Identity recognition method and system based on biological signal invariant representation learning

Abstract

The invention provides an identity recognition method and system based on biological signal invariant representation learning, and relates to the technical field of artificial intelligence and pattern recognition, and the method comprises the steps: converting waveform heart beat data into a graph structure, and extracting a topological feature graph of the graph structure through a graph structure feature extraction unit; the method comprises the following steps of: performing multi-granularity semantic correlation evaluation on waveform heart beat data through a multi-granularity semantic converter, establishing a cross-dimensional dependency relationship by adopting an attention mechanism to obtain a score feature map, inputting the score feature map and a topological feature map into a feature decoupling and recombination module, decomposing the score feature map and the topological feature map into identity sensitive invariant feature representation and physiological variable feature representation, and performing feature representation and physiological variable feature representation; and recombining the invariant feature representation and the out-of-order physiological variable feature representation to generate a recombined feature representation, inputting the recombined feature into a loss function to obtain a predicted value, and performing identity matching verification. According to the invention, efficient identity recognition is realized through cloud collaboration and model structure architecture.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence and pattern recognition technology, and in particular to an identity recognition method and system based on biological signal invariant representation learning. Background Art

[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art.

[0003] With the rapid growth of mobile Internet users and mobile devices, new social interaction paradigms such as mobile-based digital payment, instant messaging, and remote office are increasing. While this digital transformation improves service efficiency, it also poses unprecedented security challenges to device identity authentication mechanisms.

[0004] The current mainstream biometric authentication system (including fingerprint recognition, iris verification and 3D face modeling technology) and the traditional password verification method have exposed significant security vulnerabilities in practice. In recent years, with the popularity of smart wearable devices, smart bracelets have become an important carrier for daily health monitoring, sports management and security authentication. The built-in photoplethysmography (PPG) sensor of the wearable bracelet can collect the user's PPG signal in real time for identity recognition. Compared with traditional biometrics (such as fingerprints and facial recognition), PPG signals have significant technical advantages: on the one hand, it obtains blood vessel pulsation information through the principle of light reflection, which is difficult to forge; on the other hand, PPG signal acquisition does not require complex hardware, is low-cost and highly compatible, and can be seamlessly integrated into various existing wearable devices. With the rapid development of deep learning methods, various deep learning methods have shown good performance in actual identity recognition.

[0005] However, in practical applications, PPG signal identification methods still face the following problems: 1) The PPG signal collected by smart bracelets is easily affected by motion artifacts. The frequency of motion artifacts coincides with the frequency of the PPG signal itself. It is difficult to eliminate the influence of motion artifacts using existing noise elimination methods. 2) The distribution of PPG signals is easily affected by changes in physiological state, environmental noise, and device differences, causing the distribution of PPG signals to drift over time (such as user heart rate variability, changes in vascular elasticity, or sensor aging). This type of distribution drift will significantly reduce the generalization ability of the model. Especially in open dynamic scenes, the PPG signal is affected by motion artifacts and distribution drift, and the identity recognition effect is low, which limits the actual implementation of PPG signal identity recognition technology. Summary of the invention

[0006] In order to solve the above problems, the present disclosure proposes an identity recognition method and system based on biological signal invariant representation learning, extracts pure PPG waveforms synchronized with the heart cycle based on preprocessing of adaptive motion noise separation, and learns PPG signal identity recognition model based on invariant representation of graph neural network, converts PPG signal into graph structure data model through visual graph topology mapping mechanism, and constructs graph space representation matrix. The feature dimension contribution is quantified in the latent semantic space through adaptive attention mechanism, and the graph space representation is decomposed into identity sensitive invariant feature set and physiological variable feature set, and the model robustness is improved through feature recombination strategy.

