Dynamic handwritten signature recognition method, device and electronic equipment

Through radar technology, the motion information of the hand and pen is collected, dynamic signature features are extracted and matched with the template, which solves the problems of low sampling rate of handwritten signature recognition on smartphones and high equipment requirements, and achieves high security and portability handwritten signatures Identify.

CN114333075BActive Publication Date: 2024-11-26TSINGHUA UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202111667092.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-12-31
Publication Date
2024-11-26
Estimated Expiration
2041-12-31

AI Technical Summary

Technical Problem

The existing handwritten signature recognition technology has low sampling rate, insufficient hardware processing capabilities on smartphones, and high equipment requirements, making it difficult to meet portability and security needs, especially in mobile application scenarios.

Method used

Radar technology is used to collect movement and attitude information of the hand and pen, collect data through electromagnetic wave or ultrasonic radar, extract dynamic signature features, and match them with template features in the database to achieve identity identification.

Benefits of technology

The security and portability of handwritten signature recognition are improved, difficult to imitate, suitable for mobile applications, and the contactlessness and low-complexity feature extraction algorithm of radar technology make the device compact and easy to use on portable devices.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114333075B_ABST
    Figure CN114333075B_ABST
Patent Text Reader

Abstract

The present application relates to the field of biometric identification technology, and in particular to a method, device and electronic device for dynamic handwritten signature recognition, wherein the method includes: collecting radar detection data carrying the movement and posture information of the user's hand and / or pen during the writing process based on electromagnetic waves or ultrasonic radar; extracting the signature features of the user's tested signature from the radar detection data; matching the signature features of the tested signature with the signature features of the template in the database, and identifying the tested signature by the matched corresponding template. Therefore, radar technology is used as an information collection means and the dynamic features of the writing process are used as recognition features, which has the advantages of being difficult to imitate, natural writing and high device portability.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of biometric identification technology, and in particular to a dynamic handwritten signature recognition method, device and electronic device. Background Art

[0002] Biometric identification uses the unique biological characteristics of the human body to identify the user. Biometric identification uses automatic technology to measure the physiological or behavioral characteristics of a person, and compares these characteristics (samples) with the template data in the database to complete the authentication solution. Biometric identification is divided into two types: identity authentication and identity recognition. Identity authentication is to compare the sample with a specific template 1:1. If they match each other, the identity authentication is passed, otherwise it will not pass; identity recognition is to compare the sample with all templates in the database 1:N, find the template with the highest degree of matching, and then identify the sample object as the user to which the template belongs. Biometric identification technology has advantages that traditional cryptography does not have, such as easy to use, non-repudiation, difficult to forge, and no need to remember passwords. In recent years, the application of computer-based automated biometric identification technology has become increasingly widespread. Technologies such as face recognition and fingerprint recognition have been rapidly popularized in mobile payment, security inspection and other fields.

[0003] Handwritten signature is also a biometric feature that can be used for identity authentication. It can reflect the long-term unique writing behavior characteristics of the signer. It has the characteristics of convenience, easy acceptance, difficult to forge, and unique personality. From ancient times to the present, it has been widely used as an effective means of identity authentication in legal documents, contract signing, authorization management and other identity confirmation occasions. It is one of the few legally effective authentication methods. Compared with other authentication features, it has the advantages of non-invasiveness and simple equipment, so it has high acceptance. In addition, handwriting recognition is renewable. The number of physiological features such as fingerprints and faces is limited for each person. Once converted into digital information and entered into the computer system, there is a possibility of being cracked and stolen. Once the template is obtained by hackers, it will cause irreparable losses. In contrast, different handwriting feature templates can be obtained by changing the written content, and the leakage of one template will not affect others. Written text is much easier to remember than digital passwords. In addition to checking the static image features of the signature, the "dynamic signature" also measures the time information and pen information of the writing process in real time, which is a safer authentication method.

[0004] Handwritten signature authentication is divided into two types: static (offline) signature and dynamic (online) signature. The latter is more secure because it collects more information and is difficult to imitate. The input devices for dynamic signatures include tablets, input pens, touch screens, etc. Most of the research for a long time has been based on these special devices. They sample the signature trajectory to form a series of sampling points. In addition to the coordinates of each sampling point, they can also provide information such as the pen tip pressure, the pen tilt angle (pitch angle and azimuth) or acceleration corresponding to each point.

