Online signature authentication method and device, computer equipment and storage medium
By using millimeter-wave radar technology and fast Fourier transform algorithm, the time-speed characteristic data of user signatures is extracted, and the problem of online handwritten signature authentication depends on specific devices is solved, and the accuracy and reliability of authentication is improved.
Patent Information
- Application Number
- CN202310585269.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-05-23
- Publication Date
- 2024-11-26
AI Technical Summary
The existing online handwriting signature authentication scheme relies on specific handwriting equipment, and the sound-based online handwriting signature authentication is not very accurate in dynamic information recognition due to physical limitations such as frequency, resulting in low authentication accuracy.
Millimeter wave radar is used to send signals and receive echo signals, and time-speed feature data is extracted through filtering, denoising and fast Fourier transform algorithms, and the trained recognition model is used for authentication, so as to achieve the use of specific handwriting equipment without using specific handwriting equipment. Improve the accuracy of signature authentication.
The time-speed characteristic data during user signature is extracted through millimeter wave radar technology, which improves the accuracy of online handwritten signature authentication, avoids dependence on specific devices, and enhances the reliability of authentication.
Smart Images

Figure CN119025905A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an online signature authentication method, device, computer equipment and storage medium. Background Art
[0002] Since online handwritten signature authentication can obtain dynamic information during the user's signature process, it is more reliable than offline handwritten signature authentication. However, the current online handwritten signature authentication scheme relies on specific handwriting devices (such as a specific tablet or pen). In addition, sound-based online handwritten signature authentication has received increasing attention due to its universality and non-intrusiveness. However, due to physical limitations such as the frequency of sound signals, the scheme has low accuracy in dynamic information recognition, resulting in low accuracy.
[0003] How to improve the accuracy of online handwritten signature authentication without using a specific handwriting device is a problem that needs to be solved urgently in the prior art. Summary of the invention
[0004] To solve the problems in the prior art, the embodiments of this specification provide an online signature authentication method, apparatus, computer device and storage medium, which improve the authentication accuracy on the basis of performing online handwritten signature authentication without using a specific handwriting device.
[0005] In order to solve the above technical problems, the specific technical solutions of this specification are as follows:
[0006] On the one hand, the embodiments of this specification provide an online signature authentication method, including:
[0007] During the user signing process, receiving echo signals corresponding to a plurality of transmission signals of the millimeter wave radar;
[0008] Filtering and denoising the transmission signal and the echo signal to obtain a plurality of target sub-intermediate frequency signals;
[0009] Processing the plurality of target sub-intermediate frequency signals using a fast Fourier transform algorithm to obtain time-speed characteristic data; and
[0010] The trained recognition model is used to process the time and speed feature data to obtain a target authentication result corresponding to the user.
[0011] Further, the filtering and denoising of the transmission signal and the echo signal to obtain a plurality of target sub-intermediate frequency signals further includes:
[0012] Performing mixing and low-pass filtering on the transmission signal and the echo signal to obtain an intermediate frequency signal;
[0013] Performing signal interference denoising processing on the intermediate frequency signal to obtain a first intermediate frequency signal; and
[0014] Dynamic interference denoising is performed on the first intermediate frequency signal corresponding to each of the transmitted signals to obtain the multiple target sub-intermediate frequency signals.
[0015] Furthermore, the signal interference denoising process is performed on the intermediate frequency signal to obtain a first intermediate frequency signal, further comprising:
[0016]
[0017] in, Characterizes the i-th first intermediate frequency signal, represents the ith intermediate frequency signal, M represents the number of intermediate frequency signals corresponding to each of the transmitted signals, and Characterizes the mth intermediate frequency signal.
[0018] Furthermore, the step of performing dynamic interference denoising on the first intermediate frequency signal corresponding to each of the transmitted signals to obtain the multiple target sub-intermediate frequency signals further includes:
[0019]
[0020] Among them, S i,k characterizes the kth target sub-intermediate frequency signal corresponding to the i-th first intermediate frequency signal, represents the nth sampled first intermediate frequency signal in the i-th first intermediate frequency signal, N represents the number of sampling points of the first intermediate frequency signal, δ represents the unit impulse function, K represents the number of the target sub-intermediate frequency signals corresponding to the i-th first intermediate frequency signal, and L represents the length of the target sub-intermediate frequency signal, which is determined by the number of sampling points and the number of the target sub-intermediate frequency signals.
[0021] Further, the fast Fourier transform algorithm includes a distance dimension fast Fourier transform algorithm, a Doppler dimension fast Fourier transform algorithm and a summary characteristic formula, and the fast Fourier transform algorithm is used to process the multiple target sub-intermediate frequency signals to obtain the time speed characteristic data further includes:
[0022] Using the distance-dimensional fast Fourier transform algorithm to process each of the target sub-intermediate frequency signals to obtain multiple candidate signals;
[0023] Selecting a target signal to be processed from the multiple candidate signals based on the peak values of the multiple candidate signals corresponding to each of the target sub-IF signals;
[0024] Using the Doppler fast Fourier transform algorithm to process the target signal to be processed, to obtain range Doppler time series data; and
[0025] The range Doppler time series data is processed using the summary characteristic formula to obtain the time velocity characteristic data.
[0026] Furthermore, the selecting the target signal to be processed from the multiple candidate signals based on the peak values of the multiple candidate signals corresponding to each of the target sub-IF signals further includes:
[0027] For each of the target sub-intermediate frequency signals, determining a range gate corresponding to the peak value;
[0028] For a first intermediate frequency signal corresponding to a plurality of target sub-intermediate frequency signals, a plurality of range gates corresponding to the plurality of target sub-intermediate frequency signals are processed by a voting method to obtain a target range gate corresponding to the first intermediate frequency signal;
[0029] For the transmission signals corresponding to the plurality of the first intermediate frequency signals, determining a target area based on a preset window value and the target range gate; and
[0030] The candidate signals corresponding to each target gate in the target area are all used as the target signals to be processed,
[0031] Each of the transmission signals corresponds to a plurality of the first intermediate frequency signals, and each of the first intermediate frequency signals corresponds to a plurality of target sub-intermediate frequency signals.
[0032] Furthermore, the target area further includes,
[0033]
[0034] in, Characterize the target area, p min Characterizes the smallest target range gate among the target range gates, p max represents the largest target range gate among the target range gates, W represents the preset window value, N represents a positive integer, and a represents the target gate.
