Lightweight mobile PPG signal identity recognition method and system

Abstract

The invention belongs to the crossing field of biological feature recognition and mobile computing, and provides a lightweight mobile PPG signal identity recognition method and system, and the technical scheme is as follows: training a constructed lightweight feature extraction network based on PPG heartbeat cycle segments to obtain a trained lightweight feature extraction network; the construction process of the lightweight feature extraction network comprises the following steps: enhancing PPG data; the one-dimensional PPG signal after data enhancement is directly input to an input layer of a backbone network, and high-level features are output; performing decoupling analysis on the high-level features output by the backbone network, performing channel-level gating calculation and feature-level gating calculation, and fusing results obtained by calculation to obtain fused semantic features; and performing identity matching on the PPG signal of the user to be detected based on the trained lightweight feature extraction network to obtain an identity matching result. On the premise of ensuring the recognition precision, the consumption of computing resources is remarkably reduced.

Description

Technical Field

[0001] The present invention belongs to the cross - field of biometric recognition and mobile computing, and particularly relates to a lightweight mobile PPG signal identity recognition method and system. Background Technique

[0002] The statements in this part only provide background technical information related to the present invention and do not necessarily constitute prior art.

[0003] With the popularization of intelligent wearable devices, biometric - based identity recognition technology has gradually become an important research direction. Photoplethysmography (PPG) signal, as a kind of bio - electrical signal, reflects physiological characteristics such as heart rate by measuring the change of blood flow on the skin surface, and has the advantages of non - invasiveness, easy acquisition, low cost, etc. Therefore, PPG signals are widely used in fields such as health monitoring, heart rate detection, and blood oxygen saturation monitoring. Currently, the identity recognition method based on PPG signals faces the following challenges: (1) The acquisition of PPG signals is affected by devices and environments, resulting in significant differences between different devices. In addition, changes in the measurement environment (such as light intensity, temperature) and differences in individual physiological states (such as exercise, emotion) may cause significant differences in the waveform, amplitude, frequency, etc. of PPG signals in different acquisition environments and individual physiological states.

[0004] (2) Traditional PPG signal processing and identity recognition methods mainly rely on complex feature extraction and classification algorithms, which usually require high computing resources and energy consumption and are not suitable for application scenarios with limited computing resources such as mobile devices. With the rapid development of deep learning technology, neural networks perform well in pattern recognition and classification tasks, but deep neural networks usually have complex structures and a large number of parameters, which is not conducive to real - time processing on mobile devices. Summary of the Invention

[0005] In order to solve at least one of the above - mentioned technical problems in the background technique, the present invention provides a lightweight mobile PPG signal identity recognition method and system, which can significantly reduce the consumption of computing resources while ensuring the recognition accuracy.

[0006] To achieve the above object, the present invention adopts the following technical solutions: The first aspect of the present invention provides a lightweight mobile PPG signal identity recognition method, including the following steps: Obtain the PPG signal of the user; Remove the motion artifact components in the pre - processed PPG signal; Segment the PPG signal based on removing the motion artifact components, and extract the PPG heartbeat cycle segments; Train the constructed lightweight feature extraction network based on the PPG heartbeat cycle segments to obtain the trained lightweight feature extraction network; wherein, the construction process of the lightweight feature extraction network includes: Enhance the PPG data by adding Gaussian noise to the PPG heartbeat cycle segments; Directly input the one-dimensional PPG signal after data enhancement into the input layer of the backbone network, stack multiple convolutional modules, and output high-level features; perform decoupling analysis on the high-level features output by the backbone network respectively, calculate channel-level gating and feature-level gating respectively, and fuse the calculated results to obtain the fused semantic features; Perform identity matching on the PPG signal of the user to be detected based on the trained lightweight feature extraction network to obtain the identity matching result.

[0007] Further, removing the motion artifact components from the preprocessed PPG signal includes: Construct a mixed signal matrix based on the preprocessed PPG signal; Perform SDV decomposition on the mixed signal matrix, reconstruct the decomposed signal, and obtain the reconstructed mixed signal matrix; Convert the reconstructed mixed signal matrix back to the time-domain signal to obtain the reconstructed signal after removing the motion artifacts.