[0005] In recent years, online signature systems based on mobile terminals have emerged. For example, in the related art, an online handwritten signature authentication system for smartphones is designed based on the Android system. Corresponding solutions are proposed for the problems of low sampling rate of smartphones, only the position and pressure information of the pen tip (or fingertip) can be obtained, and the hardware system processing power is relatively low. At present, the touch-screen handwritten signature system for smartphones has been commercially applied in logistics, banking, electronic payment and other fields.

[0006] Related technologies also propose using Kinect (a 3D body-sensing camera) to collect virtual signatures written in the air by fingers, which can be used for identity authentication when entering a building. However, this method has high requirements for equipment, large user movements, and poor concealment, so it has not been popularized.

[0007] In terms of recognition algorithms, the features commonly used for identification include the number of strokes, the number of pauses, the difference value of signature segmentation, the aspect ratio of the signature, the tilt angle, etc. The commonly used feature compression and extraction transformations include Discrete Wavelet Transformation (DWT) and Discrete Fourier Transform (DFT). A new division method is proposed in the related art, and a neuro-fuzzy system based on fuzzy set theory is used in the classification process. In recent years, research has mainly focused on signature authentication systems for mobile devices, and the application of neural network algorithms has also increased. For example, the application of neural network technology has improved the reliability of handwritten signature authentication systems on mobile devices, and its measurement method is a multi-level pressure sensor. Someone has designed a text-dependent online handwritten signature authentication system for mobile phone unlocking applications. It uses angles, velocity spectra and Fourier descriptors as features, and uses Manhattan distance as the matching distance for template matching to meet the limitations of mobile devices on computational complexity.

[0008] Application Contents

[0009] The present application provides a dynamic handwritten signature recognition method, device and electronic device, which adopt radar technology as an information collection means and use the dynamic characteristics of the writing process as recognition features, and have the advantages of being difficult to imitate, natural writing and high device portability.

[0010] The first embodiment of the present application provides a dynamic handwritten signature recognition method, comprising the following steps:

[0011] Radar detection data collected based on electromagnetic waves or ultrasonic radars that carry information about the movement and posture of the user's hand and / or pen during the writing process;

[0012] Extracting the signature feature of the user's tested signature from the radar detection data; and

[0013] The signature feature of the signature under test is matched with the signature feature of the template in the database, and the signature under test is identified by the matched corresponding template.

[0014] Optionally, when there are multiple matching corresponding templates, the method further includes:

[0015] Calculating the matching degree between each matching corresponding template and the signature under test;

[0016] The template with the highest matching degree is used as the best matching template, so as to determine the signature to be tested based on the best matching template.

[0017] Optionally, the radar detection data collected based on electromagnetic waves or ultrasonic radar and carrying the movement and posture information of the user's hand and / or pen during the writing process includes:

[0018] Demodulate the received radar echo signal to obtain a baseband signal;

[0019] After filtering out clutter of stationary objects in the environment, processing and transforming the filtered baseband signal to generate moving target detection data;

[0020] A detection data sequence that varies with time is generated according to a preset sampling rate to generate signature data.

[0021] Optionally, extracting the signature feature of the user's tested signature from the radar detection data further includes:

[0022] Firstly, primary feature extraction is performed on the signature data to obtain primary features;

[0023] The primary features are concatenated and sent to a preset feature extractor to obtain a plurality of feature matrix sequences;

[0024] One or more signature features are obtained based on the multiple feature matrix sequences.

[0025] Optionally, matching the signature feature of the tested signature with the signature feature of a template in a database includes:

[0026] The scalar matching degree is determined according to the feature vector of the signature feature of the tested signature and the feature vector value of the signature feature of the template; and the template with the matching degree greater than a preset threshold is used as the corresponding template of the match.

[0027] The second aspect of the present application provides a dynamic handwritten signature recognition device, including:

[0028] An acquisition module, for collecting radar detection data carrying movement and posture information of a user's hand and / or pen during a writing process based on electromagnetic waves or ultrasonic radar;

[0029] an extraction module, configured to extract signature features of the user's signature under test from the radar detection data; and

[0030] The identification module is used to match the signature feature of the signature under test with the signature feature of the template in the database, and identify the signature under test by the matched corresponding template.

[0031] Optionally, when there are multiple matching corresponding templates, the identification module is further used to:

[0032] Calculating the matching degree between each matching corresponding template and the signature under test;

[0033] The template with the highest matching degree is used as the best matching template, so as to determine the signature to be tested based on the best matching template.

