includes: a handwriting database main library, a handwriting database sub-library, and an authoritative backup library. The handwriting database main library collects and stores the collected personal handwriting data, and builds an algorithm model to train, identify, and authenticate the handwriting data, and other handwriting task-related functional services; a handwriting data sub-library matrix is ​​built in the handwriting data sub-library, and the matrix includes a series of handwriting data sub-libraries (sub-libraries A, B...N). A data synchronization mechanism is used between the handwriting database main library and the sub-libraries, and the handwriting database main library and the authoritative backup library exchange handwriting data information through data backup and update.

[0026] The main database is configured with functional service modules such as data collection, data storage, algorithm model, and data synchronization, and related modules are configured in the sub-database according to business system requirements.

[0027] The handwriting database construction adopts the "main database-sub-database-backup database" architecture, in which the main database is responsible for the unified summary, processing and application of handwriting data from all data sources (including different business systems, different sub-databases); the sub-database is responsible for building, iterating and using the handwriting database in a standardized and customized manner according to the requirements of different business sub-nodes for handwriting data; the backup database is responsible for authoritative backup of the handwriting data main database information to ensure the availability and credibility of handwriting data information. Different business sub-nodes can choose to build what type of handwriting sub-database based on, but not limited to, the handwriting service network access requirements, the necessity and timeliness of handwriting model data updates. In the business sub-node, a handwriting sub-database matrix of a single type or multiple types of font combinations can be selected according to the actual situation of the specific business project.

[0028] like Figure 2 Shown is a schematic diagram of a data synchronization service in an exemplary embodiment of the present application.

[0029] The handwriting database main library and each sub-library complete the synchronization between the main library and each type of sub-library through real-time synchronization, regular synchronization, and on-demand synchronization. The handwriting database main library is taken as the highest root node, and the handwriting database sub-library is taken as the leaf node. The data synchronization service module is responsible for the data synchronization between the handwriting database main library (the highest root node) and the handwriting database sub-library (the leaf node), and different synchronization mechanisms are used according to the sub-library type. The handwriting database sub-library types are distinguished according to the different data synchronization mechanisms used for different business needs, and are distinguished based on the network environment of the service call, and are divided into three types of sub-libraries: Type A, Type B, and Type C.

[0030] When the function service call method uses the Internet public network for docking, the "Type A" sub-database uses real-time synchronization, that is, when the sub-database requests a service call from the main database, it synchronizes the relevant data required for the requested service to the main database in real time, and when the main database responds to the service request, it also synchronizes the corresponding service data in the main database to the sub-database in real time, thereby ensuring that the handwriting data in the sub-database is updated in real time and completely consistent with the handwriting data in the main database synchronized by the "Type A" sub-database in real time. Among them, the handwriting service mentioned above may include handwriting collection service, handwriting modeling service, handwriting recognition service, handwriting algorithm version update service and other services, and the "relevant data required for the service" varies according to the business purpose and service type. The service data includes but is not limited to full handwriting information, handwriting basic data, personal model data, business information data and other data. Taking the handwriting modeling service as an example, in order to meet the business needs of the modeling service, the user's personal identity information, collection device information and user's personal biometric information are required to complete the handwriting modeling service. When synchronizing data, the above categories of data need to be synchronized.

[0031] When the function service calling method is: except for the algorithm model deployed on the Internet public network, the other function modules (including: data collection, data storage, function services) are privately deployed in the intranet environment of each business system, which is called a "Type B" sub-library. For example, data collection, data storage, function services, etc. are independently operated by the sub-library builder in its own business system. When the "Type B" sub-library needs to request a handwriting service and the service needs to be accessed across the business intranet, on the premise of ensuring that the data when each function module interacts with the algorithm model meets the agreed data requirements, the data synchronization service will adopt a regular synchronization method, that is, only when the relevant links of calling the algorithm model in the handwriting service need to synchronize information with the main library, the service data corresponding to the algorithm model will be synchronized and exchanged to ensure that the handwriting data in the sub-library is completely consistent with the latest handwriting data in the main library that is regularly synchronized by the "Type B" sub-library. Because the data of the main database and the "Type B" sub-database will be synchronized regularly only when the algorithm model is called and the sub-database needs, supports and also includes the algorithm synchronization service, the data version of the "Type B" sub-database will be equal to or lag behind the data version of the main database; when the function service call method is: all function module services are implemented in the intranet environment of each business system, that is, including: algorithm model, data collection, data storage, and function service modules are deployed in its business system. The corresponding handwriting data sub-database is a "Type C" sub-database. Its data synchronization service will adopt an on-demand synchronization method, that is, only when the sub-database needs to start the algorithm synchronization service, the service data required by the sub-database will be synchronized with the corresponding service data of the main database, and the relevant handwriting data involved in the service will be synchronized at the same time.