[0035] Furthermore, the summary feature formula further includes,
[0036]
[0037] Among them, V (f,i) Characterizing the time-speed characteristic data, RDM (f,i,j) represents the corresponding value of the target gate j corresponding to the Doppler gate s in the f-th range Doppler time series data and the target signal to be processed, Rj represents the index of the target gate, N represents the first number of points corresponding to the distance dimension fast Fourier transform algorithm, and M represents the second number of points corresponding to the Doppler dimension fast Fourier transform algorithm.
[0038] Furthermore, the training process of the trained recognition model includes:
[0039] Acquire a meta-trained recognition model and sample time-speed feature data corresponding to the user;
[0040] Processing the sample time speed feature data using the meta-trained recognition model to obtain a sample recognition result;
[0041] Based on the difference between the sample recognition result and the accurate label, the meta-trained recognition model is trained to obtain the trained feature extraction model,
[0042] The meta-trained recognition model is obtained by training with a pre-training sample set, and the sample time-speed feature data is determined by a plurality of sample transmission signals and a sample echo signal corresponding to each transmission signal.
[0043] On the other hand, the embodiment of this specification also provides an online signature authentication device, including:
[0044] A receiving unit, used for receiving echo signals corresponding to multiple transmission signals of the millimeter wave radar during the user signing process;
[0045] A first processing unit is used to filter and denoise the transmission signal and the echo signal to obtain a plurality of target sub-intermediate frequency signals;
[0046] A second processing unit is used to process the plurality of target sub-intermediate frequency signals using a fast Fourier transform algorithm to obtain time-speed characteristic data; and
[0047] The third processing unit is used to process the time and speed feature data using the trained recognition model to obtain a target authentication result corresponding to the user.
[0048] Furthermore, the device also includes:
[0049] An acquisition unit, used to acquire a meta-trained recognition model and sample time and speed feature data corresponding to the user;
[0050] A fourth processing unit, used to process the sample time speed feature data using the meta-trained recognition model to obtain a sample recognition result;
[0051] A training unit is used to train the meta-trained recognition model based on the difference between the sample recognition result and the accurate label to obtain the trained feature extraction model,
[0052] The meta-trained recognition model is obtained by training with a pre-training sample set, and the sample time-speed feature data is determined by a plurality of sample transmission signals and a sample echo signal corresponding to each transmission signal.
[0053] On the other hand, an embodiment of the present specification further provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above method when executing the computer program.
[0054] On the other hand, an embodiment of the present specification further provides a computer-readable storage medium on which computer instructions are stored, and the computer instructions implement the above method when executed by a processor.
[0055] On the other hand, an embodiment of the present specification further provides a computer program product, including a computer program / instruction, and a method implemented when the computer program / instruction is executed by a processor.
[0056] Using the embodiment of this specification, during the user's signature process, the echo signal corresponding to the multiple transmission signals emitted by the millimeter wave radar is received, and the target sub-intermediate frequency signal after filtering and denoising based on the transmission signal and the echo signal is processed using the fast Fourier transform algorithm to obtain time and speed feature data, and then the trained recognition model is used to process the time and speed feature data to obtain the corresponding target authentication result. Thus, based on the transmission signal emitted by the millimeter wave radar and the received reflected echo signal, the time and speed feature data corresponding to the signature process is determined, and online handwriting signature authentication is realized without using a specific handwriting device. The extracted time and speed feature data is further used to characterize the vertical movement of the user during the signature process, thereby ensuring the accuracy of the authentication. BRIEF DESCRIPTION OF THE DRAWINGS
[0057] In order to more clearly illustrate the embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0058] Figure 1 The figure shows a schematic diagram of an implementation system of an online signature authentication method according to an embodiment of the present specification;
[0059] Figure 2It is a flowchart of an online signature authentication method according to an embodiment of this specification;
[0060] Figure 3 The figure is a flow chart of a method for determining a target sub-intermediate frequency signal according to an embodiment of the present specification;
[0061] Figure 4A It is a flow chart of a method for determining time speed characteristic data according to an embodiment of this specification;
[0062] Figure 4B The figure is a flow chart of a method for determining a target signal to be processed according to an embodiment of the present specification;
[0063] Figure 5 Shown is a schematic diagram of an online signature authentication method according to another embodiment of the present specification;
[0064] Fig. 6A It is a flowchart of a training method of a trained recognition model according to an embodiment of this specification;
[0065] Figure 6B Shown is a schematic diagram of a training method for a trained recognition model according to another embodiment of the present specification;
[0066] Fig. 7A The figure shows a schematic diagram of the structure of an online signature authentication device according to an embodiment of the present specification;
[0067] Figure 7B It is a schematic diagram of the structure of an online signature authentication device according to another embodiment of the present specification;
[0068] Figure 8 This is a schematic diagram of the structure of a computer device according to an embodiment of this specification.
[0069] [Description of Reference Numerals]
[0070] 110. Electronic equipment;
[0071] 120. Database;
[0072] 130. Recognition model after training;
[0073] 140. User terminal;
[0074] 150. Target authentication results;
[0075] 160, echo signal;
[0076] 710, receiving unit;
[0077] 720, first processing unit;
[0078] 730, second processing unit;
[0079] 740, third processing unit;
[0080] 750, acquisition unit;
[0081] 760, fourth processing unit;
[0082] 770, training unit;
[0083] 802. Computer equipment;
[0084] 804. Processing equipment;
[0085] 806. Storage resources;
[0086] 808, driving mechanism;
[0087] 810, input / output module;
[0088] 812. Input device;
[0089] 814. Output device;
[0090] 816. Presentation equipment;
[0091] 818. Graphical user interface;
[0092] 820, network interface;
[0093] 822, communication link;
[0094] 824. Communication bus. DETAILED DESCRIPTION
[0095] The following will be combined with the drawings in the embodiments of this specification to clearly and completely describe the technical solutions in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this specification.
[0096] It should be noted that the terms "first", "second", etc. in the description and claims of this specification and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of this specification described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, device, product or equipment that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or equipment.
[0097] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.
[0098] Figure 1 The diagram is a schematic diagram of an implementation system of an online signature authentication method according to an embodiment of the present specification, which may include an electronic device 110 and a database 120 .
[0099] The electronic device 110 can access the database 120 through a network, for example. The database 120 can store a meta-trained recognition model, sample time-speed feature data, and an accurate label corresponding to the sample time-speed feature data, wherein the accurate label represents an accurate recognition result corresponding to the sample time-speed feature data. It should be noted that the meta-trained recognition model is obtained by training using a pre-training sample set, and the sample time-speed feature data is determined by a plurality of sample transmission signals and a sample echo signal corresponding to each transmission signal.