[0034] Optionally, the extraction module is specifically used to:

[0035] Demodulate the received radar echo signal to obtain a baseband signal;

[0036] After filtering out clutter of stationary objects in the environment, processing and transforming the filtered baseband signal to generate moving target detection data;

[0037] A detection data sequence that varies with time is generated according to a preset sampling rate to generate signature data.

[0038] Optionally, the extraction module is further used to:

[0039] Firstly, primary feature extraction is performed on the signature data to obtain primary features;

[0040] The primary features are concatenated and sent to a preset feature extractor to obtain a plurality of feature matrix sequences;

[0041] One or more signature features are obtained based on the multiple feature matrix sequences.

[0042] Optionally, the identification module is specifically used to:

[0043] Determine a scalar matching degree according to a feature vector of the signature feature of the tested signature and a feature vector value of the signature feature of the template;

[0044] The template whose matching degree is greater than a preset threshold is used as the matching corresponding template.

[0045] The third aspect of the present application provides an electronic device, including: a radar transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program, and the radar transceiver detects and collects echo signals of the hand and / or pen during the movement of the user's signature to be measured, so as to implement the dynamic handwritten signature recognition method as described in the above embodiment.

[0046] Therefore, the dynamic handwritten signature recognition method of the embodiment of the present application has the following advantages:

[0047] (1) High security: Radar mainly uses the dynamic characteristics of the target. The radar echo signal can reflect the speed and position of the hand and pen. It can also collect the micro-movement information of the fingers and hand muscles and the shape characteristics of the hand, which greatly increases the difficulty of counterfeiting handwriting. In other words, radar measurement is a safe, convenient, humanized and highly applicable way to capture signature behavior characteristics;

[0048] (2) Free writing and suitable for mobile applications: Since it adopts a non-contact detection method, it has almost no requirements for the writing interface and writing tools, and can achieve a natural writing state, with stronger adaptability and better user experience. In addition, it does not require a writing board, so the device can be very small and does not require an additional input pen, making the device portable.

[0049] (3) Strong applicability: Compared with other possible contactless handwritten signature recognition methods, such as camera-based machine vision technology, radar solutions are more applicable. Although machine vision systems equipped with depth cameras can also capture and extract static and motion information of the target and have high resolution, first, it is difficult to obtain information about the obscured parts; second, it is sensitive to lighting conditions, and environmental factors such as the reflectivity of the writing surface, lighting, shadows, and writing distance will affect it; third, extracting the target's motion information from the video requires a more complex processing algorithm, which is limited for use on portable device platforms; fourth, the signature collection process has the risk of leaking the content of the signed document. In contrast, radar technology uses electromagnetic waves with a wavelength of several millimeters, which have a stronger detour ability than visible light and are not restricted by conditions such as lighting. Moreover, since radar directly obtains information such as the distance, speed, and shape of the target, the corresponding feature extraction algorithm is less complex than the machine vision algorithm. The monolithic integrated radar is compact and can be used in portable devices.

[0050] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0052] Figure 1 A flow chart of a dynamic handwritten signature recognition method provided according to an embodiment of the present application;

[0053] Figure 2 is an exemplary diagram of a radar-based dynamic handwritten signature recognition system according to an embodiment of the present application;

[0054] Figure 3 This is an example diagram of a signal acquisition radar configuration according to an embodiment of the present application;

[0055] Figure 4 This is an example diagram of a dynamic handwritten signature recognition device according to an embodiment of the present application;

[0056] Figure 5 is an exemplary diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0057] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0058] The following describes the dynamic handwritten signature recognition method, device and electronic device according to the embodiments of the present application with reference to the accompanying drawings.

[0059] Before introducing the dynamic handwritten signature recognition method according to the embodiment of the present application, a dynamic signature measurement device in the related art is briefly introduced.

[0060] In the related art, dynamic signature measurement equipment mainly includes handwriting tablets, input pens equipped with sensors, touch screens, etc. The handwriting pens need to be equipped separately, and the inertial measurement unit based on them has low accuracy and is cumbersome to use. Both handwriting tablets and touch screens require users to write on the panel provided by the device, and the size of the panel is inconsistent with the comfort of writing. Due to the device's restrictions on writing speed and force, the interface touch, and the writing method are different from the normal state, it is easy to cause user movement deformation, and the error rate is high when compared with signatures obtained by other methods, and the user comfort is low.