[0007] According to some embodiments, the present disclosure adopts the following technical solutions: The identity recognition method based on biological signal invariant representation learning includes: Obtain the user's pulse wave PPG signal, pre-process it based on the adaptive motion noise separation algorithm, segment and extract the waveform heartbeat data synchronized with the heart cycle; The waveform heartbeat data is input into the invariant representation learning PPG signal identity recognition model based on the graph neural network, and the user's identity recognition result is output; Among them, after the waveform heartbeat data is input into the invariant representation learning PPG signal identity recognition model based on graph neural network, it enters the graph structure feature extraction unit and the multi-granularity semantic converter unit for feature extraction respectively; after the waveform heartbeat data enters the graph structure feature extraction unit, the waveform heartbeat data is converted into a graph structure, and then the topological feature map of the graph structure is extracted by the graph structure feature extraction unit; the waveform heartbeat data is evaluated for multi-granularity semantic relevance through the multi-granularity semantic converter, and the time-sensitive attention mechanism is used to establish cross-dimensional dependencies to obtain a score feature map, and then the score feature map and the topological feature map are input into the feature decoupling and reorganization module, first decomposed into identity-sensitive invariant feature representation and physiological variable feature representation, and then the invariant feature representation and the disordered physiological variable feature representation are reorganized to generate a recombined feature representation, and the recombined feature is input into the loss function to obtain the predicted value for identity matching verification.

[0008] According to some embodiments, the present disclosure adopts the following technical solutions: The identity recognition system based on bio-signal invariant representation learning includes: The signal acquisition module is used to acquire the user's pulse wave PPG signal, pre-process it based on the adaptive motion noise separation algorithm, and segment and extract the waveform heartbeat data synchronized with the heart cycle; An identity recognition module is used to input the waveform heartbeat data into the invariant representation learning PPG signal identity recognition model based on the graph neural network, and output the user's identity recognition result; Among them, after the waveform heartbeat data is input into the invariant representation learning PPG signal identity recognition model based on graph neural network, it enters the graph structure feature extraction unit and the multi-granularity semantic converter unit for feature extraction respectively; after the waveform heartbeat data enters the graph structure feature extraction unit, the waveform heartbeat data is converted into a graph structure, and then the topological feature map of the graph structure is extracted by the graph structure feature extraction unit; the waveform heartbeat data is evaluated for multi-granularity semantic relevance through the multi-granularity semantic converter, and the time-sensitive attention mechanism is used to establish cross-dimensional dependencies to obtain a score feature map, and then the score feature map and the topological feature map are input into the feature decoupling and reorganization module, first decomposed into identity-sensitive invariant feature representation and physiological variable feature representation, and then the invariant feature representation and the disordered physiological variable feature representation are reorganized to generate a recombined feature representation, and the recombined feature is input into the loss function to obtain the predicted value for identity matching verification.

[0009] According to some embodiments, the present disclosure adopts the following technical solutions: A computer program product includes a computer program, which, when executed by a processor, implements the identity recognition method based on biological signal invariant representation learning.

[0010] According to some embodiments, the present disclosure adopts the following technical solutions: A non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by a processor, the identity recognition method based on biological signal invariant representation learning is implemented.

[0011] According to some embodiments, the present disclosure adopts the following technical solutions: An electronic device comprises: a processor, a memory and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device executes the identity recognition method based on biological signal invariant representation learning.

[0012] Compared with the prior art, the present invention has the following beneficial effects: The disclosed identity recognition method based on invariant representation learning of biological signals adopts the PPG sensor built into the smart bracelet to collect the user's pulse wave signal in real time, and transmits the signal to the user's mobile terminal using the Bluetooth protocol. The mobile terminal uploads the data to the remote server through an encrypted channel for identity recognition. The smart mobile terminal establishes a two-way communication link with the smart bracelet through the low-power Bluetooth (BLE) protocol to achieve low-latency and high-fidelity transmission of the PPG signal. In the connection initialization stage, the mobile terminal completes two-way authentication of the device identity based on the dynamic key negotiation protocol, and negotiates an adaptive clock synchronization mechanism to ensure that the sampling timing is strictly aligned with the bracelet sensor. The PPG signal can be collected by the smart bracelet in real time, providing a dynamic, continuous and covert verification method for identity recognition, which is more suitable for identity recognition in mobile device environments, and will also promote the in-depth application of mobile bracelets in financial payment, medical health and other fields, laying the foundation for the development of multimodal biometric recognition systems.

[0013] The identity recognition method based on biosignal invariant representation learning disclosed in the present invention designs a preprocessing algorithm based on adaptive motion noise separation, eliminates high-frequency and low-frequency noise, ambient light interference and baseline drift through multi-channel signal fusion and time-frequency domain filtering, and uses an improved adaptive empirical mode decomposition algorithm to extract a pure PPG waveform synchronized with the cardiac cycle, eliminating motion artifacts and signal distortion caused by limb movement in dynamic scenes.