[0100] In one embodiment, the electronic device 110 can read the sample time-speed feature data and the accurate label from the database 120, and use the sample time-speed feature data and the accurate label as samples to train the meta-trained recognition model to obtain the trained recognition model. It should be noted that the training process of the meta-trained recognition model is similar to the training process of the trained recognition model, and only the samples used for training are different, which will not be repeated here. Using the sample time-speed feature data and the accurate label as samples to train the meta-trained recognition model to obtain the trained recognition model can specifically include using the meta-trained recognition model to process the sample time-speed feature data to obtain the sample recognition result; based on the difference between the sample recognition result and the accurate label, the meta-trained recognition model is trained to obtain the trained feature extraction model. The meta-trained recognition model can, for example, include a meta-trained autoencoder model.
[0101] In one embodiment, the application scenario may also include a trained recognition model 130, a user terminal 140, a target authentication result 150, and an echo signal 160, wherein the echo signal is a signal reflected from a transmission signal emitted by the user terminal 140 using a millimeter wave radar. After receiving the echo signal 160, the user terminal 140 filters and denoises the transmission signal and the echo signal to obtain a plurality of target sub-intermediate frequency signals; and processes the plurality of target sub-intermediate frequency signals using a fast Fourier transform algorithm to obtain time and speed feature data. The user terminal 140 is connected to the electronic device 110 through a network. For example, the user terminal 140 may obtain a trained recognition model 130 (i.e., a trained recognition model obtained by training using sample time and speed feature data corresponding to the user) of the user corresponding to the echo signal from the electronic device 110, and processes the time and speed feature data based on the obtained trained recognition model 130 to obtain a target authentication result 150 corresponding to the user. It should be noted that the process of obtaining the time-speed characteristic data based on the transmitted signal and the echo signal can be determined by the user terminal 140, or can be sent by the user terminal 140 to the electronic device 110, and the electronic device 110 determines the time-speed characteristic data and sends the time-speed characteristic data to the user terminal 140.
[0102] It should be noted that the training method of the trained recognition model provided in this specification can generally be executed by the electronic device 110, or can be executed by a server connected to the electronic device 110 in communication. The online signature authentication provided in this specification can be executed by the user terminal 140 or the electronic device 110. Accordingly, the trained recognition model device provided in this specification can generally be set in the electronic device 110, or can be set in a server connected to the electronic device 110 in communication. The online signature authentication provided in this specification can be set in the user terminal 140 or the electronic device 110.
[0103] It should be understood that Figure 1 The electronic devices, databases, and echo signals in the figure are only illustrative. According to the implementation requirements, there may be any number and type of electronic devices, databases, echo signals, and user terminals.
[0104] like Figure 2 The flowchart of an online signature authentication method of an embodiment of this specification is shown. The figure describes the online signature authentication process, but it may include more or fewer operation steps based on conventional or non-creative work. The order of steps listed in the embodiment is only one way of executing the steps among many orders, and does not represent the only order of execution. When the actual system or device product is executed, it can be executed in the order of the method shown in the embodiment or the accompanying drawings or in parallel. Specifically, Figure 2 As shown, the method may include:
[0105] S210, during the user signing process, receiving echo signals corresponding to multiple transmission signals of the millimeter wave radar;
[0106] S220, filtering and denoising the transmission signal and the echo signal to obtain a plurality of target sub-intermediate frequency signals;
[0107] S230, using a fast Fourier transform algorithm to process multiple target sub-intermediate frequency signals to obtain time and speed characteristic data;
[0108] S240, using the trained recognition model to process the time and speed feature data to obtain a target authentication result corresponding to the user.
[0109] Using the embodiment of this specification, during the user's signature process, the echo signal corresponding to the multiple transmission signals emitted by the millimeter wave radar is received, and the target sub-intermediate frequency signal after filtering and denoising based on the transmission signal and the echo signal is processed using the fast Fourier transform algorithm to obtain time and speed feature data, and then the trained recognition model is used to process the time and speed feature data to obtain the corresponding target authentication result. Thus, based on the transmission signal emitted by the millimeter wave radar and the received reflected echo signal, the time and speed feature data corresponding to the signature process is determined, and online handwriting signature authentication is realized without using a specific handwriting device. The extracted time and speed feature data is further used to characterize the vertical movement of the user during the signature process, thereby ensuring the accuracy of the authentication.
[0110] According to an embodiment of the present specification, during the user signing process, the millimeter wave radar transmits multiple transmission signals, which may be frequency modulated continuous wave (FMCW) signals, for example. After the transmission signal is transmitted, the transmission signal will reflect back the corresponding signal when it encounters the user's hand signing, and the reflected signal corresponding to the transmission signal is the echo signal corresponding to the transmission signal.
[0111] The transmission signal may be, for example, a FMCW signal emitted by any millimeter wave radar. Specifically, the transmission signal may be, for example, as shown in the following formula (1).
[0112]
[0113] Among them, S T (t) represents the transmitted signal, A1 represents the transmitted amplitude, f0 represents the initial frequency, t represents the transmitting time, S represents the frequency modulation slope, and Characterizes the initial phase of the transmitted signal.
[0114] When the transmitted signal is as shown in the above formula (1), the echo signal reflected back by the moving hand can be expressed as shown in the following formula (2).
[0115]
[0116] Among them, S R (t) represents the echo signal, A2 represents the echo amplitude, f0 represents the initial frequency, t represents the sending time of the sending signal corresponding to the echo signal, τ represents the time delay, S represents the frequency modulation slope, and Characterizes the initial phase of the received signal.
[0117] The filtering and denoising processing for the transmission signal and the echo signal may be, for example, mixing the transmission signal and the echo signal to obtain an intermediate frequency signal; and denoising the intermediate frequency signal to obtain a target sub-intermediate frequency signal. Mixing (which may include low-pass filtering) and denoising processing may be, for example, processed by any method that can achieve the above functions, and this specification does not limit this. It should be noted that when denoising the intermediate frequency signal, denoising may be performed on all intermediate frequency signals, and sampling may be performed on all intermediate frequency signals first to obtain a sampled intermediate frequency signal, and then denoising may be performed on the sampled intermediate frequency signal to obtain a target sub-intermediate frequency signal.
[0118] The fast Fourier transform algorithm may be any fast Fourier transform algorithm, such as a distance dimension fast Fourier transform algorithm and a Doppler dimension fast Fourier transform algorithm, etc. The fast Fourier transform algorithm is used to process the obtained multiple target sub-intermediate frequency signals to obtain time series data.