[0061] Therefore, the methods in the related art have disadvantages such as poor portability or poor user experience, and cannot adapt to today's convenient and user-friendly application needs. The present application provides a dynamic handwritten signature recognition method, which uses a radar device to record the movement of the hand and pen during the handwriting process, and extracts features for identity recognition. The radar actively emits detection waves (such as electromagnetic waves or ultrasonic waves) to illuminate the target, and estimates various information about the properties of the target from parameters such as the flight time of the reflected wave, signal strength attenuation, and frequency shift.

[0062] Specifically, Figure 1 A flow chart of a dynamic handwritten signature recognition method provided in an embodiment of the present application.

[0063] In this embodiment, the system involved in the dynamic handwritten signature recognition method of the embodiment of the present application includes: a signal acquisition radar, a sample signature feature extraction module, a template signature feature extraction module, a matching module and a template library. Among them, the signal acquisition radar can be an ultrasonic radar or a millimeter wave or terahertz electromagnetic wave radar, which is composed of one or more transmitters and multiple receivers. The transmitter continuously transmits modulated detection waves to the detection space. If there are multiple transmitters, the detection waves emitted by each transmitter are different and should have different frequencies or orthogonal sequence modulations. The receiver continuously receives the echo of the target. Each receiver should be oriented toward the detection space and placed at different angles. The signal acquisition radar collects the motion and posture signals of the hand (and pen) (called the target) through active detection during the user's writing process. The sample signature feature extraction module extracts features that can be used to distinguish the user's identity from the data collected by the signal acquisition radar module and generates a digital feature vector. The matching module compares the sample feature vector generated by the sample signature feature extraction module with the template signature feature vector and outputs a quantitative value of the matching degree. In the identity authentication application, whether the verification is passed is determined based on whether the matching degree exceeds the preset threshold. In the identity recognition application, the best matching template is selected based on the size of the matching degree values ​​corresponding to different templates.

[0064] like Figure 1 As shown, the dynamic handwritten signature recognition method includes the following steps:

[0065] In step S101, radar detection data carrying movement and posture information of a user's hand and / or pen during a writing process is collected based on electromagnetic waves or ultrasonic radar.

[0066] It should be understood that, in conjunction with Figure 2It can be seen that the embodiments of the present application can collect detection data carrying the movement and posture information of the hand (and pen) (referred to as the target) during the user's writing process by active detection. Optionally, in some embodiments, radar detection data carrying the movement and posture information of the hand and / or pen during the user's writing process is collected based on electromagnetic waves or ultrasonic radar, including: demodulating the received radar echo signal to obtain a baseband signal; after filtering out the clutter of stationary objects in the environment, processing and transforming the filtered baseband signal to generate a moving target echo signal; generating a detection data sequence that varies with time according to a preset sampling rate, and generating signature data.

[0067] It should be understood that the modulation method of the radar's transmitting wave, that is, the radar system, can be but is not limited to a coherent pulse train, a frequency-modulated continuous wave, or a direct sequence spread spectrum phase modulation, and the receiver adopts a corresponding demodulation method to obtain a baseband signal; each receiver demodulates the received echo signal to obtain a baseband signal, and then processes and transforms the baseband signal after filtering out the clutter of stationary objects in the environment to generate a moving target echo signal, and then generates a series of time-varying sequences of range-Doppler images, short-time Fourier transform spectra, phases, and other data at a certain sampling rate, and outputs them as detection data to a sample signature feature extraction module or a template signature feature extraction module.

[0068] In step S102, the signature features of the user's signature under test are extracted from the radar detection data.

[0069] Optionally, in some embodiments, signature features of the user's tested signature are extracted from radar detection data, including: first performing primary feature extraction on the signature data to obtain primary features; concatenating the primary features and sending them into a preset feature extractor to obtain multiple feature matrix sequences; and obtaining one or more signature features based on the multiple feature matrix sequences.

[0070] It can be understood that the sample signature feature extraction module first performs primary feature extraction on the received raw detection data, and then splices the obtained primary features into the feature extractor, which is a neural network. One implementation method is a LSTM (Long short-term memory) convolutional neural network, and the output of the neural network is a set of feature vectors (high-level features). The primary feature extraction step, on the one hand, extracts several features about the sequence from the time-varying sequence of each item of the raw detection data of each receiver, and on the other hand, fuses the raw data of multiple receiver channels to generate multiple cross-channel features. The two types of features are spliced ​​to form multiple feature matrix sequences as the input of the feature extractor.