[0014] The disclosed identity recognition method based on invariant representation learning of biological signals proposes an invariant representation learning PPG signal identity recognition model of graph neural network, which includes three core components: graph structure feature extraction unit (GFEM), multi-granularity semantic converter (MTM) and feature decoupling and reorganization module (SMM). First, the PPG signal is converted into a graph structure data model through the visible graph topology mapping mechanism, and the GFEM unit is used to capture its topological features and construct a graph space representation. At the same time, the MTM component implements multi-granularity semantic relevance evaluation on the graph space features, and the feature dimension contribution is quantified in the latent semantic space through the adaptive attention mechanism. Finally, the SMM module uses a feature decoupling algorithm to decompose the graph space representation into identity-sensitive invariant features and physiological variable features, and improves the model robustness through a feature reorganization strategy; the solution achieves efficient recognition through the end-cloud collaborative architecture and the recognition model structure, while ensuring privacy and security, significantly improving the accuracy and stability of PPG identity recognition in complex scenarios, and can be extended to high-security demand scenarios such as mobile payment and medical data authorization. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The accompanying drawings constituting a part of the present disclosure are used to provide a further understanding of the present disclosure. The illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation on the present disclosure.

[0016] Figure 1 It is a flowchart of the application of the identity recognition method based on biological signal invariant representation learning according to the embodiment of the present disclosure; Figure 2 Schematic diagram of mobile identity recognition of PPG signal of the smart bracelet according to an embodiment of the present disclosure; Figure 3 A structural diagram of a PPG signal identity recognition model for learning an invariant representation of a graph neural network according to an embodiment of the present disclosure; Figure 4 It is a schematic diagram of the structure of a graph structure feature extraction unit according to an embodiment of the present disclosure; Figure 5 A schematic diagram of the structure of a multi-granularity semantic converter according to an embodiment of the present disclosure; Figure 6 It is a schematic diagram of the characteristic decoupling and reorganization module structure of an embodiment of the present disclosure. DETAILED DESCRIPTION

[0017] The present disclosure is further described below in conjunction with the accompanying drawings and embodiments.

[0018] It should be noted that the following detailed descriptions are all illustrative and are intended to provide further explanation of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present disclosure belongs.

[0019] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present disclosure. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0020] Example 1 In one embodiment of the present disclosure, a method for identity recognition based on biological signal invariant representation learning is provided, and the method steps include: Step 1: Obtain the user's pulse wave PPG signal, pre-process it based on the adaptive motion noise separation algorithm, segment and extract the waveform heartbeat data synchronized with the heart cycle; Step 2: Input the waveform heartbeat data into the invariant representation learning PPG signal identity recognition model based on the graph neural network, and output the user's identity recognition result; Among them, after the waveform heartbeat data is input into the invariant representation learning PPG signal identity recognition model based on graph neural network, it enters the graph structure feature extraction unit and the multi-granularity semantic converter unit for feature extraction respectively; after the waveform heartbeat data enters the graph structure feature extraction unit, the waveform heartbeat data is converted into a graph structure, and then the topological feature map of the graph structure is extracted by the graph structure feature extraction unit; the waveform heartbeat data is evaluated for multi-granularity semantic relevance through the multi-granularity semantic converter, and the time-sensitive attention mechanism is used to establish cross-dimensional dependencies to obtain a score feature map, and then the score feature map and the topological feature map are input into the feature decoupling and reorganization module, first decomposed into identity-sensitive invariant feature representation and physiological variable feature representation, and then the invariant feature representation and the disordered physiological variable feature representation are reorganized to generate a recombined feature representation, and the recombined feature is input into the loss function to obtain the predicted value for identity matching verification.