[0008] Further, the preprocessing process of the PPG signal includes: Perform band-pass filtering on the PPG signal; Perform moving average filtering on the PPG signal after band-pass filtering; Eliminate power interference on the PPG signal after moving average filtering.

[0009] Further, segmenting the PPG signal based on removing the motion artifact components and extracting the PPG heartbeat cycle segments includes: Set the lengths of the sliding window and the overlapping window; Slide the PPG signal with the set overlapping window, and after each slide, record the PPG signal heartbeat cycle segments within the sliding window. All the heartbeat cycle segments are the segmented PPG signal; Normalize the heartbeat cycle segments to obtain the PPG heartbeat cycle segments.

[0010] Further, the backbone network stacks multiple convolutional modules, and each convolutional module can include a convolutional layer, a batch normalization layer, and an activation function layer, expressed as: , Wherein, represents the activation function, represents the batch normalization operation, represents the one-dimensional convolution operation, is the convolution output, is the weight matrix of the convolution kernel, k is the convolution kernel size, is the offset term, and s is the stride.

[0011] Furthermore, the fusion of channel gating and feature-level gating is: , , , , , where, represents element-wise multiplication, is the high-level semantic feature of the PPG signal, is the th element of at the spatial position represents the width of Y, represents the number of channels of is the th element of is the channel feature, is the spatial feature, is the th element of represents channel-level gating, is the feature-level gating, is the sigmoid activation function, and are trainable parameters.

[0012] Furthermore, when training the lightweight feature extraction network, the minimized joint loss function for co-training is: , , , , where, , and are weight parameters, The loss function for the features extracted from the backbone network The loss function for the fused features of channel gating and feature-level gating For And The similarity constraint loss function is performed Represents the feature Of the i Th prediction probability distribution of the i-th sample Represents the feature Of the i True probability distribution of the i-th sample Represents the prediction probability distribution of the i-th sample of the fused feature S i Represents the prediction probability distribution of the i-th sample of the fused feature S Represents the i-th sample of the fused feature S i True probability distribution, and N is the number of samples

[0013] The second aspect of the present invention provides a lightweight mobile PPG signal identity recognition system, including: A signal acquisition module for acquiring the PPG signal of the user; An artifact removal module for removing the motion artifact components in the preprocessed PPG signal; A segmentation module for segmenting based on the PPG signal from which the motion artifact components are removed and extracting PPG heartbeat cycle segments; A lightweight network training module that trains the constructed lightweight feature extraction network based on the PPG heartbeat cycle segments to obtain a trained lightweight feature extraction network; wherein, the construction process of the lightweight feature extraction network includes: enhancing the PPG data by adding Gaussian noise to the PPG heartbeat cycle segments; directly inputting the one-dimensional PPG signal after data enhancement into the input layer of the backbone network, stacking multiple convolutional modules, and outputting high-level features; respectively performing decoupling analysis on the high-level features output by the backbone network, respectively performing channel-level gating and feature-level gating calculations, and fusing the calculated results to obtain fused semantic features; An identity matching module for performing identity matching on the PPG signal of the user to be detected based on the trained lightweight feature extraction network to obtain an identity matching result

[0014] The third aspect of the present invention provides a computer-readable storage medium

[0015] A computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the steps in a lightweight mobile PPG signal identity recognition method as described above

[0016] The fourth aspect of the present invention provides a computer device

[0017] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in a lightweight mobile PPG signal identity recognition method as described above are implemented.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. The present invention decouples and analyzes the high-level features extracted by the backbone network, calculates channel-level gating and feature-level gating for the high-level features respectively, and then fuses them through a dynamic weighting method to form a unified fusion gating mechanism, thereby screening the features. This design effectively reduces the computational complexity while ensuring the feature representation ability through the synergistic effect of the gating mechanism, meeting the lightweight requirement. It can significantly reduce the consumption of computing resources while ensuring the recognition accuracy. This design makes the method particularly suitable for mobile devices with limited resources, thus realizing the efficient and real-time processing of PPG signals.

[0019] 2. Due to the low computational complexity of the lightweight neural network, the present invention has a low operating power consumption on mobile devices, which helps to extend the battery life of the device. This is of great significance for long-term use, especially in applications such as wearable devices with high requirements for battery life, where it has obvious advantages.