[0071] The template signature feature extraction module first performs primary feature extraction on the received raw data, and then splices the obtained primary features into the template feature extractor. The template signature feature extraction module has the same primary feature extraction algorithm as the sample signature feature module, but its template feature extractor can have primary feature data inputs of multiple samples, and fuse them to generate a template feature vector (template). The high-level features generated by the sample signature feature extraction module have the same dimension as the feature vector corresponding to the template, and the meanings of the features on the corresponding dimensions correspond one to one. The template feature extractor is also a neural network. One implementation method is a first-level LSTM convolutional neural network with the same structure as the sample feature extractor, connected to a first-level feature fusion module, which receives outputs from multiple sample feature extractor networks and then fuses them to generate a high-level feature vector with their common characteristics. These multiple samples should come from multiple signatures of the same signature content of the same user. As a special case, the template can also be generated by only one sample.

[0072] In step S103, the signature feature of the signature under test is matched with the signature feature of the template in the database, and the signature under test is identified by the matched corresponding template.

[0073] Optionally, in some embodiments, the signature features of the signature under test are matched with the signature features of the template in the database, including: determining a scalar matching degree based on a feature vector of the signature features of the signature under test and a feature vector value of the signature features of the template; and taking a template having a matching degree greater than a preset threshold as a corresponding template for matching.

[0074] Optionally, in some embodiments, when there are multiple matching corresponding templates, it also includes: calculating the matching degree between each matching corresponding template and the signature to be tested; taking the template with the highest matching degree as the best matching template, so as to determine the signature to be tested based on the best matching template.

[0075] It should be understood that the matching module has two inputs, one is the high-level feature vector output by the sample signature feature extraction module, and the other is the template extracted from the template library. The matching module outputs a scalar value reflecting the degree of matching between the two feature vectors. For example, the embodiment of the present application can calculate the Euclidean distance between the signature feature of the signature under test and the signature feature of the template. The smaller the distance, the higher the degree of matching between the two.

[0076] The template library is a database storing signature templates of one or more users, which is stored inside the device or can be accessed remotely by the device online. There are one or more records in the library, and each record contains at least a template and a user ID corresponding to the template.

[0077] Based on the above, it can be seen that the system involved in the dynamic handwritten signature recognition method of the embodiment of the present application has three working modes:

[0078] (1) Template registration mode: This mode is used to register the user's signature information. The device collects the user's signature once or multiple times, then generates a template through the template signature feature extraction module, and stores the template in the template library;

[0079] (2) Identification mode: This mode is used to complete the identity authentication or identification of the user being tested. The user writes a signature under the device, and the signal acquisition radar of the device collects the signature process detection data. Then, the sample signature feature vector is generated through the sample signature feature extraction module. Then, the matching module is used to match the sample signature feature vector with one or more templates to be compared, and the matching result is output;

[0080] (3) Training mode: The device, which is identical to the practical device, uses a signal acquisition radar to collect multiple signature process signals of different signature contents of multiple users, and uses the primary feature extraction unit in the sample signature feature extraction module to extract the primary features of each sample as a training data set. The features of these samples are divided into recognition samples and registration samples. The registration samples are used as training samples for the feature extractor network and feature fusion network in the template signature feature extraction module, and the recognition samples are used as training samples for the feature extractor network in the sample signature feature extraction module. The outputs of the template signature feature extraction module and the sample signature feature extraction module are input to the matching module. After the machine learning step, the neural networks in the template feature extractor and the sample feature extractor, as well as the matching module, are trained and generated.

[0081] In order to facilitate those skilled in the art to further understand the dynamic handwritten signature recognition method of the embodiment of the present application, a detailed description is given below in conjunction with a specific embodiment.

[0082] Specifically, the embodiment of the present application uses a 300kHz ultrasonic transceiver as the transceiver of the radar signal. The radar signal system is a coherent pulse train, and the signal collection radar is equipped with a transmitter and two receivers with separate swing amplitudes, which are placed in the following position relationship: Figure 2 As shown, Figure 3 The medium signal acquisition radar is a one-transmit, two-receive device.

