Method, system, device and medium for recognizing signing intention based on handwriting features

Abstract

The application discloses a method for recognizing signing intention based on handwriting features, and the signatures in the active sample library and the passive sample library are mutually exclusive; online acquisition of user's handwritten electronic signature is compared and verified with the corresponding signature in the active intention sample library, if the verification is consistent, the active intention setting is executed, if the verification is inconsistent, further comparison and verification with the signature in the passive intention sample library is performed, if the verification is consistent, the passive intention setting is executed, the pre-warning is started, if the number of inconsistent verification exceeds the predetermined number, the signing is locked. The two kinds of intention samples are consistent in static features such as writing method, but the dynamic features are different, and the handwriting verification algorithm is used for verification, which is mutually exclusive. It is difficult to be imitated, and is a kind of private and safe passive sample method. It is beneficial to protect the personal safety and property safety of the person under duress.

Description

Technical Field

[0001] This invention relates to the field of information technology, and more specifically, to a method for recognizing signing intent based on handwriting features. Background Technology

[0002] With the development of informatization and datafication, the signature medium has gradually shifted from traditional paper to electronic devices such as handwriting tablets and electronic screens. In addition to traditional pen-and-paper signatures, electronic signatures, stored electronically, have also emerged. The promulgation of the "Electronic Signature Law of the People's Republic of China" legalized electronic signatures, clarifying that reliable electronic signatures have the same legal effect as traditional signatures and seals. Electronic signatures are now widely used in various fields such as government affairs, healthcare, banking and finance, fast-moving consumer goods, real estate, and logistics. However, how can the reliability and authenticity of electronic signatures be guaranteed? Generally, during relevant business processes, users sign electronic documents. At this time, the business system assumes that the user has signed voluntarily, assuming that the operation represents the user's true intention, without considering the user's environment, personal safety, or whether the user signed voluntarily or under duress. This leads to the signing of some electronic documents without the user's voluntary consent, reducing the authenticity, legality, and reliability of electronic signatures.

[0003] Chinese invention patent application CN110348466A, entitled "Method and Apparatus for Identity Recognition," discloses a method and apparatus for identity recognition. The method involves acquiring the current passive features of a current user, matching these features with preset passive features in a feature library, and determining the matching degree of the current passive features. When the matching degree is not less than a first threshold, the current user's identity is recognized based on the current passive features; when the matching degree is less than the first threshold, the current user's identity is recognized based on the current user's active features. This identity recognition method employs a fusion of active verification and passive features, and performs interval processing based on the matching degree of the current passive features, thereby improving the accuracy, security, and user experience of identity recognition.

[0004] Passive features are user characteristics that can be obtained without requiring active user interaction, and are used to improve the accuracy of identity authentication.

[0005] Chinese invention patent application CN103440447A, entitled "Online Handwriting Authentication Method with Attacker Identification Capability," describes a method for identifying the true identity of a user during the online handwriting authentication phase of an authentication system. If the user fails authentication, the method compares the physiological characteristics related to the writing hand in the test handwriting feature information submitted by the user with the physiological characteristics related to the writing hand in the peripheral device's physiological characteristic information database. This method is resistant to replay attacks; attackers cannot pre-create test handwriting feature information; and it has the ability to identify attackers. The method compares the separated handwriting information and the physiological characteristics related to the writing hand with the corresponding registration template information in the authentication system's self-built handwriting feature information database. Based on the comparison result, it determines whether the user has passed authentication. If authentication fails, and the character recognition algorithm indicates that the handwriting matches the presented registered standard characters, the method compares the physiological characteristics related to the writing hand in the test handwriting feature information submitted by the user with the physiological characteristics related to the writing hand in the peripheral device's physiological characteristic information database to identify the true identity of the user.

[0006] The aforementioned existing technologies primarily target signer authentication, focusing on optimizing algorithms or verification strategies to improve accuracy, prevent replay attacks, and mitigate repeated copying attacks. Besides identity authentication, signature recognition (authentication) also functions as a form of intent authentication, a key distinction between signature recognition and other biometric features. None of the aforementioned literature addresses the issue of signature intent, or assumes that signatures that pass authentication are written with voluntary intent. However, some signatures are produced under duress, not as an expression of the signer's voluntary will; these are called passive intent signatures. For example, a criminal suspect might threaten a signer with a knife, demanding they sign to pass the verification system or face threats to their safety. In such cases, the signer desires to pass the system without endangering their own safety, while also hoping to be identified as having passive intent, allowing the system to respond with appropriate settings based on this passive intent, such as protecting property, implementing covert alarm measures, or preserving evidence. Existing technologies do not address intent authentication. Summary of the Invention

[0007] This application employs a novel handwriting recognition scheme that enables the identification of the signatory's intent, especially passive intent, in order to protect the personal and property safety of coerced signatories.