[0032] The data update and backup service is mainly responsible for the data update and backup functions between the main database and the authoritative backup database, thereby ensuring the availability and credibility of the handwriting data main database. Its specific data interaction mechanism is determined by the functions provided by the authoritative backup database, which include: only providing backup, or providing handwriting recognition services while backing up.

[0033] When the backup library only needs to provide the backup function, only the desensitized and encrypted personal handwriting feature data in the personal model data in the main library data storage will be backed up in the backup library. Among them, data interaction is limited to the main library regularly synchronizing the latest user handwriting features to the backup library, and the backup library regularly synchronizes all data from the newly added collection sources collected by the main library collection module, including but not limited to personal biometric information data, personal identity information data, collection device information data, etc. from different sources of various business systems.

[0034] When the backup library not only provides backup functions but also provides handwriting recognition services, in addition to the desensitized and encrypted personal handwriting feature data in the personal model data, the handwriting recognition capability will also be deployed in the backup library. At this time, in addition to user feature information, the main library will also regularly synchronize the latest handwriting recognition algorithm with the backup library, and the backup library will regularly return the handwriting data used for recognition in the handwriting recognition service to ensure the continuous iteration of the handwriting recognition service. The handwriting database function control module includes: data collection module, data storage module, algorithm model module, function service module and other sub-function module combinations.

[0035] like Figure 3 The figure shows the handwriting core data flow diagram in the exemplary embodiment of the present application. The handwriting data from various business systems and different sources (data source A, data source B...data source N) are collected, the data is stored in a prescribed format, and the algorithm module is called to process the handwriting sea data. Functional services are then used to support various basic and application businesses of the handwriting database, among which the functional services include: sub-library data synchronization, backup library data interaction, and business applications.

[0036] like Figure 4The data collection module in the exemplary embodiment of the present application is shown as a schematic diagram of the data collection module architecture. The data collection module is responsible for the unified aggregation of multiple types of data from multiple business system sources, multiple scenarios, multiple equipment types, etc., after preliminary screening and preprocessing of the data, the preprocessing includes data item repair, historical data cleaning, business docking data quality verification, abnormal data marking, data classification and other functional processing for the acquired handwriting data. After the collection, the preliminary qualified handwriting data that meets the needs of its own handwriting business is obtained: the basic data format is complete, the business data information is complete, the data from each source is compatible, the handwriting data is marked, and the handwriting data is preliminarily classified and graded. Among them, the data collection module aggregates the data sources on demand, which can specifically include the "standardization" principle, that is, for the unified collection standards of the data basic format, business data format, and source identification format that have been set, the incomplete data items in each source data are reasonably repaired and supplemented, the data items that meet the standards are confirmed and identified, and the additional data items that exceed the standards are uniformly stored and recorded, and the additional data items are included in the unified collection standards after they meet the universality, commonality and commonality of the data statistics level.

[0037] Among them, the preliminary qualified handwriting data must at least include: basic handwriting data information (also called "full handwriting information"), additional handwriting data information, data collection business information, etc.

[0038] The data collection module pre-processes the original data of personal biometric information, personal identity information, and collection device information collected from different sources, and unifies them into a full set of handwriting information in a standard format. The handwriting algorithm model extracts features, calculates, learns, and adjusts parameters of the handwriting data to generate a personal handwriting model that can uniquely represent the individual's handwriting writing habits; the personal handwriting model is used to compare and identify the handwriting of unknown writers, and all associated data of the algorithm model generate comparison model data, which serves as data governance process data for intermediate data in the process of status identification of handwriting data.

[0039] like Figure 5 The following is a diagram of the data storage module architecture in an exemplary embodiment of the application. The data storage module architecture encrypts and stores the data involved in all processes in the database. The data storage module provides data encryption services and data version control services for the full handwriting information, personal model data, comparison model data, and data governance process data. Among them, the full handwriting information consists of handwriting core information and handwriting basic information: the handwriting core information is the unique representation of each writing behavior of the user, and its content includes but is not limited to the handwriting unique identifier, user unique identifier, and identity authentication identifier; and the handwriting basic information is the original information set of the digital handwriting sequence itself, and its content includes but is not limited to basic point information, device information, user information, business information, etc.