[0020] The advantages of the additional aspects of the present invention will be partially given in the following description, partially will become obvious from the following description, or will be understood through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The schematic diagrams in the specification forming a part of the present invention are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation to the present invention.

[0022] Figure 1 is a flowchart of a lightweight mobile PPG signal identity recognition method provided by an embodiment of the present invention; Figure 2 is a schematic diagram of the structure of a lightweight feature extraction network provided by an embodiment of the present invention; Figure 3 is the ROC curve of the existing MobileNetV4 network and the method of the present invention on the MARSH database provided by an embodiment of the present invention; Figure 4 is the ROC curve of the existing MobileNetV4 network and the method of the present invention on the noisy MARSH database provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0024] It should be noted that the following detailed description is exemplary and is intended to provide further illustration of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the technical field to which the present invention belongs.

[0025] 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 invention. 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, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0026] Due to the limited storage capacity and computing resources of mobile devices, the PPG signal identity recognition method for mobile devices should occupy less storage and computing resources. The present invention proposes a lightweight mobile PPG signal identity recognition method. First, the wearable device is used to collect the PPG signal, and the collected PPG signal is preprocessed. Then, the lightweight neural network is used to extract the discriminative features of the PPG signal. Finally, the Euclidean distance is used to match the identity of the query individual. Due to the advantages of high classification accuracy and low memory consumption of the lightweight neural network, it is more suitable for identity recognition in the mobile device environment.

[0027] Embodiment 1 As Figure 1 shown, this embodiment provides a lightweight mobile PPG signal identity recognition method, including the following steps: Step 1: Obtain the PPG signal of the user; In this embodiment, the PPG signal of the user can be collected by a mobile wearable device with a built-in PPG sensor, and then the PPG signal of the user to be detected collected by the mobile wearable device is transmitted to the mobile terminal; Specifically, the type of the mobile wearable device can be selected according to the actual situation, as long as the built-in PPG sensor can collect data; Specifically, when transmitting the PPG signal of the user to be detected collected by the mobile wearable device to the mobile terminal, it can be connected and transmitted through a wireless transmission method such as Bluetooth. Specifically, it includes: Bluetooth pairing: When the user uses it for the first time, Bluetooth pairing needs to be performed between the wearable device and the mobile terminal; this process includes steps such as searching for devices and entering a pairing code to ensure the security of the connection.

[0028] Signal acquisition: The signal acquisition sensor irradiates the skin with green or red light, receives the reflected or transmitted optical signal, and indirectly measures it based on the change in blood volume; Signal transmission: After the acquisition and processing of the PPG signal, the identity feature data packet is sent to the mobile terminal using the Bluetooth protocol. The data packet should contain information such as timestamp, identity feature code, and signal strength.

[0029] Data synchronization: After the mobile terminal receives the data sent by the wearable device, it needs to verify and synchronize the received data, including confirming the integrity and accuracy of the data; Continuous reception: The mobile terminal continuously listens to the Bluetooth transmission from the wearable device. Once a data packet is received, it immediately performs decryption processing to restore the original PPG signal and related information.

[0030] Step 2: Preprocess the obtained PPG signal of the user to be detected to obtain the preprocessed PPG signal; The collected PPG signal is easily interfered by many factors such as external environmental changes and human movements. Therefore, the PPG signal must be preprocessed by an effective method before use.

[0031] The specific preprocessing process includes the following steps: Step 201: Perform band-pass filtering on the PPG signal; In this embodiment, a 4th-order Butterworth band-pass filter is used to filter high-frequency and low-frequency noise; the lower cut-off frequency , the upper cut-off frequency .

[0032] It should be noted that in other embodiments, other band-pass filters or other methods can also be used for filtering processing as long as the filtering effect can be achieved; Step 202: Perform moving average filtering on the PPG signal after band-pass filtering; In this embodiment, a 5-point moving window (window length 100ms) is used to smooth high-frequency spikes.