[0083] After each receiver receives the echo signal, it obtains the baseband signal through I / Q demodulation and moving target detection (MTD) filtering, and then processes the baseband signal according to each N cpiEach repetition period is a coherent processing interval (CPI), and the corresponding range-Doppler image (Range-Dopple Matrix, RDM) is generated by processing each CPI. The method of generating RDM is to first form a sequence of data at corresponding time points (fast time sampling moments) in each repetition period, and the sampling interval of the sequence is the repetition period, which is called a slow time series; then Fourier transform the slow time series to obtain its spectrum. The spectrum of each fast time sampling moment is arranged in order to form a two-dimensional matrix, which is RDM. The fast time dimension is the distance dimension, and the frequency dimension is the Doppler (speed) dimension. For the complex baseband signal obtained by I / Q demodulation, calculate its average value within a repetition period, and then calculate the phase angle of the average value vector, and form a vector of the phase angle within a CPI. Each receiver is used as a channel, and the sequence of RDM and phase angle vectors of each channel is output as raw data to the sample signature feature extraction module or the template signature feature extraction module.

[0084] The entire signal acquisition radar is implemented by hardware. The template signature feature extraction module is implemented in the computer. The sample signature feature extraction module is implemented by hardware and together with the radar hardware, it constitutes the hardware part of signature recognition.

[0085] The sample signature feature extraction module first generates a micro-Doppler sequence and a one-dimensional range image sequence from the RDM sequence. The micro-Doppler sequence is an amplitude-frequency vector sequence obtained by integrating each RDM along the range dimension. The one-dimensional range image sequence is an amplitude-time vector sequence obtained by integrating each RDM along the Doppler dimension. The micro-Doppler sequence, the one-dimensional range image sequence, and the phase angle vector sequence are respectively regarded as the data of a channel. The data of the two receiver channels are input into an LSTM neural network. The LSTM neural network is a feature extractor that outputs a set of high-level feature vectors of length M.

[0086] The template signature feature extraction module uses the same primary feature extraction module and the same LSTM network as the sample signature feature extraction module to generate an M-dimensional high-level feature vector. In the template registration mode, the signal acquisition radar is used to collect signature registration samples of multiple users, and then the high-level feature vectors extracted from each registration sample are averaged and fused to obtain an M-dimensional average high-level feature vector, which is used as a template and stored in the template library in the computer, which records the user's information and the corresponding template.

[0087] When performing user identification or authentication, the user writes a signature and uses this device to collect signals. The sample signature feature extraction module generates a sample feature vector and sends it to the matching module. The matching module extracts the template to be compared from the template at the same time, and then calculates the Euclidean distance between the two feature vectors as the matching degree. The decision is made based on the set distance threshold. If the relationship value output by the matching module is greater than the distance threshold, the authentication is rejected, otherwise it is passed. In the recognition application, the user of the template with the smallest distance is selected as the recognition result.

[0088] The LSTM network in the feature extractor of the sample signature feature extraction module and the template signature feature extraction module can use the same weight parameters. The training method of the neural network is: use the hardware of the device to collect multiple signature process signals of different signature contents of multiple users, and use the primary feature extraction unit in the sample signature feature extraction module to extract the primary features of each sample as a training data set, and divide the features of these samples into recognition samples and registration samples. The registration samples are used as training samples for the feature extractor network and feature fusion network in the template signature feature extraction module, and the recognition samples are used as training samples for the feature extractor network in the sample signature feature extraction module. The outputs of the template signature feature extraction module and the sample signature feature extraction module are input to the matching module to train the network. The optimization goal in the training is to make the output of the same signature matching module for the same user less than the threshold, and the output of the different signature matching modules for different users or the same user should be greater than the threshold. After the machine learning step, the decision threshold of the network weight and the matching degree is obtained.

[0089] According to the dynamic handwritten signature recognition method proposed in the embodiment of the present application, the radar detection data carrying the movement and posture information of the user's hand and / or pen during the writing process can be collected by electromagnetic waves or ultrasonic radars, and the signature features of the user's tested signature can be extracted from the radar detection data, and the signature features of the tested signature can be matched with the signature features of the template in the database, and the tested signature can be recognized by the matched corresponding template. Therefore, radar technology is used as an information collection means and the dynamic features of the writing process are used as recognition features, which solves the problems of high requirements for the use scenarios and low security of handwritten signature recognition technology in related technologies, and has the advantages of being difficult to imitate, natural writing, and high portability of the device.