[0008] In view of this, this application proposes a method for recognizing signing intent based on handwriting features, including: verifying the user's identity through information such as ID card and verification code; sending a prompt for active intent sampling; collecting multiple electronic signatures of the user and storing them in a sampling library to construct an active intent signature sampling library; sending a prompt for passive intent sampling; collecting multiple electronic signatures of the user and performing mutual exclusion determination with the corresponding electronic signatures of the user in the active intent sampling library; storing electronic signatures that meet the mutual exclusion requirement in the passive intent sampling library; obtaining the user's handwritten electronic signature online; comparing and verifying it with the corresponding signature in the active intent sampling library; if the verification matches, executing the active intent setting; if the verification does not match, further comparing and verifying it with the signature in the passive intent sampling library; if the verification matches, executing the passive intent setting and initiating a setting warning; if the number of verification discrepancies exceeds a predetermined number, locking the signature.

[0009] Further preferably, the electronic signatures satisfy the mutual exclusion requirement as follows: the electronic signatures in the active intention sample library and the electronic signatures in the passive intention sample library have the same static characteristics but significantly different dynamic characteristics, and cannot be matched by any handwriting comparison algorithm; any sample in the active intention sample library cannot be matched by a sample in the passive intention sample library by a handwriting comparison algorithm.

[0010] Further optimization involves the following scenarios: if the online electronic signature and the signature in the active intention sample database pass the consistency verification, the system will trigger an active intention response and log in to the user account through signature verification; if the online electronic signature and the signature in the passive intention sample database pass the consistency verification, it will be determined as a passive intention signature, triggering a passive intention response in the system. The front end will indicate that the verification has passed, while the back end will refuse to log in to the user account or restrict transactions, save transaction records, and send alarm information.

[0011] Further optimization involves using an attention-weighted dynamic time warping method to compare and verify online electronic signatures with signatures in the sample database. The weights of dynamic features of the signature data are set higher than those of static features, and the weights of the start and end stroke features are set higher than those of the stroke features.

[0012] Further optimization involves employing a dynamic time warping algorithm combined with dynamic signature features to align the features of the two compared signature points during signature mutual exclusion determination and comparison verification. In the attention-weighted dynamic time warping method, the difference between the dynamic signature sequence features predicted by the mask pre-trained model and the obtained online signature features is used as the dynamic feature sequence weight. and Call the formula:

[0013] Calculate the quantization difference d between the two signatures. If the difference is greater than a threshold, the two signatures are considered inconsistent; if the difference is less than the threshold, the two signatures are considered consistent. and Let represent the dynamic feature sequences of signature A and signature B, respectively, where n is the length of the aligned dynamic feature sequences. and Let A and B represent the weights of the dynamic feature sequences of signature A and signature B at point i, respectively.

[0014] Further optimization involves calculating a verification distance threshold based on multiple signatures in the active intention signature sample library. The maximum or average distance between all signatures with the same content from the same signer in the active intention signature sample library is used as the verification distance threshold. The mutual exclusion determination threshold between active intention sampling and passive intention sampling is greater than or equal to the threshold for signature comparison and verification in the verification stage.

[0015] According to another aspect of this application, the present invention also proposes a system for recognizing signing intent based on handwriting features, comprising: a signature acquisition unit, a signature comparison unit, a verification unit, and an execution unit. The signature acquisition unit is used to acquire electronic signatures signed multiple times by the user under both active and passive intent prompts after the user has passed authentication, and store these signatures in a sample library to construct an active intent signature sample library and a passive intent sample library; it also acquires the user's handwritten electronic signature online. The signature comparison unit is used to determine the mutual exclusion of the electronic signature acquired under passive intent prompts with the corresponding user's electronic signature in the active intent sample library, and stores the electronic signatures that meet the mutual exclusion requirement in the passive intent sample library. The verification unit compares and verifies the user's handwritten electronic signature acquired online with the corresponding electronic signatures in both the active and passive intent sample libraries. The execution unit executes the active intent setting if the verified online electronic signature matches the signature in the active intent sample library, and passes user signature authentication; if it matches the signature in the passive intent sample library, it executes the passive intent setting and initiates a setting warning; if the number of inconsistent verifications exceeds a predetermined number, the signature acquisition unit is locked.