[0119] The horizontal movement of a user's handwritten signature usually increases linearly with small oscillations on the linear curve, while the vertical movement changes back and forth between positive and negative values, and the changes are more frequent and the oscillations are more obvious. Therefore, only vertical movement can be considered and horizontal movement can be ignored. Compared with changes in distance, millimeter-wave radar is more sensitive to changes in speed, so changes in speed in the vertical direction can be used as features of handwritten signatures. In order to obtain information on speed changes in the vertical direction when handwriting signatures, the information representing the vertical movement in the time series data is extracted, and the information representing the vertical movement is used as time-speed feature data. This can reduce computing costs, separate the interference of horizontal movement, and further improve recognition accuracy.
[0120] The trained recognition model is used to process the time and speed characteristic data to obtain a target authentication result corresponding to the user in the signing process.
[0121] The trained recognition model can be, for example, trained based on the training sample time-speed feature data and the sample standard recognition results corresponding to the training sample time-speed feature data. Specifically, the training sample time-speed feature data is processed using a preset recognition model to obtain a prediction result, and further, the difference between the prediction result and the sample standard recognition result is used to train the preset recognition model until a trained recognition model that meets the conditions is obtained. It should be noted that the condition can be, for example, the number of training times, the convergence of a series of loss function values that characterize the difference, etc. The preset recognition model can be, for example, a preset autoencoder model (Transformer).
[0122] The training sample time-speed characteristic data may be a training sample receiving signal corresponding to a plurality of training sample sending signals of the millimeter wave radar collected for the sample user. Further, the operations such as S220 to S230 are adopted to process the plurality of training sample sending signals and the plurality of training sample receiving signals to obtain the training sample time-speed characteristic data.
[0123] Figure 3 The flowchart of a method for determining a target sub-IF signal in an embodiment of the present specification is shown. In this figure, a process for determining a target sub-IF signal is described, but based on conventional or non-creative work, more or fewer operation steps may be included. Specifically, Figure 3 As shown, the method may include:
[0124] S321, performing mixing and low-pass filtering on the transmission signal and the echo signal to obtain an intermediate frequency signal;
[0125] S322, performing signal interference denoising processing on the intermediate frequency signal to obtain a first intermediate frequency signal;
[0126] S323: Perform dynamic interference denoising on the first intermediate frequency signal corresponding to each transmitted signal to obtain multiple target sub-intermediate frequency signals.
[0127] According to another embodiment of the present specification, the transmission signal and the echo signal are mixed and low-pass filtered to obtain an intermediate frequency signal, which may be processed based on the following formula (3).
[0128]
[0129] Among them, S IF (t) represents the intermediate frequency signal, LPF{·} represents the low-pass filtering operation, S T (t) represents the transmitted signal, S R (t) represents the echo signal, A represents the mixing amplitude, and f IF Characterization frequency, f IF =Sτ, S represents the frequency modulation slope, τ represents the time delay, and t represents the sending time of the sending signal. Characterizes the phase of the intermediate frequency signal.
[0130] In addition to the handwritten signature movement, the transmitted signal and the echo signal also include other interference signals (such as static objects in the environment and the user's torso, etc.). In order to improve the accuracy of signature recognition, the embodiment of the present application performs signal interference denoising on the intermediate frequency signal after obtaining the intermediate frequency signal. The signal interference denoising process can, for example, determine all the intermediate frequency signals corresponding to each transmitted signal; and determine the mean of all the intermediate frequency signals as interference noise. Furthermore, the interference noise is subtracted from each intermediate frequency signal to obtain a first intermediate frequency signal. Thus, the intermediate frequency signal corresponds to the first intermediate frequency signal.
[0131] It should be noted that a sampling step can be added before the intermediate frequency signal is subjected to signal interference denoising, that is, a sampling process is performed on the intermediate frequency signal to obtain a sampled intermediate frequency signal, and then a signal interference denoising process is performed on the sampled intermediate frequency signal to obtain a first intermediate frequency signal. Specifically, for each transmitted signal, all corresponding sampled intermediate frequency signals are determined; and the mean of all sampled intermediate frequency signals is determined as the interference noise. Further, the interference noise is subtracted from each sampled intermediate frequency signal to obtain a first intermediate frequency signal. Thus, the sampled intermediate frequency signal corresponds to the first intermediate frequency signal. If the intermediate frequency signal is sampled before signal interference denoising, the sampling can be, for example, as shown in the following formula (4).
[0132] S SIF (n) = S IF (t)·μ(n),n=0,1,…,N-1 Formula (4)
[0133] Among them, S IF (t) represents the intermediate frequency signal, S SIF (n) represents the sampled intermediate frequency signal, μ(n) represents the unit step sequence, and N represents the number of samples (fast time number) of the corresponding intermediate frequency signal.
[0134] According to another embodiment of the present specification, the intermediate frequency signal is subjected to signal interference denoising to obtain the first intermediate frequency signal. For example, the intermediate frequency signal is subjected to processing based on the following formula (5) to obtain the first intermediate frequency signal.
[0135]
[0136] in, Characterizes the i-th first intermediate frequency signal, represents the i-th intermediate frequency signal, M represents the number of intermediate frequency signals corresponding to each transmitted signal, and It should be noted that if sampling is performed before signal interference denoising, Characterizes the i-th sampled intermediate frequency signal, represents the i-th sampled first intermediate frequency signal, M represents the number of sampled intermediate frequency signals corresponding to each transmitted signal, and Characterize the mth sampled intermediate frequency signal.
[0137] The first intermediate frequency signal is obtained to represent the interference-free signal obtained during the user's signature process. However, the interference-free signal includes not only the position signal of the hand during the user's signature process, but also the position signals of other parts. In order to extract the hand position signal in the first intermediate frequency signal, this specification performs dynamic interference denoising on the obtained first intermediate frequency signal to obtain a target sub-intermediate frequency signal.
[0138] According to another embodiment of the present specification, dynamic interference denoising is performed on the first intermediate frequency signal corresponding to each transmitted signal to obtain multiple target sub-intermediate frequency signals. For example, the first intermediate frequency signal is processed based on the following formula (6) to obtain multiple target sub-intermediate frequency signals.
[0139]
[0140] Among them, S i,k Represents the kth target sub-IF signal corresponding to the i-th first IF signal, represents the nth sampled first intermediate frequency signal in the i-th first intermediate frequency signal, N represents the number of sampling points of the first intermediate frequency signal, δ represents the unit impulse function, K represents the number of target sub-intermediate frequency signals corresponding to the i-th first intermediate frequency signal, and L represents the length of the target sub-intermediate frequency signal, which is determined by the number of sampling points and the number of target sub-intermediate frequency signals.