[0021] As an embodiment, the specific implementation process of the identity recognition method based on biological signal invariant representation learning disclosed in the present invention is as follows: Step 1: Obtain the user's pulse wave PPG signal, pre-process it based on the adaptive motion noise separation algorithm, segment and extract the waveform heartbeat data synchronized with the heart cycle; Specifically, firstly, the user's pulse wave PPG signal is collected by a wearable device, which can be a mobile bracelet. The user's pulse wave signal is collected in real time through the built-in PPG sensor of the mobile bracelet, and the signal is transmitted to the mobile terminal using the Bluetooth protocol. The mobile terminal can be a smart phone. The smart phone establishes a two-way communication link with the smart bracelet through the low-power Bluetooth (BLE) protocol. The mobile terminal uploads the data to the remote server through an encrypted channel for identity recognition, and finally returns the recognition result to the mobile terminal to complete the identity authentication. Low-latency and high-fidelity transmission of PPG signals is achieved. In the connection initialization stage, the mobile terminal completes the two-way authentication of the device identity based on the dynamic key negotiation protocol, and negotiates the adaptive clock synchronization mechanism to ensure that the sampling timing is strictly aligned with the bracelet sensor.

[0022] Furthermore, the collected user's pulse wave PPG signal is divided into a training set and a test set.

[0023] Then, the collected PPG signal is preprocessed, and the preprocessing includes two stages. The first stage eliminates high-frequency and low-frequency noise, ambient light interference and baseline drift, and the second stage eliminates motion artifacts. The specific process is as follows: Phase 1: Eliminate high-frequency and low-frequency noise, ambient light interference, and baseline drift.

[0024] 1) Eliminate baseline drift. Remove low-frequency baseline offsets through moving average or polynomial fitting, detrend and normalize the raw PPG signal, and eliminate baseline drift.

[0025] 2) Initialize the filter coefficients and step size factor μ , the cutoff frequency of the low-pass filter is set to 50Hz, and the cutoff frequency of the high-pass filter is set to 0.5Hz.

[0026] 3) Design a finite impulse response filter whose output is expressed as:

[0027] in, y [ n ] is the output signal of the finite impulse response filter, x[n] is the input signal, w k [n] is the adaptive coefficient of the filter, M is the order of the filter. The filter coefficient It will adjust adaptively as the input signal changes to minimize the error.

[0028] 4) According to the principle of minimizing the sum of squared errors, the coefficients of the filter are gradually updated. The update formula is:

[0029] in, e [ n ] = d [ n ] - y [ n ] is the error signal, d [ n ] is the expected signal (usually a reference signal), is the step size factor, which controls the convergence speed of the algorithm.

[0030] 5) Use the finite impulse response filter to design low-pass filters and high-pass filters. The low-pass filter is used to remove high-frequency noise (such as noise above 50Hz), and the high-pass filter is used to remove low-frequency noise (such as baseline drift below 0.5Hz).

[0031] 6) Combine low pass filter and high pass filter into a band filter.

[0032] 7) Adjust the filter coefficients at each sampling point using the least mean square error algorithm.

[0033] Stage 2: Eliminating motion artifacts.

[0034] 1) Adaptive decomposition. The improved adaptive empirical mode decomposition algorithm is used to decompose the signal y after the first stage of denoising. Perform empirical mode decomposition to obtain several intrinsic mode functions (IMFs):

[0035] in, For the i A modal function, is the residual item.

[0036] 2) Time-frequency analysis. Perform time-frequency analysis on each modal function and calculate its spectrum characteristics. ,calculate Integrate in the low frequency band [0,5], the specific formula is:

[0037] if Above a predetermined threshold , then determine The component contains motion artifacts, otherwise it means that is a valid PPG signal.

[0038] 3) Motion artifact removal and signal reconstruction. After removing the motion artifact, the artifact-free PPG signal is obtained:

[0039] in, S For a valid PPG signal gather.

[0040] 4) The Pan-Tompkins algorithm is used to accurately detect the R peak in the PPG signal and segment the PPG signal into heart beats. The segmented heart beats are used as the input of the PPG signal identity recognition model based on the invariant representation learning of the graph neural network.

[0041] Specifically, the Pan-Tompkins algorithm is used to accurately detect the R peak in the PPG signal and segment the PPG signal into heart beats as follows: S1: Preprocessing, including: using the first-order difference to calculate the signal slope and enhance the fast rising edge characteristics. The calculation formula is: y[n] = y[n+1]- y[n-1]. Square the differential signal point by point to highlight the high-frequency components and eliminate negative interference.

[0042] Furthermore, a moving window integration is performed, including integrating the squared signal using a 150 ms window (assuming a sampling rate of 100 Hz, corresponding to 15 sampling points) to generate a smooth envelope and aggregate the pulse energy.