[0090] Next, the dynamic handwritten signature recognition device proposed according to the embodiment of the present application is described with reference to the accompanying drawings.

[0091] Figure 4 It is a block diagram of a dynamic handwritten signature recognition device according to an embodiment of the present application.

[0092] like Figure 4As shown, the dynamic handwritten signature recognition device 10 includes: an acquisition module 100 , an extraction module 200 and a recognition module 300 .

[0093] The acquisition module 100 is used to acquire radar detection data carrying the movement and posture information of the user's hand and / or pen during the writing process, which is collected based on electromagnetic waves or ultrasonic radars;

[0094] The extraction module 200 is used to extract the signature features of the user's tested signature from the radar detection data; and

[0095] The identification module 300 is used to match the signature feature of the signature under test with the signature feature of the template in the database, and identify the signature under test by the matched corresponding template.

[0096] Optionally, when there are multiple matching corresponding templates, the identification module 300 is further used to:

[0097] Calculate the matching degree between each matching corresponding template and the signature under test;

[0098] The template with the highest matching degree is used as the best matching template, so as to determine the signature to be tested based on the best matching template.

[0099] Optionally, the acquisition module 100 is specifically used for:

[0100] Demodulate the received radar echo signal to obtain a baseband signal;

[0101] After filtering out the clutter of stationary objects in the environment, the filtered baseband signal is processed and transformed to generate the moving target echo signal;

[0102] A detection data sequence that varies with time is generated according to a preset sampling rate to generate signature data.

[0103] Optionally, the extraction module 200 is used to:

[0104] Firstly perform primary feature extraction on the signature data to obtain primary features;

[0105] The primary features are concatenated and sent to a preset feature extractor to obtain multiple feature matrix sequences;

[0106] One or more signature features are obtained based on multiple feature matrix sequences.

[0107] Optionally, the identification module 300 is specifically used for:

[0108] Determine the scalar matching degree according to the feature vector of the signature feature of the tested signature and the feature vector value of the signature feature of the template;

[0109] The templates with matching degree greater than the preset threshold are taken as the matching corresponding templates.

[0110] It should be noted that the above explanations of the embodiment of the dynamic handwritten signature recognition method are also applicable to the dynamic handwritten signature recognition device of this embodiment, and will not be repeated here.

[0111] According to the dynamic handwritten signature recognition device proposed in the embodiment of the present application, the radar detection data carrying the movement and posture information of the user's hand and / or pen during the writing process can be collected based on electromagnetic waves or ultrasonic radars, and the signature features of the user's tested signature can be extracted from the radar detection data, and the signature features of the tested signature can be matched with the signature features of the template in the database, and the tested signature can be recognized by the matched corresponding template. Therefore, radar technology is used as an information collection means and the dynamic features of the writing process are used as recognition features, which solves the problems of high requirements for the use scenarios and low security of handwritten signature recognition technology in related technologies, and has the advantages of being difficult to imitate, natural writing, and high portability of the device.

[0112] Figure 5 The electronic device may include a radar transceiver 500 , a memory 501 , a processor 502 , and a computer program stored in the memory 501 and executable on the processor 502 .

[0113] The processor 502 executes the program, and the radar transceiver 500 detects and collects the echo signal of the user's hand and / or pen during the movement of the signature to be measured, so as to implement the dynamic handwritten signature recognition method provided in the above embodiment.

[0114] Furthermore, the electronic device further comprises:

[0115] The communication interface 503 is used for communication between the memory 501 and the processor 502 .

[0116] The memory 501 is used to store computer programs that can be executed on the processor 502 .

[0117] The memory 501 may include high-speed RAM and ROM memory, and may also include non-volatile memory (non-volatile memory), such as at least one disk memory.

[0118] If the memory 501, the processor 502 and the communication interface 503 are implemented independently, the communication interface 503, the memory 501 and the processor 502 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 5 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0119] Optionally, in a specific implementation, if the radar transceiver 500, the memory 501, the processor 502 and the communication interface 503 are integrated on a chip, the radar transceiver 500, the memory 501, the processor 502 and the communication interface 503 can communicate with each other through an internal interface.