[0141] It should be noted that in the above formula (6) The meaning of characterizing the nth sampled first intermediate frequency signal in the i-th first intermediate frequency signal is that, in the case of sampling before processing the intermediate frequency signal, the Represents the nth sampled first intermediate frequency signal; if the intermediate frequency signal is not sampled before processing, each first intermediate frequency signal is a corresponding sampled first intermediate frequency signal, and there are as many sampled first intermediate frequency signals as there are first intermediate frequency signals. Therefore, the sampled first intermediate frequency signal here does not represent the sampling of the first intermediate frequency signal, but represents whether the intermediate frequency signal is sampled before signal interference denoising. Therefore, when the intermediate frequency signal is sampled before signal interference denoising, the number of sampling points of the first intermediate frequency signal is the same as the number of sampling points of the intermediate frequency signal; when the intermediate frequency signal is not sampled before signal interference denoising, the number of sampling points of the first intermediate frequency signal is the same as the number of first intermediate frequency signals.
[0142] The length is determined by the number of sampling points and the number of target sub-IF signals and can be specifically determined based on the following formula (7).
[0143]
[0144] Among them, L represents the length of the target sub-IF signal, N represents the number of sampling points of the first IF signal, and K represents the number of target sub-IF signals corresponding to the i-th first IF signal. Therefore, the determined target sub-IF signals all utilize the entire original bandwidth, so the obtained distance resolution is not reduced.
[0145] It can be seen that a transmitted signal corresponds to multiple intermediate frequency signals (or multiple sampled intermediate frequency signals), an intermediate frequency signal (or a sampled intermediate frequency signal) corresponds to a first intermediate frequency signal (or a sampled first intermediate frequency signal), and a first intermediate frequency signal (or a sampled first intermediate frequency signal) corresponds to multiple target sub-intermediate frequency signals.
[0146] Figure 4A The flowchart of a method for determining time and speed characteristic data according to an embodiment of the present specification is shown. Figure 4A A process for determining time and speed characteristic data is described in , but it may include more or fewer operation steps based on conventional or non-creative labor. Figure 4A As shown, the method may include:
[0147] S431, using a distance-dimensional fast Fourier transform algorithm to process each target sub-intermediate frequency signal to obtain multiple candidate signals;
[0148] S432, selecting a target signal to be processed from the multiple candidate signals based on the peak values of the multiple candidate signals corresponding to each target sub-IF signal;
[0149] S433, using a Doppler fast Fourier transform algorithm to process the target signal to be processed to obtain range Doppler time series data;
[0150] S434, using the summary characteristic formula to process the range Doppler time series data to obtain time velocity characteristic data.
[0151] According to another embodiment of the present specification, the fast Fourier transform algorithm includes a range dimension fast Fourier transform algorithm (Range FFT), a Doppler dimension fast Fourier transform algorithm (Doppler FFT) and a summary characteristic formula.
[0152] Range FFT is used to process each target sub-IF signal to obtain multiple candidate signals (slow time samples). It should be noted that the multiple candidate signals may be, for example, signals corresponding to each range gate in the spectrum obtained by the range-dimensional fast Fourier transform algorithm.
[0153] After each target sub-intermediate frequency signal passes through the distance dimension fast Fourier transform algorithm, multiple candidate signals corresponding to it are obtained. The peak value of the candidate signal is determined from the multiple candidate signals, such as the largest candidate signal among the multiple candidate signals. Based on the peak value, the candidate signal to be processed corresponding to each target sub-intermediate frequency signal is selected from the multiple candidate signals. It should be noted that, for example, the candidate signal within the target neighborhood of the range gate corresponding to the peak value can be selected as the candidate signal to be processed. Then, based on a similar method, for each transmitted signal, the corresponding target signal to be processed is determined.
[0154] Then, each target signal to be processed is processed using Doppler FFT to obtain range Doppler time series data, thereby obtaining range Doppler time series data corresponding to each transmitted signal.
[0155] In order to reduce the computational overhead, the multiple range Doppler time series data can be aggregated into a feature graph. Specifically, the range Doppler time series data are processed using the aggregate feature formula to obtain time velocity feature data. For example, the multiple range Doppler time series data can be processed using the aggregate feature formula shown in the following formula (8) to obtain time velocity feature data.
[0156]
[0157] Among them, V (f,i) Characterizing time-speed characteristic data, RDM (f,i,j) Represents the corresponding value of the target gate j corresponding to the Doppler gate s and the target signal to be processed in the f-th range Doppler time series data, R j The index representing the target gate, N represents the first number of points corresponding to the distance dimension fast Fourier transform algorithm, and M represents the second number of points corresponding to the Doppler dimension fast Fourier transform algorithm. It should be noted that the target gate is the range gate corresponding to the target signal to be processed in the frequency spectrum corresponding to the target signal to be processed (which can be understood as the horizontal coordinate).
[0158] Figure 4B The flowchart of a method for determining a target signal to be processed according to an embodiment of the present specification is shown. Figure 4B A process for determining a target signal to be processed is described in , but it may include more or fewer operation steps based on conventional or non-creative work. Figure 4B As shown, the method may include:
[0159] S4321, for each target sub-intermediate frequency signal, determining a range gate corresponding to a peak value;
[0160] S4322, for a first intermediate frequency signal corresponding to a plurality of target sub-intermediate frequency signals, using a voting method to process a plurality of range gates corresponding to the plurality of target sub-intermediate frequency signals to obtain a target range gate corresponding to the first intermediate frequency signal;
[0161] S4323, determining a target area based on a preset window value and a target range gate for transmission signals corresponding to the plurality of first intermediate frequency signals;
[0162] S4324, taking the candidate signals corresponding to each target gate in the target area as target signals to be processed.
[0163] According to another embodiment of the present specification, for each target sub-IF signal, multiple candidate signals corresponding thereto are determined; peak values of the multiple candidate signals are determined; and a range gate corresponding to the peak value is determined. The above peak values and range gates can be obtained from the spectrum corresponding to the target sub-IF signal, for example.
[0164] The first intermediate frequency signal includes a signal obtained after filtering and signal interference denoising the transmission signal and the echo signal, and the target sub-intermediate frequency signal includes a signal obtained after dynamic interference denoising is performed on the first intermediate frequency signal. Each transmission signal corresponds to a plurality of first intermediate frequency signals, and each first intermediate frequency signal corresponds to a plurality of target sub-intermediate frequency signals.
