Method, device, electronic device and storage medium for identifying identity using magnetic cardiogram signals

Abstract

The present invention provides a method, device, electronic device and storage medium for identifying the identity of a cardiomagnetic signal, and relates to the technical field of cardiomagnetic signal identification. In an embodiment of the present invention, the measured cardiomagnetic signal is first subjected to noise reduction processing; the waveform is then converted into a time-frequency matrix; and the data is reorganized into a target matrix according to the spatial position distribution; the cardiomagnetic signal recognition model is used for identification, and the closest identity is matched. In an embodiment of the present invention, a deep learning model (cardiomagnetic signal recognition model) is used to identify the target matrix, which can make full use of the spatial correlation information of the cardiomagnetic signal, and because the deep learning model is used, the process of distinguishing the characteristic information of the cardiomagnetic signals of different users can be automatically optimized, which will not be affected by the subjective factors of the operator and is more practical. At the same time, a convolutional neural network structure specialized for identifying cardiomagnetic information is used to make the verification result more accurate.

Description

Technical Field

[0001] Embodiments of the present invention relate to the technical field of magnetocardiographic signal recognition, and in particular to a magnetocardiographic signal identity recognition method, device, electronic device, and storage medium. Background Art

[0002] With the increasing demand for information security in today's society, traditional identification methods such as fingerprints and irises are no longer effective. Therefore, magnetocardiography, a biometric feature with higher security, holds great promise for identification. Related technologies typically use image recognition technology to identify magnetocardiographic signals. However, using image recognition for identification loses spatial correlation information and introduces redundant channel information, resulting in identification accuracy and efficiency that fall short of application requirements.

[0003] Therefore, a new method for identifying the identity of magnetic cardiogram signals is urgently needed. Summary of the Invention

[0004] Embodiments of the present invention provide a method, device, electronic device, and storage medium for identifying an identity using a magnetocardiogram signal, to at least partially resolve the problems existing in the related art.

[0005] A first aspect of an embodiment of the present invention provides a method for identifying an identity using a magnetocardiogram signal, the method comprising:

[0006] receiving a to-be-identified magnetocardiographic signal, wherein the to-be-identified magnetocardiographic signal comprises: a corresponding magnetocardiographic signal time series measured by one or more probes;

[0007] performing noise reduction processing on the magnetocardiographic signal time series;

[0008] Converting the denoised magnetocardiographic signal time series into a time-frequency matrix to obtain one or more time-frequency matrices;

[0009] Determine the time slice mode and position combination mode according to the matrix reorganization mode used in the training of the magnetocardiographic signal recognition model;

[0010] reorganizing the one or more time-frequency matrices into a target matrix according to a time slicing pattern and a position combination pattern;

[0011] Inputting the target matrix into a pre-trained magnetocardiographic signal recognition model to determine the identity information corresponding to the user's magnetocardiographic signal;

[0012] The training samples of the magnetocardiographic signal recognition model are a sample matrix and its identity information, and the sample matrix is ​​obtained based on the sample magnetocardiographic signal time series after noise reduction processing and the matrix reorganization pattern.

[0013] Optionally, performing noise reduction processing on the magnetocardiographic signal time series includes:

[0014] Removing high-frequency noise and low-frequency noise from the magnetocardiographic signal time series by digital filtering;

[0015] The power frequency noise of the magnetocardiographic signal time series is suppressed by superimposing a sine wave.

[0016] Optionally, suppressing the power frequency noise of the magnetocardiographic signal time series by superimposing a sine wave includes:

[0017] The mid-edge deformation region of the filtered magnetocardiographic signal time series is truncated;

[0018] The power frequency noise amplitude is determined by measuring the average absolute value of all maxima and minima in the magnetocardiographic signal time series, and the phase of the power frequency noise is determined by the position of the first maximum value.

[0019] A 50 Hz sine wave with the same phase and opposite amplitude is generated and superimposed on the magnetocardiographic signal time series to cancel out the power frequency noise.

[0020] Optionally, the method of suppressing the power frequency noise of the magnetocardiographic signal time series by superimposing a sine wave further includes:

[0021] By sweeping the frequency and comparing the signal variance of the superimposed magnetocardiographic signal time series, the offset of the power frequency noise relative to 50Hz in the current cycle is determined;

[0022] The amplitude and phase of the power frequency noise are re-determined for each preset period, and the power frequency noise is eliminated according to the corresponding amplitude and phase.

[0023] Optionally, the training step of the magnetocardiographic signal recognition model includes:

[0024] The sample matrix data in the training set is passed through the convolutional neural network to obtain the predicted value;

[0025] The error between the predicted value and the true value is quantified by the cross entropy loss function;

[0026] Optimize network parameters through training error through back propagation algorithm;

[0027] It is determined whether the training error is less than an error threshold. If so, the next set of sample matrix data is used for model training until a magnetocardiographic signal recognition model with an error less than the error threshold is obtained through training.

[0028] Optionally, the sample matrix data is determined by the following steps:

[0029] Acquire sample magnetocardiographic signals, wherein the sample magnetocardiographic signals include a plurality of sample magnetocardiographic signal time series;

[0030] performing noise reduction processing on the multiple sample magnetocardiographic signal time series respectively;

[0031] Convert the sample magnetocardiographic signal time series after noise reduction into a sample time-frequency matrix;

[0032] Divide each sample time-frequency matrix into multiple sample time-frequency matrix slices according to the time division interval;

[0033] Combining sample time-frequency matrix slices of a plurality of adjacent positions into a sample matrix according to the same time division interval; determining the time division interval as a time slicing mode, determining the plurality of adjacent positions as a position combination mode, and determining the time slicing mode and the position combination mode as a matrix reorganization mode;

[0034] Multiple sample matrices are divided into training sets and test sets, and each sample matrix is ​​labeled with corresponding identity information.