[0043] S2: Perform feature enhancement.

[0044] S3: Adaptive threshold detection, including: setting signal threshold = 0.75 × noise threshold, noise threshold = 0.25 × signal threshold. The average amplitude of the initial 2-second signal is calculated as the noise threshold, and the absolute refractory period of 200ms is set to avoid pseudo-peak interference. The threshold is updated every 0.5 seconds.

[0045] S4: R wave localization, including: searching for the local maximum within a ±100 ms window of the original PPG signal, excluding abnormal peaks with adjacent intervals <300 ms or >1500 ms, and accurately locating the systolic peak.

[0046] S5: heart beat segmentation, including: taking the R wave peak as the reference, taking the first 200ms and the last 400ms to form the heart beat segment.

[0047] S6: Length normalization, including: unifying each heart beat to a fixed length through cubic spline interpolation.

[0048] Step 2: Input the waveform heartbeat data into the invariant representation learning PPG signal identity recognition model based on the graph neural network, and output the user's identity recognition prediction label; Specifically, the graph neural network-based invariant representation learning PPG signal identity recognition model includes a graph structure feature extraction unit (GFEM), a multi-granularity semantic transformer (MTM) and a feature decoupling and reorganization module (SMM).

[0049] Among them, the graph-based feature extraction module (GFEM) consists of two parts: the visual graph construction module and the feature extraction module. Figure 4 shown.

[0050] The visibility graph construction module converts ECG data into a visibility graph, and the feature extraction module is able to extract feature representations of its topological features.

[0051] The Multi-scale Transformer Module (MTM) includes a Multi-scale Extraction Module (MEM), a Bidirectional Transformer (BiTransformer) module, a Multi-Feature Attention (MFA) module, and a linear layer, such as Figure 5 shown.

[0052] The feature decoupling and recombination module includes a decoupling module and a recombination module. The feature decoupling and recombination module is responsible for separating and mixing invariant features and false features, and capturing the invariant relationship between features and labels through a self-supervised learning mechanism. Figure 6 shown.

[0053] As an embodiment, the specific process of training the PPG signal identity recognition model based on the invariant representation learning of the graph neural network is as follows: Step 1: After the waveform heartbeat data of the original PPG signal is input into the invariant representation learning PPG signal identity recognition model based on graph neural network, it enters the graph structure feature extraction unit and the multi-granularity semantic converter unit for processing, and the topological feature map and score feature map are extracted respectively. The specific process is as follows: 1) After the waveform heartbeat data is input into the PPG signal identity recognition model based on the invariant representation learning of the graph neural network, the waveform heartbeat data enters the graph structure feature extraction unit (GFEM), and the graph structure conversion of the waveform heartbeat data is performed using the visible graph construction module of the graph structure feature extraction unit (GFEM). Given a waveform heartbeat data , Indicated by The transformed visible graph, represents a node set, Specifically, consider the amplitude point in each heartbeat as a node in the visible graph. .node and nodes There is an edge between , if and only if the following conditions are met:

[0054] in, represents the interval of sampling points, Represents the maximum slope between the target node and its right adjacent node. The initial setting is ,In each iteration, the slope between the target node and the neighbor nodes is updated.

[0055] Furthermore, the visual graph structure data is input into the feature extraction module in the graph structure feature extraction unit (GFEM), which includes four graph isomorphism layers and two linear layers. The graph isomorphism layer captures the structure and features of the visual graph by iteratively aggregating information from nodes and edges. The role of the linear layer is to perform a linear transformation on the input features to obtain the topological feature map of the graph structure. The specific representation is as follows:

[0056]

[0057]

[0058]

[0059]

[0060]

[0061] in, Indicates A graph isomorphism layer, Used to realize feature splicing, Represents the input features Perform a linear transformation, Represented as a topological feature map of the PPG signal.

[0062] 2) After the waveform heartbeat data of the original PPG signal is input into the multi-granularity semantic transformer (MTM), the multi-granularity semantic transformer (MTM) designed in the present invention realizes the dynamic evaluation of the contribution of feature dimensions to identity discrimination in the latent feature space through a hierarchical semantic association analysis framework. The multi-granularity semantic transformer (MTM) consists of three-order processing units: Multi-scale Extraction Module (MEM), Bidirectional Transformer (BiTransformer) module, and Multi-Feature Attention (MFA) module. The specific process is as follows: (1) The waveform heartbeat data of the original PPG signal first enters the multi-granularity feature extraction module and is input into three convolutions with different kernel sizes respectively. Then, the feature maps of different scales are fused and formally expressed as:

[0063]

[0064]

[0065]

[0066] in, It is a multi-granularity feature.