[0165] For each first intermediate frequency signal, multiple target sub-intermediate frequency signals corresponding thereto are determined. The range gate corresponding to each target sub-intermediate frequency signal corresponding to the multiple target sub-intermediate frequency signals is processed using a voting method to obtain a target range gate corresponding to each first intermediate frequency signal.
[0166] For each transmitted signal, a plurality of first intermediate frequency signals corresponding thereto are determined. Based on a preset window value and a target range gate corresponding to each first intermediate frequency signal, a target area is determined.
[0167] According to another embodiment of the present specification, the expression of the target area corresponding to each transmission signal is shown in the following formula (9).
[0168]
[0169] in, Characterize the target area, p min Characterizes the smallest target range gate among the target range gates, p max represents the largest target range gate in the target range gates, W represents the preset window value, N represents a positive integer, and a represents the target gate. It should be noted that the target area is an area in the spectrum, the area has multiple range gates, and each range gate in the area is the target gate.
[0170] Figure 5 Shown is a schematic diagram of an online signature authentication method according to another embodiment of the present specification.
[0171] like Figure 5 As shown, during the user signature process, a millimeter wave radar is used to send multiple transmission signals, and an echo signal corresponding to each transmission signal is received. After obtaining the transmission signal and the echo signal, the transmission signal and the echo signal are preprocessed, specifically filtering and denoising the transmission signal and the echo signal (specifically, the processing can be based on the above formula (1) to formula (5)) to obtain the first intermediate frequency signal (IF signal) after noise elimination. It should be noted that in Figure 5 Only one first intermediate frequency signal is schematically shown, which does not mean that there is only one first intermediate frequency signal. The embodiments of this specification do not limit the number of first intermediate frequency signals.
[0172] Formula (6) is used to process each first intermediate frequency signal to obtain a plurality of target sub-intermediate frequency signals (sub-signal 1, sub-signal 2, ..., sub-signal k).
[0173] The target sub-IF signals are processed based on the distance dimension fast Fourier transform algorithm (distance FFT) to obtain the target area as shown in formula (9). The target to-be-processed signals corresponding to the target gate in the target area are processed using the Doppler dimension fast Fourier transform algorithm (Doppler FFT) to obtain the distance Doppler data. Figure 5 The region of interest in is composed of the range Doppler time series data of the target signal to be processed corresponding to the target gate in the target area after Doppler FFT processing. Then, the range Doppler time series data is subjected to summary feature processing, specifically, the range Doppler time series data is processed using formula (8) to obtain time-velocity feature data (2D time-velocity feature diagram).
[0174] Furthermore, the trained recognition model (a neural network based on Transformer) obtained by training based on small sample learning is used to process the time speed feature data to obtain a target authentication result, which is information indicating whether the authentication is successful or failed.
[0175] Fig. 6A FIG. 1 is a flowchart of a training method for a trained recognition model according to an embodiment of the present specification. Figure 6B Shown is a schematic diagram of a training method for a trained recognition model according to another embodiment of the present specification. Fig. 6AFIG. 1 is a schematic diagram of a training method for a trained recognition model according to an embodiment of the present specification. In this figure, another process for determining the first candidate interval is described, but more or fewer operation steps may be included based on conventional or non-creative labor. Fig. 6A As shown, the method may include:
[0176] S650, obtaining a meta-trained recognition model and sample time and speed feature data corresponding to the user;
[0177] S660, using the meta-trained recognition model to process the sample time and speed feature data to obtain a sample recognition result;
[0178] S670, based on the difference between the sample recognition result and the accurate label, train the meta-trained recognition model to obtain a trained recognition model.
[0179] According to another embodiment of the present specification, for new users, in order to enable them to quickly adapt to the system based on a small number of training samples and to improve the trained recognition model trained in this case, the present specification relates to a small sample learning method.
[0180] The meta-trained recognition model is trained using a pre-training sample set, which includes pre-training samples, each of which is based on Figure 2 S210~S230 are obtained. Specifically, the recognition model after meta-training is obtained by training with a pre-training sample set, and the pre-training samples included in the pre-training sample set are processed with a preset recognition model to obtain a prediction result, and then based on the difference between the prediction result and the accurate label corresponding to the pre-training sample included in the pre-training sample set, the preset recognition model is trained to obtain the trained recognition model. It should be noted that the preset recognition model is, for example, a Transformer model, which specifically includes: processing the pre-training sample with a channel embedding layer (patch embedding layer) to divide the pre-training sample into multiple channels (patches); adding position information to each channel with a position embedding layer (position embedding layer) to obtain an embedded vector sequence; processing the embedded vector sequence with an autoencoder (Transformer encoder) to obtain a prediction result. It should be noted that the Transformer model may include, for example, two autoencoders, each of which includes a multi-head self-attention module (MSA), a normalization layer (LN layer) and a multi-layer perceptron (MLP), and multiple self-attention modules are composed of two linear projections separated nonlinearly by an activation function (GELU).
[0181] After obtaining the meta-trained recognition model, the sample time-speed feature data corresponding to the user is obtained. The user is the user for handwritten signature recognition. The sample time-speed feature data is determined by multiple sample transmission signals and sample echo signals corresponding to each transmission signal. Specifically, for example, it can be based on Figure 2 The multiple sample transmission signals and sample echo signals are the user's registration process, based on the following Figure 2 Obtained from S210.
[0182] The sample time and speed feature data are processed using the meta-trained recognition model to obtain the sample recognition result. Since the sample recognition speed feature data are all data from user registration, the labels corresponding to the sample time and speed feature data are all accurate labels (i.e., data that represent successful authentication).
[0183] Further, based on the difference between the sample recognition result and the accurate label, the meta-trained recognition model is trained to obtain a trained recognition model that meets the preset conditions. It should be noted that the preset conditions may be, for example, that the number of trainings meets the number threshold and / or that the series consisting of the loss function values corresponding to the difference in each training process converges. It should be noted that, in the application scenario of user dynamic signatures, the sample time-speed feature data corresponding to the user (for example, collected by the user during the registration phase) is relatively small, and the training of the recognition model with relatively small sample time-speed feature data will lead to the problem of low recognition accuracy of the trained recognition model. The embodiment of this specification adopts a small sample learning method to train the recognition model, and realizes the training of a trained recognition model with high recognition accuracy based on a small amount of sample time-speed feature data corresponding to the user.
[0184] Specifically, the small sample learning method can be as follows Figure 6B As shown. Pre-training sample set It is divided into training data (also called support set) and test data (also called query set). When receiving the training task, the support set and query set are analyzed using the preset recognition model. Fig. 6A The training shown in FIG. 1 obtains the recognition model (meta-learner) after meta-training.