[0033] Specifically, in the 5-point moving window, the 5 points include: the previous point and 2 points before and after the current point (a total of 5 points) to calculate the average value. The specific calculation formula is: , where represents the signal before sliding smoothing filtering, represents the signal after sliding smoothing filtering, represents the position of the previous signal point, represents the local index within the sliding window.

[0034] Step 203: Eliminate power interference from the PPG signal after sliding average filtering; In this embodiment, a power frequency notch filter is designed to eliminate 50 / 60 Hz power interference in the PPG signal.

[0035] Specifically, an infinite impulse response notch filter with a Q value of 35 is used to calculate the bandwidths of the 50HZ and 60HZ power interferences respectively, and the dual finite difference method is used to calculate the filter coefficients.

[0036] Step 3: Remove the motion artifact components from the preprocessed PPG signal; Specifically, it includes the following steps: Step 301: Construct a mixed signal matrix based on the preprocessed PPG signal; Specifically, by means of a sliding window method, the signal is rearranged into an embedding matrix : , where M = N - L + 1, N is the total length of the signal and L is the window length.

[0037] Step 302: Perform SDV decomposition on the mixed signal matrix, reconstruct the decomposed signal, and obtain the reconstructed mixed signal matrix; Specifically, the decomposition formula is: , where is an L×L orthogonal matrix containing the left singular vectors, is an L×M diagonal matrix containing the singular values, and V is an M×M orthogonal matrix containing the right singular vectors.

[0038] Specifically, assume that the first k singular values and the corresponding singular vectors are selected for signal reconstruction, and the reconstruction formula is: , where and only contain the first k columns, is a diagonal matrix composed of the first k singular values.

[0039] Step 303: Convert the reconstructed matrix back to the time-domain signal to obtain the reconstructed signal after removing the motion artifacts; Specifically, it is achieved by averaging the elements of each column: , j = 1, 2, … M. Among them, is the reconstructed signal after removing motion artifacts, is the i-th element in the j-th sliding window of the reconstructed signal.

[0040] Step 4: Segment the PPG signal based on the PPG signal after removing motion artifact components, and extract PPG heartbeat cycle segments; Since the waveform of the PPG signal collected by the mobile device is greatly affected by external factors and it is difficult to determine the reference point of the PPG signal, in this embodiment, a PPG signal segmentation method based on a sliding window is used to extract PPG heartbeat cycle segments, which specifically includes the following steps: Step 401: Set the lengths of the sliding window and the overlapping window; In this embodiment, the length of the sliding window is set to 1 second, which is greater than one cardiac cycle of the PPG signal, and the length of the overlapping window is 0.4 seconds.

[0041] Step 402: Slide the PPG signal with the set overlapping window. After each slide, record the PPG signal heartbeat cycle segment within the sliding window. All the heartbeat cycle segments are the segmented PPG signal; Step 403: Normalize the heartbeat cycle segments; In order to eliminate the influence caused by different magnitudes of numerical values in different dimensions, in this embodiment, the min-max normalization method is used to normalize the heartbeat cycle segments so that the corresponding numerical values of the heartbeat cycle segments are within the range of [0, 1].

[0042] Step 5: Train the constructed lightweight feature extraction network based on the PPG heartbeat cycle segments to obtain the trained lightweight feature extraction network; Specifically, it includes the following steps: Step 501: Construct a training set and a test set from the extracted PPG heartbeat cycle segments; Step 502: Construct a lightweight feature extraction network. The network schematic diagram is as Figure 2 shown, and specifically includes the following steps: Step 5021: Enhance the PPG data by adding Gaussian noise to the PPG heartbeat cycle segments; In this embodiment, for the PPG signal heartbeat cycle segments , M is the signal length, and the enhanced signal is: , where , is the standard deviation of the noise.

[0043] Step 5022: Directly input the one-dimensional PPG signal after data augmentation into the input layer of the backbone network, stack multiple convolutional modules, and output high-level features. In this embodiment, MobileNetV4 is used as the backbone network feature. MobileNetV4 is modified to change all two-dimensional convolutional operations to one-dimensional convolutional operations.

[0044] Among them, each convolutional module can include a convolutional layer, a batch normalization layer (Batch Normalization), and an activation function layer, which can be expressed as: , Among them, represents the activation function, represents the batch normalization operation, represents the one-dimensional convolutional operation, is the convolutional output, is the weight matrix of the convolutional kernel, k is the convolutional kernel size, is the offset term, and s is the stride.