[0016] Further optimization involves the signature comparison unit and verification unit employing a dynamic time warping algorithm combined with dynamic signature features to align the features of the two compared signature points during signature mutual exclusion determination and verification. An attention-weighted dynamic time warping method is used, with the difference between the signature dynamic sequence features predicted by the mask pre-training model and the acquired user handwritten signature features serving as the dynamic feature sequence weight. and Call the formula:

[0017] Calculate the quantization difference d between the two signatures. If the difference is greater than a threshold, the two signatures are considered inconsistent; if the difference is less than the threshold, the two signatures are considered consistent. and These represent the dynamic feature sequences of signature A and signature B, respectively, where n is the length of the aligned dynamic feature sequence. The signature comparison unit calculates the verification distance threshold based on multiple signatures in the active intention signature retention library. It uses the maximum or average distance between signatures with the same content from the same signer in the active intention signature retention library as the threshold. The mutual exclusion determination threshold between active intention retention and passive intention retention is greater than or equal to the signature comparison and verification threshold in the verification stage.

[0018] According to another aspect of this application, an electronic device is proposed, comprising: a processor; and a memory storing a program, wherein the program includes instructions that, when executed by the processor, cause the processor to perform the method for recognizing signature intent based on handwriting features as described above.

[0019] According to another aspect of this application, a non-transitory computer-readable storage medium storing computer instructions is proposed, wherein the computer instructions are used to cause the computer to perform the method for identifying signature intent based on handwriting features as described above.

[0020] This application involves at least two types of intention sampling, and the two types of intention sampling are verified to be mutually exclusive through handwriting verification algorithms. It proposes that the two types of sample signatures should be as consistent as possible in static features such as handwriting style, but differ in dynamic features. Such passive intention sampling is less likely to be detected by feature differences and is difficult to imitate, making it a privacy-preserving and secure passive sampling method.

[0021] Existing time-series alignment-based methods lack weighting for point features, leading to the averaging of differences between active and passive signatures. This application employs an attention-weighted method to calculate the differences between two signatures, giving higher weight to unconventional stroke features. This is more beneficial for the algorithm to identify personalized dynamic features in the signature that distinguish it from ordinary signatures, resulting in higher accuracy and better suitability for personalized features in unconventional passive signatures. During the attention weight calculation, the difference between the common dynamic features and the actual dynamic features of the signature point features reflects the degree of personalization. A signature feature recognition method based on a masked pre-trained model can be used to predict common dynamic features corresponding to the signature style. Because it has been trained on a large amount of data, its output smooths out personalized features, predicting common dynamic features corresponding to the signature style.

[0022] Existing verification systems typically only include active intention verification and response processes. This application also includes passive intention verification and passive intention response processes. The passive intention response process is configurable and includes features such as withdrawal limits, intention storage, and default alarms, which are distinct from active intention response processes and protect the rights of signatories. Attached Figure Description

[0023] Further details, features, and advantages of this application are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which:

[0024] Figure 1 The diagram shown is a schematic representation of the construction of a user sample library according to an exemplary embodiment of this application.

[0025] Figure 2 The diagram shown is a schematic representation of a user sample retention intention verification process according to an exemplary embodiment of this application.

[0026] Figure 3 The diagram shown is a structural block diagram of an exemplary electronic device that can be used to implement embodiments of this application;

[0027] Figure 4 The diagram shown is a schematic representation of the DTW alignment result according to an exemplary embodiment of this application. Detailed Implementation

[0028] Embodiments of this application will now be described in more detail with reference to the accompanying drawings. While some embodiments of this application are shown in the drawings, it should be understood that this application can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this application. It should be understood that the drawings and embodiments of this application are for illustrative purposes only and are not intended to limit the scope of protection of this application.

[0029] It should be understood that the steps described in the method embodiments of this application may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of this application is not limited in this respect.

[0030] The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Definitions of other terms will be given in the following description. It should be noted that the concepts of "first", "second", etc., mentioned in this application are used only to distinguish different devices, modules, or units, and are not intended to limit the order of functions performed by these devices, modules, or units or their interdependencies.