[0185] Furthermore, the classifier included in the meta-learner is used to classify the sample time speed feature data ( The sample time-speed feature data (including support set and query set) is processed to obtain the authentication result (sample recognition result), and then based on the difference between the authentication result and the accurate label, the classifier is fine-tuned to obtain the trained recognition model.
[0186] Specifically, when training the meta-learner, Select any two users' data as tasks Each task has a corresponding support set ( Each user has K training data) and query set ( The rest of the training data). For the preset recognition model with initial parameters θ, the parameter update algorithm (MAML) is used to update the model parameters. Specifically, for each task, K training samples are used to evaluate The updated parameters can be shown as the following formula (10).
[0187]
[0188] Among them, θ′ i represents the updated parameters, θ represents the parameters to be updated, α is the step size hyperparameter, Based on the task The loss function obtained by the corresponding support set.
[0189] After getting the updated parameters for all tasks, use the query set corresponding to each task To calculate the loss The total loss of all tasks can be expressed as follows:
[0190]
[0191] Among them, M represents the number of tasks, Representation based on task The loss function obtained by the corresponding query set. The parameters θ of the meta-trained recognition model (meta-learner) can be obtained by minimizing the objective function through the stochastic gradient descent (SGD), specifically, the minimum parameters obtained by the following formula (12).
[0192]
[0193] Among them, β represents the step size hyperparameter of the meta-learner, θ f represents the parameters of the meta-learner, θ represents the initial parameters of the meta-learner at the beginning of the current training, Representation based on task The loss function obtained by the corresponding query set, and M represents the number of tasks.
[0194] It should be noted that the above only uses one parameter as an example to illustrate the training process of the meta-trained recognition model, but the meta-trained recognition model may also include multiple parameters, and the specific parameters included are not limited in this specification.
[0195] Fig. 7AThe structure diagram of an online signature authentication device according to an embodiment of the present specification is shown in FIG7A. As shown in FIG7A , the device comprises:
[0196] The receiving unit 710 is used to receive the echo signal corresponding to the multiple transmission signals of the millimeter wave radar during the user signing process;
[0197] The first processing unit 720 is used to filter and denoise the transmission signal and the echo signal to obtain multiple target sub-intermediate frequency signals;
[0198] The second processing unit 730 is used to process the multiple target sub-intermediate frequency signals using a fast Fourier transform algorithm to obtain time-speed characteristic data; and
[0199] The third processing unit 740 is used to process the time and speed feature data using the trained recognition model to obtain a target authentication result corresponding to the user.
[0200] Since the principle of solving the problem by the above device is similar to that of the above method, the implementation of the above device can refer to the implementation of the above method, and the repeated parts will not be repeated.
[0201] Figure 7B FIG. 7B shows a schematic diagram of the structure of an online signature authentication device according to another embodiment of the present specification. As shown in FIG. 7B , the device comprises:
[0202] An acquisition unit 750 is used to acquire the meta-trained recognition model and sample time and speed feature data corresponding to the user;
[0203] The fourth processing unit 760 is used to process the sample time speed feature data using the meta-trained recognition model to obtain a sample recognition result;
[0204] The training unit 770 is used to train the meta-trained recognition model based on the difference between the sample recognition result and the accurate label to obtain a trained feature extraction model.
[0205] The meta-trained recognition model is obtained by training with a pre-training sample set, and the sample time-speed feature data is determined by a plurality of sample transmission signals and a sample echo signal corresponding to each transmission signal.
[0206] Since the principle of solving the problem by the above device is similar to that of the above method, the implementation of the above device can refer to the implementation of the above method, and the repeated parts will not be repeated.
[0207] like Figure 8The structure diagram of a computer device in an embodiment of this specification is shown. The apparatus in this specification can be a computer device in this embodiment, and the method of this specification is executed. The computer device 802 may include one or more processing devices 804, such as one or more central processing units (CPUs), and each processing unit may implement one or more hardware threads. The computer device 802 may also include any storage resource 806, which is used to store any kind of information such as code, settings, data, etc. Non-limiting, for example, the storage resource 806 may include any one or more combinations of the following: any type of RAM, any type of ROM, flash memory device, hard disk, optical disk, etc. More generally, any storage resource can use any technology to store information. Further, any storage resource can provide volatile or non-volatile retention of information. Further, any storage resource can represent a fixed or removable component of the computer device 702. In one case, when the processing device 804 executes an associated instruction stored in any storage resource or a combination of storage resources, the computer device 802 can perform any operation of the associated instruction. The computer device 802 also includes one or more drive mechanisms 808 for interacting with any storage resources, such as a hard disk drive mechanism, an optical disk drive mechanism, and the like.
[0208] The computer device 802 may also include an input / output module 810 (I / O) for receiving various inputs (via input devices 812) and for providing various outputs (via output devices 814). A specific output mechanism may include a presentation device 816 and an associated graphical user interface (GUI) 818. In other embodiments, the input / output module 810 (I / O), the input device 812, and the output device 814 may not be included, and the computer device 802 may be used as a computer device in a network. The computer device 802 may also include one or more network interfaces 820 for exchanging data with other devices via one or more communication links 822. One or more communication buses 824 couple the components described above together.
[0209] The communication link 822 may be implemented in any manner, for example, through a local area network, a wide area network (e.g., the Internet), a point-to-point connection, etc., or any combination thereof. The communication link 822 may include any combination of hardwired links, wireless links, routers, gateway functions, name servers, etc. governed by any protocol or combination of protocols.
[0210] The embodiments of the present specification also provide a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above method is implemented.
[0211] The embodiments of this specification also provide a computer program product, which includes a computer program. When the computer program is executed by a processor, the above method is implemented.
[0212] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems, or computer program products. Therefore, this specification may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.
[0213] This specification is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of this specification. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0214] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0215] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0216] The above specific embodiments further illustrate the purpose, technical solutions and beneficial effects of this specification. It should be understood that the above are only specific embodiments of this specification and are not intended to limit the scope of protection of this specification. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of this specification should be included in the scope of protection of this specification.
Claims
1. An online signature authentication method, It is characterized in that include: During the user signing process, receiving echo signals corresponding to a plurality of transmission signals of the millimeter wave radar; Filtering and denoising the transmission signal and the echo signal to obtain a plurality of target sub-intermediate frequency signals; Using a fast Fourier transform algorithm to process the multiple target sub-intermediate frequency signals to obtain time-speed characteristic data; as well as The trained recognition model is used to process the time and speed feature data to obtain a target authentication result corresponding to the user.