[0045] Step 5023: Perform decoupling analysis on the high-level features output by the backbone network respectively, calculate channel-level gating and feature-level gating respectively, and fuse the features obtained from the decoupling analysis with the features output by the backbone network to obtain the fused semantic features. The feature decomposition module aims to perform decoupling analysis on the high-level features extracted by the backbone network. This module first calculates channel-level gating and feature-level gating for the high-level features respectively, and then fuses the two through a dynamic weighting method to form a unified fusion gating mechanism, thereby screening the features. This design effectively reduces the computational complexity while ensuring the feature representation ability through the synergistic effect of the gating mechanism, meeting the lightweight requirements.

[0046] Specifically, it includes the following steps: Step 50231: Calculate the channel-level gating: , , Among them, is the th channel of at the spatial position represents the width of Y, represents the number of channels of, and Y represents the features extracted by the backbone network, is the th element in is the channel feature, is the sigmoid activation function, are trainable parameters, represents channel-level gating; Step 50232, calculate the feature-level gating: , , wherein, represents the spatial feature, is a trainable parameter, is the feature-level gating; Obviously, the number of parameters required to calculate is , and the total number of parameters required to calculate the channel gating and the feature-level gating is .

[0047] Step 50233, fuse the high-level feature Y, the channel-level gating and the feature-level gating : , wherein, represents element-wise multiplication, is the high-level semantic feature of the PPG signal.

[0048] Step 503, use the training set to train the constructed lightweight feature extraction network, and use the test set to perform identity matching on the PPG signal; Divide the test sample data set into two parts: the template sample set and the verification sample set. The template sample set generates the registration matching template, and the verification set is used to test the performance of the proposed method.

[0049] In this embodiment, when the feature extraction network is trained, the minimized joint loss function for co-training is: , wherein, , and are weight parameters; The loss function of the feature extracted by the backbone network is: , where N is the number of samples, represents the predicted probability distribution of the -th sample of the feature i , represents the true probability distribution of the -th sample of the feature i ; Loss function of the fused feature S of channel gating and feature-level gating is as follows: , where N is the number of samples, represents the predicted probability distribution of the i -th sample of the fused feature S, represents the true probability distribution of the i -th sample of the fused feature S; To maintain the intra-class similarity of the same individual and minimize the influence of common features in PPG signal identity recognition, and perform similarity constraints: , After co-training, the trained lightweight feature extraction network is finally obtained.

[0050] Specifically, it includes the following steps: Step 5031: Obtain the feature distribution encoding of each user for the template sample set through the trained lightweight feature extraction network, and use it as the registration matching template set ; Step 5032: For the PPG signal verification sample , obtain the feature distribution encoding through the trained lightweight feature extraction network ; Step 5033: Calculate the Euclidean distance between the feature distribution encoding and the registration matching template set , and the identity label of the verification sample is obtained by the following formula: , where, is the indicator function, which only keeps the coefficient corresponding to the i -th class element in the vector non-zero, and all other elements are 0.

[0051] Step 504: Evaluate the identity of the recognition result; In this embodiment, the false acceptance rate (FAR), false rejection rate (FRR), and equal error rate (EER) are used to measure the effectiveness of the proposed identity recognition method; Among them, the false acceptance rate: , The false rejection rate: , The equal error rate: , Wherein, NGRA is the total number of intra-class tests, NIRA is the total number of inter-class tests; NFR and NFA are the numbers of false rejections and false acceptances.

[0052] Step 6: Perform identity matching on the PPG signal of the user to be detected based on the trained lightweight feature extraction network to obtain an identity matching result.

[0053] Experimental verification The present invention is verified on the MARSH database and the noisy MARSH database, as Figure 3 shown is the ROC curve of the existing MobileNetV4 network and the method of the present invention on the MARSH database, Figure 4 shown is the ROC curve of the existing MobileNetV4 network and the method of the present invention on the noisy MARSH database; experiments show that the recognition efficiency of the method proposed by the present invention achieves the best results on both datasets. It shows that the feature decomposition module and co-training strategy designed by the present invention can increase the generalization ability of the model.