[0031] It should be noted that the terms "a" and "a plurality of" used in this application are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0032] The names of the messages or information exchanged between multiple devices in the embodiments of this application are for illustrative purposes only and are not intended to limit the scope of these messages or information.

[0033] Personal handwriting signatures are distinctive; once writing habits are formed, the external appearance of the handwriting becomes fixed. A user's signature biometrics, including strokes, stroke order, pen pressure, and pen speed, remain relatively stable under conscious control. Pen movement characteristics are also reflected in five aspects: starting stroke characteristics, stroke movement characteristics, ending stroke characteristics, pen force characteristics, and basic stroke shape characteristics. Even differences in the starting, moving, and ending strokes will affect the basic stroke shape characteristics. Pen movement is the most delicate and complex writing action, therefore, its characteristics are highly specific, stable, and valuable for authentication.

[0034] The implementation of the present invention will be described in detail below with reference to the accompanying drawings and specific examples.

[0035] like Figure 1 The diagram illustrates the process of building a user signature database. It includes a user signature database with active intent and a passive intent database. Active intent sampling prompts a signature, performs a consistency check, and if the signature satisfies the requirement, it is stored in the active intent database. Passive intent sampling prompts a signature, performs a consistency check, and checks for mutual exclusion between the passive and active intent signatures. If the mutual exclusion requirement is met, the passive intent signature is stored in the passive intent database.

[0036] This system for recognizing signing intent based on handwriting features includes: a signature acquisition unit, a signature comparison unit, a verification unit, and an execution unit. The signature acquisition unit collects electronic signatures signed multiple times by the user under prompts for active and passive intent signature retention and stores them in a retention library, constructing both an active intent signature retention library and a passive intent signature retention library. It also collects user electronic signatures and obtains handwritten electronic signatures online. The signature comparison unit performs a mutual exclusion check between the electronic signature obtained under passive intent signature retention and the corresponding user's electronic signature in the active intent signature retention library, storing the mutually exclusive electronic signatures in the passive intent signature retention library. The verification unit compares the user's handwritten electronic signatures obtained online with the corresponding electronic signatures in both the active and passive intent signature retention libraries. If the verified online electronic signature matches the signature in the active intent signature retention library, the active intent setting is executed, and the execution unit authenticates the signature. If the signature matches the signature in the passive intent signature retention library, the passive intent setting is executed, and the execution unit initiates a setting warning. If the number of inconsistent verifications exceeds a predetermined number, the signature acquisition unit is locked.

[0037] In a trusted environment (such as when handling business in person using an ID card), the system acquires and verifies the user's identity information. After verifying the user's true identity, it collects and samples the user's electronic signature data. The system sends an active intention sampling prompt, collects multiple electronic signatures from the user on the same signing device, and stores them in the active sampling database. The system also sends a passive intention sampling prompt, collects multiple electronic signatures from the user on the same signing device, and performs a mutual exclusion check between the electronic signatures obtained under the passive intention sampling prompt and the corresponding electronic signatures in the active sampling database. Electronic signatures that meet the mutual exclusion requirement are stored in the passive sampling database.

[0038] When voluntarily retaining a signature, the system's handwriting retention specifications can be followed. It can be a signature that has been retained multiple times (e.g., 3 times) and can be determined using a 1:1 signature authentication algorithm. The voluntarily signed electronic signature is stored in the voluntary retention database.

[0039] When passively sampling signatures, the requirements for active signature handwriting sampling should be followed. The signatures should be mutually exclusive with the active signature, and the two types of sampled signatures should be as consistent as possible in static features such as handwriting style, while having differences in dynamic features. Such passively sampled signatures are less likely to be detected and harder to imitate. Passive signatures are stored in a passive sample database.

[0040] In the creation of the passive intention signature database, the signature sample cannot be matched by any sample in the active intention signature database using a handwriting comparison algorithm. Passive intention samples and active intention samples can be very similar in form, but have significant differences in dynamic features such as pressure and stroke order. This way, passive intention signature samples can be correctly distinguished by the algorithm and are not easily detected by others.

[0041] The active and passive signature databases are mutually exclusive. A user's signature sample must belong to either the active or passive database; that is, no sample from one database can be matched with a sample from the other using a handwriting comparison algorithm. The mutual exclusivity of the active and passive signature databases can be determined using a handwriting comparison algorithm. Specifically, any combination of signatures from the active and passive databases must not be matched. If a match is found, the databases are not mutually exclusive, and the passive signature needs to be resampled. This process continues until a signature that fails to match a signature in the active signature database is used as a passive signature sample.