[0040] Personal model data is the sum of data that can uniquely represent an individual's handwriting writing habits after the full amount of handwriting information has been processed through the standard operating procedures for handwriting basic data processing, and then extracted, calculated, learned, and adjusted by handwriting algorithms, including but not limited to deep learning algorithms and traditional sequence algorithms. Its content includes but is not limited to encrypted personal handwriting sets, personal handwriting feature information and model information, personal handwriting recognition history information, and personal model style group status information.

[0041] The comparison model data is the sum of all associated data of the algorithm model called by the handwriting recognition service when using the personal handwriting model to compare and identify the handwriting of an unknown writer. The comparison model has the characteristics of continuously improving and expanding the comparison and identification capabilities as the full amount of handwriting information is accumulated. Its model can include but is not limited to Transformer, RNN, LSTM, CNN, etc. It should be noted that when the handwriting recognition service is called in the handwriting sub-library, the personal handwriting model in the sub-library is used, and when the handwriting recognition service is called in the handwriting main library, the personal handwriting model in the main library is used.

[0042] Data governance process data is not directly used for personal model building and comparison model construction, but is only used to identify the status of handwriting data and can be used for handwriting function services. It is the sum of intermediate data, including but not limited to suspected qualified data, algorithm pre-preparation data, identity authentication shared tags and other data.

[0043] At the same time, the data storage module is responsible for ensuring the basic services of encryption management and data version management of data storage and data storage during the storage process in accordance with the basic requirements for the data life cycle management of biometric information in GB / T 43697-2024 "Data Security Technology Data Classification and Grading Rules" and the customized needs of the handwriting library builder.

[0044] The algorithm model determines the personal handwriting model according to the handwriting modeling algorithm objectives and algorithm implementation path, specifically including: the algorithm model selects and calls the corresponding network model, determines the structure, number of layers and sequential construction of hidden layers including convolutional layers, pooling layers and fully connected layers, and calls activation function design and loss function design to construct the personal handwriting model.

[0045] In the exemplary embodiment of the present application, the functional services include but are not limited to: data services, basic handwriting services, algorithm synchronization services, handwriting application services and other service sub-modules. The data service consists of data statistics related services, "main library-sub-library" data synchronization services and "main library-backup library" data backup update services.

[0046] Data service is the "flow" of handwriting database data, and basic handwriting service is the foundation of handwriting application. The combination of basic handwriting service and business scenarios can build a variety of handwriting application services. Therefore, basic handwriting service can include but not limited to handwriting collection service, handwriting modeling service, handwriting recognition service and personal model management function. Algorithm synchronization service is the guarantee of continuous growth of handwriting database data efficiency. It will identify and control the corresponding model data version, algorithm version, personal model version, etc. respectively to ensure the full life cycle management of each version of data. Its specific service content may include but is not limited to personal style group addition and update, sample modeling algorithm synchronization, recognition and comparison algorithm synchronization and personal model synchronization. Handwriting application service is a combination of business scenarios and basic handwriting services. It can be combined on demand according to the target customer groups of each handwriting database main library. Its practice may include but is not limited to handwriting recognition login, handwriting regional encryption, handwriting identification assistance, etc.

Claims

1. A system for generating an electronic signature handwriting database, characterized in that: It includes the handwriting database main database, handwriting database sub-database, and authoritative backup database. The handwriting data sub-database includes a handwriting data sub-database matrix established according to different business system data sources. The handwriting database sub-database type is determined according to different service module calling methods. The handwriting data main database adopts different data synchronization mechanisms according to different types of sub-databases to synchronize handwriting data and related information. The handwriting database main database and the authoritative backup database exchange handwriting data and related information through data backup and update.

2. The system according to claim 1, characterized in that The data collection module standardizes all handwriting data obtained from the main database and stores them in the specified format; the algorithm module builds personal handwriting models in the main database and sub-database according to different business system standards.

3. The system according to claim 1, characterized in that The method of determining the handwriting database sub-library type according to different calling methods and adopting different data synchronization mechanisms includes: when the function service module calling methods all use the Internet public network for docking, the sub-library without deployed function service modules is a "Type A" sub-library, and its data synchronization service method adopts real-time synchronization; when the function module service calling method is that the algorithm model is deployed in the main library, and the remaining function modules are deployed in the sub-library, it is a "Type B" sub-library, and its data synchronization service method adopts periodic synchronization; when the function service module calling method is that all function service module calls are deployed in the sub-library intranet environment, it is a "Type C" sub-library, and its data synchronization service adopts on-demand synchronization.