[0054] In summary, the present invention has the advantages of wide applicability and easy integration: Wide applicability: The method of the present invention is not only applicable to common mobile devices such as smart phones and smart watches, but also can be extended and applied to other devices with PPG signal acquisition capabilities, such as medical monitoring devices and fitness trackers, etc. Its wide applicability provides new technical support for identity recognition in multiple fields.

[0055] Easy integration: The lightweight neural network has low computational and storage requirements, making the method of the present invention easy to integrate into existing mobile device platforms. Developers can quickly deploy this identity recognition system through simple hardware and software upgrades to enhance the functions and security of the devices.

[0056] Embodiment 2 The present embodiment provides a lightweight mobile PPG signal identity recognition system, including: A signal acquisition module, which is used to acquire the PPG signal of the user; An artifact removal module, which is used to remove the motion artifact components in the preprocessed PPG signal; A segmentation module, which is used to segment the PPG signal based on the PPG signal after removing the motion artifact components and extract PPG heartbeat cycle segments; The lightweight network training module trains the constructed lightweight feature extraction network based on PPG heartbeat cycle segments to obtain the trained lightweight feature extraction network. Among them, the construction process of the lightweight feature extraction network includes: enhancing the PPG data by adding Gaussian noise to the PPG heartbeat cycle segments; directly inputting the one-dimensional PPG signal after data enhancement into the input layer of the backbone network, stacking multiple convolutional modules, and outputting high-level features; respectively performing decoupling analysis on the high-level features output by the backbone network, calculating channel-level gating and feature-level gating respectively, and fusing the calculated results to obtain the fused semantic features. The identity matching module is used to perform identity matching on the PPG signal of the user to be detected based on the trained lightweight feature extraction network to obtain an identity matching result.

[0057] It should be noted that the specific implementation manner of a lightweight mobile PPG signal identity recognition system according to an embodiment of the present invention is similar to the specific implementation manner of a lightweight mobile PPG signal identity recognition method according to an embodiment of the present invention. For details, please refer to the description in the method part. To reduce redundancy, it will not be elaborated here.

[0058] Embodiment III This embodiment provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the steps in a lightweight mobile PPG signal identity recognition method as described above.

[0059] Embodiment IV This embodiment provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the steps in a lightweight mobile PPG signal identity recognition method as described above.

[0060] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can be implemented in the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can be implemented in the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0061] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows 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 the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices produce means for implementing the functions specified in one or more of the flows Figure 1 one or more of the flows and / or blocks Figure 1 or means for implementing the functions specified in one or more of the blocks.

[0062] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory produce a manufactured article including instruction means that implement the functions specified in one or more of the flows Figure 1 one or more of the flows and / or blocks Figure 1 or means for implementing the functions specified in one or more of the blocks.

[0063] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operational steps are performed on the computer or other programmable device to generate a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more of the flows Figure 1 one or more of the flows and / or blocks Figure 1 or means for implementing the functions specified in one or more of the blocks.

[0064] Those of ordinary skill in the art can understand that all or part of the processes in the above-described method embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above-described method embodiments. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0065] The above are only preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included within the protection scope of the present invention.

Claims

1. A lightweight mobile PPG signal identity recognition method, characterized in that, including: Obtain the user's PPG signal; Remove the motion artifact components from the preprocessed PPG signal; Segment based on the PPG signal after removing the motion artifact components, and extract PPG heartbeat cycle segments; Train the constructed lightweight feature extraction network based on the PPG heartbeat cycle segments to obtain the trained lightweight feature extraction network; wherein, the construction process of the lightweight feature extraction network includes: Enhance the PPG data by adding Gaussian noise to the PPG heartbeat cycle segments; Directly input the one-dimensional PPG signal after data augmentation into the input layer of the backbone network, stack multiple convolutional modules, and output high-level features; perform decoupling analysis on the high-level features output by the backbone network respectively, calculate channel-level gating and feature-level gating respectively, and fuse the calculated results to obtain the fused semantic features; Based on the trained lightweight feature extraction network, perform identity matching on the PPG signal of the user to be detected to obtain an identity matching result.