[0042] Configure system responses for different intentions. Determine whether the electronic signature data represents an active or passive intention, and trigger different responses based on the intention to which the signature belongs. For active intentions, the system response can be set to reflect the identity of the signer or the signer's active intention.

[0043] The system collects the user's electronic signature and compares it with electronic signature samples in the database. If it is determined to be a voluntary signature, the system confirms that the electronic signature represents the user's true intention, triggering a voluntary response. After signature authentication, the system initiates the next step of operation. This applies to scenarios such as bank transfers, withdrawals, contract signings, and online transactions, allowing the system to proceed to the next step, initiating the transfer or completing the transaction.

[0044] Passive consent signing occurs when a signature is made under duress, rather than as a voluntary expression of intent. For example, a criminal suspect might threaten a signatory with a knife, demanding they sign to pass a verification system, threatening their safety otherwise. In such cases, the signatory wants both to receive a verification notification to avoid personal danger and to protect their assets. This signing could be identified as passive consent. The system should indicate successful verification at the front end, but the back end should not proceed with any substantive transaction processing, but should retain relevant evidence and implement password-based alarm measures, such as issuing alerts.

[0045] The system can respond to relevant settings based on passive intentions, such as refusing transfers, secretly alerting the authorities, preserving evidence, or even displaying false user information.

[0046] In online transaction scenarios, the system can set a single transaction limit of 10,000 yuan and an unlimited number of transactions per day if the transaction is initiated by the user. If the transaction is initiated by the user, the system can set a single transaction limit of 1,000 yuan and a maximum of 3 transactions per day. The system will also store the transaction information and send an alert to the system administrator for future use as evidence in case of an emergency.

[0047] This method greatly protects users who sign under duress, and provides reliable evidence for subsequent loss recovery and judicial proceedings.

[0048] Figure 2 The diagram illustrates the user's intention verification process. The system prompts the user to write, and the user writes according to the prompt. Based on the signature handwriting verification algorithm, the system performs an active intention verification by comparing the user's signature with the signatures in the active intention database. If the consistency check passes, the active intention setting and response are executed. If the consistency check fails, the system performs a passive intention verification by comparing the signature with the signatures in the passive intention database. If the consistency check passes, the passive intention setting and response are executed. If the consistency check fails, the user is prompted to sign again, and a consistency check is performed again. An alert is also set, and the signature is locked.

[0049] During the system verification phase, the signatory's electronic signature is first compared with the electronic signatures in the active intent sample database using a handwriting comparison algorithm to verify whether the electronic signature represents the true intention of the party concerned. If so, the system executes the active intent setting; otherwise, it proceeds to passive intent verification.

[0050] If the signatory's electronic signature fails the active intent verification, a passive intent verification is performed. This involves comparing the signature with electronic signatures in the passive intent sample database using a handwriting comparison algorithm to verify whether the signature was signed under duress. If so, the passive intent setting is executed. If not, it is determined that the signature does not match either the active or passive sample databases, and further processes such as rewriting and locking the signature are initiated.

[0051] If a user is not used to the writing device or their writing style has changed, they can choose to rewrite the document. If multiple signature attempts fail, it is suspected that someone else may be forging the signature. Depending on the settings, appropriate actions may be taken, such as prompting for a sample signature or locking the signature verification function.

[0052] Signature handwriting comparison or signature handwriting verification methods are used in the user signature verification process, as well as in the signature stability and mutual exclusion verification during the sample retention process.

[0053] This embodiment proposes to leverage the characteristics of both active and passive intention signature data, paying greater attention to distinguishable dynamic features. An attention-weighted method is used to calculate and compare the differences between the two signatures, assigning higher weights to dynamic features such as pen movement characteristics (e.g., pen pressure, pen speed, pen sequence) and lower weights to static features such as stroke shape, character spacing, and structure. This approach facilitates the algorithm's identification of personalized dynamic features in signatures that differentiate them from ordinary individuals, resulting in higher accuracy. It is particularly beneficial for recognizing the personalized features of unconventional passive signature writing. This effectively improves the accuracy of the active and passive intention signature matching algorithm. Specifically, it can be implemented using the following method:

[0054] Generally, personal handwriting signatures are distinctive; once writing habits are formed, the external presentation of the handwriting becomes fixed. A user's signature biometrics, including strokes, stroke order, pen pressure, and pen speed, remain relatively stable under conscious control. Pen movement is the most delicate and complex writing action, therefore its characteristics are highly specific, stable, and valuable for identification. However, not all pen movement characteristics vary significantly from person to person. For example, writing any stroke involves three aspects: the beginning, the middle, and the end. The differences in the beginning and end of strokes are usually greater than those in the middle. The DTW (Dynamic Time Warping) algorithm typically treats all points with the same weight when calculating differences, neglecting more discriminative features.