[0120] The processor 502 may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0121] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0122] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0123] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0124] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or N wirings (electronic devices), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0125] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiment, the N steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0126] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0127] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0128] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A dynamic handwritten signature recognition method, characterized in that: The following steps are involved: Based on radar detection data collected by electromagnetic waves or ultrasonic radars and carrying information on the movement and posture of the user's hand and pen during the writing process, the received radar echo signal is demodulated to obtain a baseband signal. After filtering out clutter of stationary objects in the environment, the filtered baseband signal is processed and transformed to generate moving target detection data, and a detection data sequence that varies with time is generated according to a preset sampling rate, and signature data is generated; wherein, according to the generation of the detection data sequence that varies with time according to the preset sampling rate, generating the signature data includes: generating a micro-Doppler sequence and a one-dimensional range image sequence from an RDM sequence, the micro-Doppler sequence is an amplitude-frequency vector sequence obtained by integrating each RDM along the range dimension, and the one-dimensional range image sequence is an amplitude-time vector sequence obtained by integrating each RDM along the Doppler dimension; The posture information is the micro-movement information of the fingers and hand muscles and the hand shape characteristics; Ultrasonic radar consists of multiple transmitters and multiple receivers; Extracting the signature feature of the user's tested signature from the radar detection data; The method of extracting the signature features of the user's tested signature from the radar detection data includes: firstly performing primary feature extraction on the signature data to obtain primary features; splicing the primary features and sending them to a preset feature extractor to obtain multiple feature matrix sequences, and obtaining one or more signature features based on the multiple feature matrix sequences, wherein the feature extractor is a neural network, and the primary feature extraction includes: extracting a number of features about the sequence from the time-varying sequence of each item of the original detection data of each receiver, and fusing the original data of multiple receiver channels to generate multiple cross-channel features, and splicing the two types of features to form multiple feature matrix sequences as inputs of the feature extractor; And, the signature feature of the signature under test is matched with the signature feature of the template in the database, and the signature under test is identified by the matched corresponding template.

2. The method according to claim 1, characterized in that When there are multiple matching corresponding templates, it also includes: Calculating the matching degree between each matching corresponding template and the signature under test; The template with the highest matching degree is used as the best matching template, so as to determine the signature to be tested based on the best matching template.

3. The method according to any one of claims 1 to 2, characterized in that: The matching of the signature feature of the signature under test with the signature feature of the template in the database includes: Determine a scalar matching degree according to a feature vector of the signature feature of the tested signature and a feature vector value of the signature feature of the template; The template whose matching degree is greater than a preset threshold is used as the matching corresponding template.

4. A dynamic handwritten signature recognition device, characterized in that: The dynamic handwritten signature recognition device is used to execute the method described in any one of claims 1 to 3, comprising: An acquisition module is used to collect radar detection data carrying the movement and posture information of the user's hand pen during the writing process based on electromagnetic waves or ultrasonic radar; an extraction module, configured to extract signature features of the user's signature under test from the radar detection data; and The identification module is used to match the signature feature of the signature under test with the signature feature of the template in the database, and identify the signature under test by the matched corresponding template.

5. The device according to claim 4, characterized in that When there are multiple matching corresponding templates, the identification module is further used to: Calculating the matching degree between each matching corresponding template and the signature under test; The template with the highest matching degree is used as the best matching template, so as to determine the signature to be tested based on the best matching template.

6. The device according to claim 4, characterized in that The acquisition module is specifically used for: Demodulate the received radar echo signal to obtain a baseband signal; After filtering out clutter of stationary objects in the environment, processing and transforming the filtered baseband signal to generate moving target detection data; A detection data sequence that varies with time is generated according to a preset sampling rate to generate signature data.

7. The device according to claim 6, characterized in that The extraction module is specifically used for: Firstly, primary feature extraction is performed on the signature data to obtain primary features; The primary features are concatenated and sent to a preset feature extractor to obtain a plurality of feature matrix sequences; One or more signature features are obtained based on the multiple feature matrix sequences.

8. The device according to any one of claims 4 to 7, characterized in that: The identification module is specifically used for: Determine a scalar matching degree according to a feature vector of the signature feature of the tested signature and a feature vector value of the signature feature of the template; The template whose matching degree is greater than a preset threshold is used as the matching corresponding template.

9. An electronic device, characterized in that: include: A radar transceiver, a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the program, and the radar transceiver detects and collects the echo signal of the hand and / or pen during the movement of the user's signature to be measured, so as to implement the dynamic handwritten signature recognition method as described in any one of claims 1 to 3.