[0067] (2) The bidirectional transformer module performs reverse mapping on the obtained multi-granularity feature map to obtain the reverse mapped feature map. The multi-granularity feature map and its reverse mapped feature map are respectively input into the BiTransformer model for encoding. The outputs are respectively expressed as:

[0068]

[0069] in, express Transformer encoding layer, Represents the reverse mapping method, and They represent the outputs corresponding to the multi-granularity feature maps and the reverse-mapped feature maps respectively.

[0070] (3) The Multi-Feature Attention (MFA) module adds the output of the multi-scale feature map and its reverse-mapped feature map as follows:

[0071] Will The key channel features are emphasized in the input channel attention mechanism, and the output features are defined as follows:

[0072] Next, Input the linear layer to further process the features and obtain the score feature map which is the same as the separated features :

[0073] in, Represents the Sigmoid activation function, the purpose is to Each value in is constrained to be between (0,1), Represents a linear layer.

[0074] 3) The score feature map and the topological feature map are simultaneously input into the feature decoupling and recombination module, and decomposed into identity-sensitive invariant feature representation and physiological variable feature representation. The invariant feature representation and the disordered physiological variable feature representation are then recombined to generate a recombined feature representation. The recombined feature is input into the loss function to obtain the predicted value for identity matching verification.

[0075] Specifically, after obtaining the score feature map And the topological feature map, based on Separate graph-level representation , to capture both invariant representations (identity-sensitive invariant feature representations) and spurious representations (physiologically variable feature representations):

[0076]

[0077] in, Represents unchanged representation, Indicates false representation, Represents element-wise product.

[0078] Furthermore, the invariant representation and spurious representations of scrambled order Then, the unchanged representation and false representation Connect them to generate a new connection representation .

[0079] Furthermore, the learning objective loss function can be defined as:

[0080]

[0081]

[0082]

[0083]

[0084]

[0085] in, represents the predicted label, represents the true label, represents the cross entropy loss function, and Respectively and The predicted value of .

[0086] in, and is a hyperparameter used to control the weights of the invariant prediction loss and the invariant hybrid loss.

[0087] Step 3: Divide all heartbeat sample data into three parts: training set, template set and validation set. The template set generates a registered matching template, and the validation set is used to test the performance of the proposed method. The present disclosure first uses the training set to train the invariant representation learning PPG signal identity recognition model based on the graph neural network, and then uses the trained model to perform identity matching on the test set. The specific process is as follows: (1) Model training. The training PPG signal set is used to train the PPG signal identity recognition model based on the graph neural network’s invariant representation learning.

[0088] (2) Construct a template set. The feature extraction model based on the trained graph neural network-based invariant representation learning is used to obtain the feature distribution encoding of each user for the template sample set, and this is used as the registration matching template set. . (3) PPG signal verification sample , the feature distribution encoding is obtained by learning the PPG signal identity recognition model based on the trained graph neural network invariant representation .

[0089] (4) Calculation Matching template set with registration The Euclidean distance between the validation samples The identity label is obtained by the following formula:

[0090]

[0091] in, express With template set The Euclidean distance between Representation and The closest template is the sample identity tag.

[0092] Step 4: Matching result evaluation. Finally, the false recognition rate (FAR), rejection rate (FRR) and equal error rate (EER) are calculated to measure the effectiveness of the proposed identification method. The specific formula is as follows: False recognition rate:

[0093] Rejection rate:

[0094] Equal error rate:

[0095] Among them, NGRA is the total number of intra-class tests, NIRA is the total number of inter-class tests; NFR and NFA are the number of false rejections and false acceptances.