[0055] The active intention signature sampling and passive intention signature sampling involved in the exemplary embodiments of this application can be consistent at many feature points, and more attention needs to be paid to the differences in feature points.

[0056] Before performing signature comparison and verification and selecting samples from the retention library, the collected electronic signatures are preprocessed, including standard signature preprocessing such as uniform sampling rate and removal of interfering strokes. Assuming preprocessed signatures A (active intent signature retention library) and B (passive intent signature retention library), standardized signature dynamic features such as signature stroke speed and acceleration are extracted. A dynamic time normalization algorithm is used in conjunction with the signature dynamic features to align the features of the two signature points. and Let A and B represent the dynamic feature sequences of signature A and signature B, respectively, where n is the total length of the aligned dynamic feature sequences. The formula is:

[0057] Calculate the quantization difference d between signature B and signature A.

[0058] Where d represents the quantitative difference between signature B and signature A, the smaller d is, the more similar the signatures are. and Let represent the weights of signature A and signature B in the dynamic feature sequence at point i, respectively. The larger the dynamic weight at a point, the more the difference at that point is amplified. and Choose a larger value to represent the weight of that point.

[0059] The difference between two signatures is calculated using the attention-weighted dynamic time warping (DTW) method. If the difference is greater than a threshold, the two signatures are considered inconsistent; if the difference is less than the threshold, the two signatures are considered consistent.

[0060] The verification distance threshold is calculated based on multiple signatures in the proactive intention signature sample library. This can be achieved by using all signatures with the same content from the same signer in the proactive intention signature sample library, and calculating the maximum or average distance between the signatures using the formula described above. The threshold for determining the stability and consistency of signatures in the proactive intention sample library is greater than or equal to the threshold for determining the mutual exclusion between proactive and passive intention samples, and is greater than or equal to the threshold for signature verification during the verification phase.

[0061] The above method can pay more attention to the impact of regions with significant feature changes on signature handwriting. By adopting the attention-weighted dynamic feature difference method, it can significantly improve the algorithm for correctly distinguishing between active intention sampling and passive intention sampling.

[0062] To obtain signature features, a signature feature recognition method based on a mask pre-trained model is adopted. The difference between the dynamic sequence features of the signature predicted by the mask pre-trained model after signature feature pre-training and the size of the actual signature features is used as the weight of the dynamic feature sequence. and The calculation can be based on the L2 distance.

[0063] This application employs a pre-training method. Due to extensive training with large amounts of data, its output smooths out individual characteristics, predicting the general dynamic characteristics corresponding to the signature style. The magnitude of the difference reflects the degree of individualization. This method is more effective in identifying dynamic characteristics in signatures that distinguish them from ordinary people, resulting in higher algorithm accuracy and greater benefit for the individualization characteristics of unconventional passive signature styles.

[0064] An exemplary embodiment of this application also provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor. The memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to cause the electronic device to perform a method according to an embodiment of this application.

[0065] An exemplary embodiment of this application also provides a non-transitory computer-readable storage medium storing a computer program, wherein the computer program, when executed by a computer's processor, is used to cause the computer to perform a method according to an embodiment of this application.

[0066] An exemplary embodiment of this application also provides a computer program product, including a computer program, wherein, when executed by a computer's processor, the computer program is used to cause the computer to perform a method according to an embodiment of this application.

[0067] refer to Figure 3 The present invention describes a structural block diagram of an electronic device 300 that can serve as a server or client of this application, which is an example of a hardware device that can be applied to various aspects of this application. The electronic device is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the application described and / or claimed herein.