2. The lightweight mobile PPG signal identity recognition method according to claim 1, wherein, Removing the motion artifact components from the preprocessed PPG signal includes: Construct a mixed signal matrix based on the preprocessed PPG signal; Perform SDV decomposition on the mixed signal matrix, reconstruct the decomposed signal, and obtain the reconstructed mixed signal matrix; Convert the reconstructed mixed signal matrix back to the time-domain signal to obtain the reconstructed signal after removing the motion artifact.

3. The lightweight mobile PPG signal identity recognition method according to claim 1, characterized in that, The preprocessing process of the PPG signal includes: Perform band-pass filtering on the PPG signal; Perform moving average filtering on the PPG signal after band-pass filtering; Eliminate power interference on the PPG signal after moving average filtering.

4. A lightweight mobile PPG signal identity recognition method according to claim 1, characterized in that, Segmenting based on the PPG signal after removing the motion artifact components and extracting PPG heartbeat cycle segments includes: Set the lengths of the sliding window and the overlapping window; Slide the PPG signal with the set overlapping window, and record the PPG signal heartbeat cycle segments within the sliding window after each step of sliding. All the heartbeat cycle segments are the segmented PPG signal; Normalize the heartbeat cycle segments to obtain the PPG heartbeat cycle segments.

5. The lightweight mobile PPG signal identity recognition method according to claim 1, characterized in that, The backbone network stacks multiple convolutional modules, and each convolutional module can include a convolutional layer, a batch normalization layer, and an activation function layer, expressed as: ， Among them, represents the activation function, represents the batch normalization operation, represents the one-dimensional convolution operation, is the convolution output, is the weight matrix of the convolution kernel, k is the convolution kernel size, is the offset term, and s is the stride.

6. The lightweight mobile PPG signal identity recognition method according to claim 1, characterized in that, The fusion of channel gating and feature-level gating is: ， ， ， ， ， Among them, represents element-wise multiplication, is the high-level semantic feature of the PPG signal, is the th element at the spatial position of the represents the width of Y, represents the number of channels of , Y represents the feature extracted by the backbone network, is the th element in , is the channel feature, is the spatial feature, is the th element in , represents channel-level gating, is feature-level gating, is the sigmoid activation function, and are trainable parameters.

7. A lightweight mobile PPG signal identity recognition method according to claim 1, characterized in that When training the lightweight feature extraction network, the minimized joint loss function for co-training is: ， ， ， ， in, , and is the weight parameter, is the loss function of the features extracted by the backbone network, is the loss function of the fusion feature of channel gating and feature-level gating, for and Perform similarity constraint loss function, Representative features No. i The predicted probability distribution of samples is Representative features No. i The true probability distribution of samples is Represents the first i The predicted probability distribution of samples is Represents the first i is the true probability distribution of samples, N is the number of samples.

8. A lightweight mobile PPG signal identity recognition system, characterized in that, including: A signal acquisition module, which is used to obtain the user's PPG signal; An artifact removal module, which is used to remove the motion artifact components from the preprocessed PPG signal; A segmentation module, which is used to segment based on the PPG signal after removing the motion artifact components and extract PPG heartbeat cycle segments; A lightweight network training module trains the constructed lightweight feature extraction network based on PPG heartbeat cycle segments to obtain the trained lightweight feature extraction network. Among them, the construction process of the lightweight feature extraction network includes: enhancing the PPG data of the PPG heartbeat cycle segments by adding Gaussian noise; directly inputting the one-dimensional PPG signal after data enhancement into the input layer of the backbone network, stacking multiple convolutional modules, and outputting high-level features; respectively performing decoupling analysis on the high-level features output by the backbone network, respectively calculating channel-level gating and feature-level gating, and fusing the calculated results to obtain the fused semantic features. An identity matching module is used to perform identity matching on the PPG signal of the user to be detected based on the trained lightweight feature extraction network to obtain an identity matching result.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by a processor, it implements the steps in a lightweight mobile PPG signal identity recognition method according to any one of claims 1-7.

10. A computer device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in a lightweight mobile PPG signal identity recognition method according to any one of claims 1-7.

Images