2. The method according to claim 1, It is characterized in that The filtering and denoising of the transmission signal and the echo signal to obtain a plurality of target sub-intermediate frequency signals comprises: Performing mixing and low-pass filtering on the transmission signal and the echo signal to obtain an intermediate frequency signal; Performing signal interference denoising processing on the intermediate frequency signal to obtain a first intermediate frequency signal; and Dynamic interference denoising is performed on the first intermediate frequency signal corresponding to each of the transmitted signals to obtain the multiple target sub-intermediate frequency signals.
3. The method according to claim 2, It is characterized in that The performing signal interference denoising processing on the intermediate frequency signal to obtain a first intermediate frequency signal comprises: in, Characterizes the i-th first intermediate frequency signal, represents the ith intermediate frequency signal, M represents the number of intermediate frequency signals corresponding to each of the transmitted signals, and Characterizes the mth intermediate frequency signal.
4. The method according to claim 2, It is characterized in that The performing dynamic interference denoising on the first intermediate frequency signal corresponding to each of the transmitted signals to obtain the multiple target sub-intermediate frequency signals comprises: Among them, S i,k characterizes the kth target sub-intermediate frequency signal corresponding to the i-th first intermediate frequency signal, represents the nth sampled first intermediate frequency signal in the i-th first intermediate frequency signal, M represents the number of sampling points of the first intermediate frequency signal, δ represents the unit impulse function, K represents the number of the target sub-intermediate frequency signals corresponding to the i-th first intermediate frequency signal, and L represents the length of the target sub-intermediate frequency signal, which is determined by the number of sampling points and the number of the target sub-intermediate frequency signals.
5. The method according to claim 1, It is characterized in that The fast Fourier transform algorithm includes a distance dimension fast Fourier transform algorithm, a Doppler dimension fast Fourier transform algorithm and a summary characteristic formula. The fast Fourier transform algorithm is used to process the multiple target sub-intermediate frequency signals to obtain time velocity characteristic data including: Using the distance-dimensional fast Fourier transform algorithm to process each of the target sub-intermediate frequency signals to obtain multiple candidate signals; Selecting a target signal to be processed from the multiple candidate signals based on the peak values of the multiple candidate signals corresponding to each of the target sub-IF signals; Using the Doppler fast Fourier transform algorithm to process the target signal to be processed, to obtain range Doppler time series data; and The range Doppler time series data is processed using the summary characteristic formula to obtain the time velocity characteristic data.
6. The method according to claim 5, It is characterized in that The selecting a target signal to be processed from the multiple candidate signals based on the peak values of the multiple candidate signals corresponding to each of the target sub-IF signals comprises: For each of the target sub-intermediate frequency signals, determining a range gate corresponding to the peak value; For a first intermediate frequency signal corresponding to a plurality of target sub-intermediate frequency signals, a plurality of range gates corresponding to the plurality of target sub-intermediate frequency signals are processed by a voting method to obtain a target range gate corresponding to the first intermediate frequency signal; For the transmission signals corresponding to the plurality of the first intermediate frequency signals, determining a target area based on a preset window value and the target range gate; and The candidate signals corresponding to each target gate in the target area are all used as the target signals to be processed, Each of the transmission signals corresponds to a plurality of the first intermediate frequency signals, and each of the first intermediate frequency signals corresponds to a plurality of target sub-intermediate frequency signals.
7. The method according to claim 6, It is characterized in that The target areas include: in, Characterize the target area, p min Characterizes the smallest target range gate among the target range gates, p max represents the largest target range gate among the target range gates, W represents the preset window value, N represents a positive integer, and a represents the target gate.
8. The method according to claim 5, It is characterized in that The summary feature formula includes: Among them, V (f,i) Characterizing the time-speed characteristic data, RDM (f,i,j) represents the corresponding value of the target gate j corresponding to the Doppler gate s in the f-th range Doppler time series data and the target signal to be processed, R j represents the index of the target gate, N represents the first number of points corresponding to the distance dimension fast Fourier transform algorithm, and M represents the second number of points corresponding to the Doppler dimension fast Fourier transform algorithm.
9. The method according to claim 1, It is characterized in that The training process of the trained recognition model includes: Acquire a meta-trained recognition model and sample time-speed feature data corresponding to the user; Processing the sample time speed feature data using the meta-trained recognition model to obtain a sample recognition result; Based on the difference between the sample recognition result and the accurate label, the meta-trained recognition model is trained to obtain the trained recognition model, The meta-trained recognition model is obtained by training with a pre-training sample set, and the sample time-speed feature data is determined by a plurality of sample transmission signals and a sample echo signal corresponding to each transmission signal.
10. An online signature authentication device, It is characterized in that include: A receiving unit, used for receiving echo signals corresponding to multiple transmission signals of the millimeter wave radar during the user signing process; A first processing unit is used to filter and denoise the transmission signal and the echo signal to obtain a plurality of target sub-intermediate frequency signals; A second processing unit is used to process the plurality of target sub-intermediate frequency signals using a fast Fourier transform algorithm to obtain time-speed characteristic data; as well as The third processing unit is used to process the time and speed feature data using the trained recognition model to obtain a target authentication result corresponding to the user.
11. The device according to claim 10, It is characterized in that Also includes: An acquisition unit, used to acquire a meta-trained recognition model and sample time and speed feature data corresponding to the user; A fourth processing unit, used to process the sample time speed feature data using the meta-trained recognition model to obtain a sample recognition result; A training unit is used to train the meta-trained recognition model based on the difference between the sample recognition result and the accurate label to obtain the trained feature extraction model, The meta-trained recognition model is obtained by training with a pre-training sample set, and the sample time-speed feature data is determined by a plurality of sample transmission signals and a sample echo signal corresponding to each transmission signal.
12. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the computer program, the method according to any one of claims 1 to 9 is implemented.
13. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method of any one of claims 1 to 9 is executed.
14. A computer program product comprising a computer program / instructions, It is characterized in that The computer program / instructions, when executed by a processor, implement a method according to any one of claims 1-9.
Citation Information
Patent Citations
Hand gesture motion detection method based on millimeter wave radar
CN109188414A
Air handwritten character recognition method and device of FMCW (Frequency Modulated Continuous Wave) single millimeter wave radar
CN114168058A
Dynamic handwritten signature recognition method and device and electronic equipment
CN114333075A
Online handwritten signature authentication method based on meta-learning
CN114882600A
Target identification system and identification method
CN115453516A