[0068] like Figure 3As shown, the electronic device 300 includes a computing unit 301, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 302 or a computer program loaded from a storage unit 308 into a random access memory (RAM) 303. The RAM 303 may also store various programs and data required for the operation of the device 300. The computing unit 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0069] Multiple components in electronic device 300 are connected to I / O interface 305, including: input unit 306, output unit 307, storage unit 308, and communication unit 309. Input unit 306 can be any type of device capable of inputting information to electronic device 300. Input unit 306 can receive input digital or character information and generate key signal inputs related to user settings and / or function control of electronic device. Output unit 307 can be any type of device capable of presenting information and may include, but is not limited to, a display, speaker, video / audio output terminal, vibrator, and / or printer. Storage unit 308 may include, but is not limited to, disk and optical disk. Communication unit 309 allows electronic device 300 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks, and may include, but is not limited to, modems, network cards, infrared communication devices, wireless communication transceivers, and / or chipsets, such as Bluetooth devices, WiFi devices, WiMax devices, cellular communication devices, and / or the like.

[0070] The computing unit 301 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 301 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 301 performs the various methods and processes described above. For example, in some embodiments, the reconstruction and decomposition of the muscle movement trajectory of the signature stroke based on its original trajectory, and the decomposition of its logarithmic velocity curve, can be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as storage unit 308. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 300 via ROM 302 and / or communication unit 309. In some embodiments, the computing unit 301 can be configured by any other suitable means (e.g., by means of firmware) to perform a signature handwriting dynamic acquisition implementation method.

[0071] The program code used to implement the methods of this application may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0072] In the context of this application, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0073] As used in this application, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, device, and / or apparatus (e.g., disk, optical disk, memory, programmable logic device (PLD)) for providing machine instructions and / or data to a programmable processor, including machine-readable media that receive machine instructions as machine-readable signals. The term "machine-readable signal" refers to any signal for providing machine instructions and / or data to a programmable processor.

[0074] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0075] The systems and techniques described herein can be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected to each other by digital data communication in any form or medium (e.g., a communication network). Examples of communication networks include: local area network (LAN), wide area network (WAN), and the Internet.

[0076] A computer system can include clients and servers. The clients and servers are generally far from each other and typically interact through a communication network. The client - server relationship is created by computer programs that run on the respective computers and have a client - server relationship with each other.

[0077] Figure 4 Represents the DTW alignment result. The thicker the dots, the greater the weight. It can be seen that there are more thick dots in the "Jiu" on the left, and the overall personalized characteristics are stronger. The weights of the dots at some details are very obvious, and the dynamic characteristics of such dots will be given greater weights to participate in the difference calculation.

[0078] Although only two types of intentions, active intention and passive intention, are listed, it is equally applicable to scenarios with more than two types of intentions. It is also applicable to the留样库 (the description here might be incorrect or unclear in the original Chinese, but translated as is), and it is not limited to signatures and is equally applicable to symbols in dynamic sequences. It should be noted that the term "留样库" in the original text seems to be an incorrect or unclear expression. If there is more context or a correct term, the translation can be further refined.

Claims

1. A method for recognizing signing intent based on handwriting features, characterized in that, Verify user identity, send a prompt for active intention to retain samples, collect multiple electronic signatures from the user and store them in the sample database, and build an active intention signature sample database; Send a passive intention sampling prompt, collect multiple electronic signatures from the user and determine their mutual exclusion with the corresponding user's electronic signature in the active intention sampling database, and store the electronic signatures that meet the mutual exclusion requirements into the passive intention sampling database; The system retrieves the user's handwritten electronic signature online and compares it with the corresponding signature in the active intention sample library. If the verification matches, the active intention setting is executed. If the verification does not match, the system further compares and verifies it with the signature in the passive intention sample library. If the verification matches, the passive intention setting is executed and a setting warning is activated. If the number of verification discrepancies exceeds a predetermined number, the signature is locked. The electronic signatures must meet the mutual exclusion requirement as follows: the electronic signatures in the active intention sample library and the electronic signatures in the passive intention sample library have the same static characteristics but significantly different dynamic characteristics. They cannot be matched by any handwriting comparison algorithm, and any sample in the active intention sample library cannot be matched by a sample in the passive intention sample library by a handwriting comparison algorithm.

2. The method as described in claim 1, characterized in that, If the online electronic signature and the signature in the active intention sample database are found to be consistent, the system will trigger an active intention response and log in to the user account through signature verification. If the online electronic signature and the signature in the passive intention sample database are found to be consistent, the signature will be determined to be a passive intention signature and the system will trigger a passive intention response. The front end will indicate that the verification has been passed, while the back end will refuse to log in to the user account or restrict transactions, save transaction records, and send alarm information.