[0096] Example 2 In one embodiment of the present disclosure, an identity recognition system based on bio-signal invariant representation learning is provided, comprising: The signal acquisition module is used to acquire the user's pulse wave PPG signal, pre-process it based on the adaptive motion noise separation algorithm, and segment and extract the waveform heartbeat data synchronized with the heart cycle; An identity recognition module is used to input the waveform heartbeat data into the invariant representation learning PPG signal identity recognition model based on the graph neural network, and output the user's identity recognition result; Among them, after the waveform heartbeat data is input into the invariant representation learning PPG signal identity recognition model based on graph neural network, it enters the graph structure feature extraction unit and the multi-granularity semantic converter unit for feature extraction respectively; after the waveform heartbeat data enters the graph structure feature extraction unit, the waveform heartbeat data is converted into a graph structure, and then the topological feature map of the graph structure is extracted by the graph structure feature extraction unit; the waveform heartbeat data is evaluated for multi-granularity semantic relevance through the multi-granularity semantic converter, and the time-sensitive attention mechanism is used to establish cross-dimensional dependencies to obtain a score feature map, and then the score feature map and the topological feature map are input into the feature decoupling and reorganization module, first decomposed into identity-sensitive invariant feature representation and physiological variable feature representation, and then the invariant feature representation and the disordered physiological variable feature representation are reorganized to generate a recombined feature representation, and the recombined feature is input into the loss function to obtain the predicted value for identity matching verification.

[0097] As an embodiment, the application process of the identity recognition method based on biological signal invariant representation learning in the identity recognition system based on biological signal invariant representation learning includes: Step 1: Use a smart bracelet to collect PPG signals, and a mobile phone obtains the collected PPG signals through Bluetooth or other means; Step 2: Preprocess the PPG signal.

[0098] Step 3: Use the preprocessed PPG signal set to learn the PPG signal identity recognition model based on the invariant representation of the graph neural network for identity recognition; Step 4: Use the template to perform identity matching on the PPG signal recognition results.

[0099] Step 5: Perform identity evaluation on the recognition results.

[0100] Example 3 In one embodiment of the present disclosure, a computer program product is provided, including a computer program, which, when executed by a processor, implements the identity recognition method based on biological signal invariant representation learning.

[0101] Example 4 In one embodiment of the present disclosure, a non-transitory computer-readable storage medium is provided, wherein the non-transitory computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by a processor, the identity recognition method based on biological signal invariant representation learning is implemented.

[0102] Example 5 In one embodiment of the present disclosure, an electronic device is provided, including: a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory, so that the electronic device executes the identity recognition method based on biological signal invariant representation learning.

[0103] The present disclosure 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 the present disclosure. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks 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 generate 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.

[0104] 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.

[0105] Although the above describes the specific implementation methods of the present disclosure in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present disclosure. Technical personnel in the relevant field should understand that on the basis of the technical solution of the present disclosure, various modifications or variations that can be made by those skilled in the art without creative work are still within the scope of protection of the present disclosure.

Claims

1. An identity recognition method based on biological signal invariant representation learning, characterized in that: include: Obtain the user's pulse wave PPG signal, pre-process it based on the adaptive motion noise separation algorithm, segment and extract the waveform heartbeat data synchronized with the heart cycle; The waveform heartbeat data is input into the invariant representation learning PPG signal identity recognition model based on the graph neural network, and the user's identity recognition result is output; Among them, after the waveform heartbeat data is input into the invariant representation learning PPG signal identity recognition model based on the graph neural network, it enters the graph structure feature extraction unit and the multi-granularity semantic converter unit for feature extraction; after the waveform heartbeat data enters the graph structure feature extraction unit, the waveform heartbeat data is converted into a graph structure, and then the topological feature graph of the graph structure is extracted by the graph structure feature extraction unit; The waveform heartbeat data is evaluated for multi-granularity semantic relevance through a multi-granularity semantic converter, and the temporal sensitive attention mechanism is used to establish cross-dimensional dependencies to obtain a score feature map. The score feature map and the topological feature map are then input into the feature decoupling and reorganization module, first decomposed into identity-sensitive invariant feature representation and physiologically variable feature representation, and then the invariant feature representation and the disordered physiologically variable feature representation are recombined to generate a recombined feature representation. The recombined feature is input into the loss function to obtain the predicted value for identity matching verification.

2. The identity recognition method based on biological signal invariant representation learning as claimed in claim 1, characterized in that: The preprocessing is performed based on an adaptive motion noise separation algorithm. The preprocessing process includes two stages. In the first stage, the low-frequency baseline offset is removed by moving average or polynomial fitting, the original pulse wave PPG signal is detrended and normalized, a low-pass filter is used to remove high-frequency noise, and a high-pass filter is used to remove low-frequency noise.