4. The system according to claim 2, characterized in that The construction of the personal handwriting model includes: the algorithm model calls the corresponding network model according to the business sub-node standard, determines the structure, number of layers and sequential construction of hidden layers including convolution layer, pooling layer and full connection layer, and calls the activation function design and loss function design to construct the personal handwriting model in the sub-library.

5. The system according to claim 2, characterized in that The data collection module pre-processes multiple types of data and then collects them uniformly, that is, it repairs and supplements the incomplete data items obtained by each business system, identifies the data items that meet the standards, stores and records the data items that do not meet the standards, and performs normalized data processing; The personal biometric information, personal identity information, and collection device information are unified into a full set of handwriting information in a standard format; the handwriting feature data are repaired, cleaned, data quality checked, abnormal data marked, and data classified and graded to obtain basic handwriting data.

6. The system according to claim 2, characterized in that The handwriting algorithm model learns handwriting style features from handwriting data, adjusts algorithm model parameters, and generates a personal handwriting model that represents personal handwriting writing habits; the personal handwriting model calls the algorithm model to associate the basic handwriting data to form comparison data for handwriting comparison and recognition, and performs status identification as data governance process data.

7. The system according to any one of claims 2 to 6, characterized in that: The real-time synchronization includes: when the sub-library requests a service call from the main library, it synchronizes the requested service-related data including handwriting basic data, personal model data, and business information data to the main library in real time. When the main library responds to the service request, it also synchronizes the corresponding service data to the sub-library in real time; the periodic synchronization is to synchronize the handwriting basic data required by the algorithm model only when the algorithm model is called; the on-demand synchronization is to synchronize the sub-library algorithm and its handwriting data with the main library when the sub-library enables the algorithm synchronization service.

8. The system according to any one of claims 1 to 6, characterized in that: A data update backup service is adopted between the main library and the backup library. When the backup library only provides backup services, the main library regularly synchronizes the handwriting feature data of the latest personal handwriting model data to the backup library. When the backup library not only provides backup services but also provides handwriting recognition services, the main library also regularly synchronizes the latest handwriting recognition algorithm with the backup library, and the backup library regularly flows back recognition request data.

9. A method for generating an electronic signature handwriting database, characterized in that: It includes establishing a handwriting data sub-library matrix according to different business system data sources and building different types of handwriting data sub-libraries based on different functional service calling methods; determining the handwriting database sub-library types and different data synchronization mechanisms between the handwriting data main library and different types of sub-libraries according to different business needs; and exchanging handwriting data information between the handwriting database main library and the authoritative backup library through data backup and update.

10. The method according to claim 9, characterized in that The data collection module standardizes the handwriting data of all sub-databases obtained from the main database and stores them in the prescribed format; the algorithm model calls the corresponding network model according to the business sub-node standard, determines the structure, number of layers and sequential construction of hidden layers including convolutional layer, pooling layer and fully connected layer, and calls the activation function design and loss function design to build a personal handwriting model in the sub-database.

11. The method according to claim 10, characterized in that The handwriting algorithm model learns handwriting style features from handwriting data, adjusts algorithm model parameters, and generates a personal handwriting model that represents personal handwriting writing habits; the personal handwriting model calls the algorithm model to associate the basic handwriting data to form comparison data for handwriting comparison and recognition, and performs status identification as data governance process data.

12. The method according to any one of claims 9 to 11, characterized in that: When the function service module calling mode uses the Internet public network for docking, the sub-library without deployed function service modules is a "Type A" sub-library, and its data synchronization service mode adopts real-time synchronization; when the function service module calling mode is that the algorithm model is deployed in the main library, and the other function modules are deployed in the sub-library, it is a "Type B" sub-library, and its data synchronization service mode adopts periodic synchronization; when the function service module calling mode is that all function service modules are deployed in the intranet environment of the sub-library corresponding to the business system, it is a "Type C" sub-library, and its data synchronization service adopts on-demand synchronization; the real-time synchronization includes: when the sub-library requests a service call from the main library, it synchronizes the requested service related data including handwriting basic data, personal model data, and business information data to the main library in real time, and when the main library responds to the service request, it also synchronizes the corresponding service data to the sub-library in real time; the periodic synchronization is to synchronize the required handwriting basic data only when the algorithm model is called; the on-demand synchronization is to synchronize the sub-library algorithm and its handwriting data with the main library when the sub-library enables the algorithm synchronization service.

Images