3. The method according to any one of claims 1-2, characterized in that, An attention-weighted dynamic time warping method is used to compare and verify online electronic signatures with signatures in the sample database. The weight of dynamic features of signature data is set higher than that of static features of signature data, and the weight of the start and end features is higher than that of the stroke features.

4. The method as described in claim 3, characterized in that, In the signature mutual exclusion determination and comparison verification, a dynamic time warping algorithm is used in conjunction with signature dynamic features to align the features of the two signature points being compared. In the attention-weighted dynamic time warping method, the difference between the signature dynamic sequence features predicted by the mask pre-trained model and the obtained online signature features is used as the dynamic feature sequence weight. and Call the formula: Calculate the quantization difference d between the two signatures. If the difference is greater than a threshold, the two signatures are considered inconsistent; if the difference is less than the threshold, the two signatures are considered consistent. and Let represent the dynamic feature sequences of signature A and signature B, respectively, where n is the length of the aligned dynamic feature sequences. and Let A and B represent the weights of the dynamic feature sequences of signature A and signature B at point i, respectively.

5. The method according to any one of claims 1-2 or 4, characterized in that, The verification distance threshold is calculated based on multiple signatures in the active intention signature sample library. The maximum or average distance between all signatures with the same content from the same signer in the active intention signature sample library is used as the verification distance threshold. The mutual exclusion determination threshold between active intention sampling and passive intention sampling is greater than or equal to the threshold for signature comparison and verification in the verification stage.

6. A system for recognizing signing intent based on handwriting features, characterized in that, include: The system comprises a signature acquisition unit, a signature comparison unit, a verification unit, and an execution unit. The signature acquisition unit is used to collect the electronic signatures signed by the user multiple times under both active and passive intention prompts after the user has passed authentication, and store them in a retention library. This constructs an active intention signature retention library and a passive intention signature retention library with mutually exclusive signature features, and allows for online acquisition of the user's handwritten electronic signature. The signature comparison unit is used to determine the mutual exclusivity of the electronic signatures obtained under passive intention retention prompts with the corresponding user's electronic signatures in the active intention retention library, and stores the electronic signatures that meet the mutual exclusivity requirements in the passive intention retention library. Verification Unit: Compares and verifies the handwritten electronic signatures acquired online with the corresponding electronic signatures in the active intention sample library and the passive intention sample library, respectively. Execution Unit: If the verified online electronic signature matches the signature in the active intention sample library, the active intention setting is executed, and user signature authentication is passed. If it matches the signature in the passive intention sample library, the passive intention setting is executed, and a setting warning is initiated. If the number of inconsistent verifications exceeds a predetermined number, the signature acquisition unit is locked. The electronic signatures must meet the mutual exclusion requirement as follows: the electronic signatures in the active intention sample library and the electronic signatures in the passive intention sample library have the same static characteristics but significantly different dynamic characteristics. They cannot be matched by any handwriting comparison algorithm, and any sample in the active intention sample library cannot be matched by a sample in the passive intention sample library using a handwriting comparison algorithm.

7. The system as described in claim 6, characterized in that, In the signature comparison and verification units, a dynamic time warping algorithm is used in conjunction with dynamic signature features to align the features of the two signature points being compared during signature mutual exclusion determination and verification. An attention-weighted dynamic time warping method is employed, using the difference between the signature dynamic sequence features predicted by the mask pre-trained model and the acquired user handwritten signature features as the dynamic feature sequence weights. and Call the formula: Calculate the quantization difference d between the two signatures. If the difference is greater than a threshold, the two signatures are considered inconsistent; if the difference is less than the threshold, the two signatures are considered consistent. and These represent the dynamic feature sequences of signature A and signature B, respectively, where n is the length of the aligned dynamic feature sequence. The signature comparison unit calculates the verification distance threshold based on multiple signatures in the active intention signature retention library. It uses the maximum or average distance between signatures with the same content from the same signer in the active intention signature retention library as the threshold. The mutual exclusion determination threshold between active intention retention and passive intention retention is greater than or equal to the signature comparison and verification threshold in the verification stage.

8. An electronic device, comprising: processor; And a memory storing a program, wherein the program includes instructions that, when executed by the processor, cause the processor to perform the method for recognizing signature intent based on handwriting features according to any one of claims 1-5.

9. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method for recognizing signature intent based on handwriting features according to any one of claims 1-5.

Images