3. The identity recognition method based on biological signal invariant representation learning as claimed in claim 2, characterized in that: In the second stage, an improved adaptive empirical mode decomposition algorithm is used to perform empirical mode decomposition on the denoised signal in the first stage to obtain several intrinsic mode functions. Time-frequency analysis is performed on each intrinsic mode function, and its spectral characteristics are calculated and artifacts are removed to obtain the artifact-free PPG signal. The Pan-Tompkins algorithm is used to detect the R peak in the artifact-free PPG signal, and the signal is segmented and the waveform heartbeat data synchronized with the cardiac cycle is extracted.

4. The identity recognition method based on biological signal invariant representation learning as claimed in claim 1, characterized in that: After the waveform heartbeat data is input into the graph structure feature extraction unit, the visible graph construction module of the graph structure feature extraction unit is used to regard the amplitude point in each waveform heartbeat as a node in the visible graph, and the waveform heartbeat data is converted into a visible graph structure; then the topological feature graph is extracted by the feature extraction module of the graph structure feature extraction unit. The feature extraction module includes four graph isomorphism layers and two linear layers. The graph isomorphism layer obtains the structure and features of the visible graph by iteratively aggregating the information of the nodes and edges from the visible graph structure. The function of the linear layer is to perform linear transformation on the input features to obtain the topological feature graph of the graph structure.

5. The identity recognition method based on biological signal invariant representation learning as claimed in claim 1, characterized in that: The multi-granularity semantic converter comprises a multi-granularity feature extraction module, a bidirectional converter module and a multi-granularity attention module. The multi-granularity feature extraction module comprises three convolutions with different kernels, extracts feature maps of different scales, and then fuses the feature maps of different features to obtain multi-granularity features. The bidirectional converter module reversely maps the obtained multi-granularity feature map to obtain a reverse mapped feature map, and respectively inputs the multi-granularity feature map and its reverse mapped feature map into the BiTransformer model for encoding, and weights the respectively obtained encoded representations using the multi-granularity attention module to obtain a score feature map.

6. The identity recognition method based on biological signal invariant representation learning as claimed in claim 1, characterized in that: After obtaining the score feature map and the topological feature map, based on the separated graph-level representation, the feature decoupling and reorganization module is used to capture the invariant representation and the false representation, and the invariant representation and the shuffled false representation are connected to generate another new reorganized feature representation.

7. An identity recognition system based on learning invariant representations of biological signals, characterized in that: include: The signal acquisition module is used to acquire the user's pulse wave PPG signal, pre-process it based on the adaptive motion noise separation algorithm, and segment and extract the waveform heartbeat data synchronized with the heart cycle; An identity recognition module is used to input the waveform heartbeat data into the invariant representation learning PPG signal identity recognition model based on the graph neural network, and output the user's identity recognition result; Among them, after the waveform heartbeat data is input into the invariant representation learning PPG signal identity recognition model based on the graph neural network, it enters the graph structure feature extraction unit and the multi-granularity semantic converter unit for feature extraction; after the waveform heartbeat data enters the graph structure feature extraction unit, the waveform heartbeat data is converted into a graph structure, and then the topological feature graph of the graph structure is extracted by the graph structure feature extraction unit; The waveform heartbeat data is evaluated for multi-granularity semantic relevance through a multi-granularity semantic converter, and the temporal sensitive attention mechanism is used to establish cross-dimensional dependencies to obtain a score feature map. The score feature map and the topological feature map are then input into the feature decoupling and reorganization module, first decomposed into identity-sensitive invariant feature representation and physiologically variable feature representation, and then the invariant feature representation and the disordered physiologically variable feature representation are recombined to generate a recombined feature representation. The recombined feature is input into the loss function to obtain the predicted value for identity matching verification.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the identity recognition method based on biological signal invariant representation learning according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium, characterized in that: The non-transitory computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by the processor, the identity recognition method based on biological signal invariant representation learning as described in any one of claims 1-6 is implemented.

10. An electronic device, characterized in that: include: A processor, a memory and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device executes the identity recognition method based on biological signal invariant representation learning as described in any one of claims 1-6.

Images