[0035] Optionally, the verification step of the magnetocardiographic signal recognition model includes:

[0036] The sample matrix in the test set is calculated using the trained magnetocardiographic signal recognition model, and its output value is compared to see if it meets the verification threshold of a certain identity information. If so, it is determined to be the identity information;

[0037] The determined identity identification information is compared with the identity identification information corresponding to the sample matrix to verify the accuracy of the magnetocardiographic signal recognition model.

[0038] Based on the same inventive concept, a second aspect of an embodiment of the present invention provides a device for identifying an individual using a magnetocardiogram signal, the device comprising:

[0039] A receiving module is configured to receive a to-be-identified magnetocardiographic signal, wherein the to-be-identified magnetocardiographic signal comprises: a corresponding magnetocardiographic signal time series measured by one or more probes;

[0040] a noise reduction processing module, configured to perform noise reduction processing on the magnetocardiographic signal time series;

[0041] a conversion module, configured to convert the denoised magnetocardiographic signal time series into a time-frequency matrix by wavelet transform, thereby obtaining one or more time-frequency matrices;

[0042] A first determination module is used to determine a time slicing mode and a position combination mode according to a matrix reorganization mode used in training a magnetocardiographic signal recognition model;

[0043] a reorganization module, configured to reorganize the one or more time-frequency matrices into a target matrix according to a time slicing pattern and a position combination pattern;

[0044] A second determination module is configured to input the target matrix into a pre-trained magnetocardiographic signal recognition model to determine the identity information corresponding to the user's magnetocardiographic signal;

[0045] The training samples of the magnetocardiographic signal recognition model are a sample matrix and its identity information, and the sample matrix is ​​obtained based on the sample magnetocardiographic signal time series after noise reduction processing and the matrix reorganization pattern.

[0046] Optionally, the noise reduction processing module specifically includes:

[0047] A digital filtering submodule, configured to remove high-frequency noise and low-frequency noise from the magnetocardiographic signal time series through digital filtering;

[0048] The superposition submodule is used to suppress the power frequency noise of the magnetocardiographic signal time series by superimposing a sine wave.

[0049] Optionally, the superposition submodule is specifically configured to:

[0050] The mid-edge deformation region of the filtered magnetocardiographic signal time series is truncated;

[0051] The power frequency noise amplitude is determined by measuring the average absolute value of all maxima and minima in the magnetocardiographic signal time series, and the phase of the power frequency noise is determined by the position of the first maximum value.

[0052] A 50 Hz sine wave with the same phase and opposite amplitude is generated and superimposed on the magnetocardiographic signal time series to cancel out the power frequency noise.

[0053] Optionally, the superposition submodule is further configured to:

[0054] By sweeping the frequency and comparing the signal variance of the superimposed magnetocardiographic signal time series, the offset of the power frequency noise relative to 50 Hz is determined.

[0055] The amplitude and phase of the power frequency noise are re-determined for each preset period, and the power frequency noise is eliminated according to the corresponding amplitude and phase.

[0056] Optionally, the training step of the magnetocardiographic signal recognition model includes:

[0057] The sample matrix data in the training set is passed through the convolutional neural network to obtain the predicted value;

[0058] The error between the predicted value and the true value is quantified by the cross entropy loss function;

[0059] Optimize network parameters through training error through back propagation algorithm;

[0060] It is determined whether the training error is less than an error threshold. If so, the next set of sample matrix data is used for model training until a magnetocardiographic signal recognition model with an error less than the error threshold is obtained through training.

[0061] Optionally, the sample matrix data is determined by the following steps:

[0062] Acquire sample magnetocardiographic signals, wherein the sample magnetocardiographic signals include a plurality of sample magnetocardiographic signal time series;

[0063] performing noise reduction processing on the multiple sample magnetocardiographic signal time series respectively;

[0064] Convert the sample magnetocardiographic signal time series after noise reduction into a sample time-frequency matrix;

[0065] Divide each sample time-frequency matrix into multiple sample time-frequency matrix slices according to the time division interval;

[0066] Combining sample time-frequency matrix slices of a plurality of adjacent positions into a sample matrix according to the same time division interval; determining the time division interval as a time slicing mode, determining the plurality of adjacent positions as a position combination mode, and determining the time slicing mode and the position combination mode as a matrix reorganization mode;

[0067] Multiple sample matrices are divided into training sets and test sets, and each sample matrix is ​​labeled with corresponding identity information.

[0068] Optionally, the verification step of the magnetocardiographic signal recognition model includes:

[0069] The sample matrix in the test set is calculated using the trained magnetocardiographic signal recognition model, and its output value is compared to see if it meets the verification threshold of a certain identity information. If so, it is determined to be the identity information;

[0070] The determined identity identification information is compared with the identity identification information corresponding to the sample matrix to verify the accuracy of the magnetocardiographic signal recognition model.

[0071] Based on the same inventive concept, the third aspect of an embodiment of the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and runnable on the processor, wherein when the processor executes the computer program, the steps in the method described in the first aspect of the present invention are implemented.

[0072] Based on the same inventive concept, a fourth aspect of an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps in the method described in the first aspect of the present invention are implemented.

[0073] In this embodiment of the present invention, a deep learning model (a MCG signal recognition model) is used to identify the identity of a person using MCG signals. Furthermore, the MCG signals are converted into a target matrix for identification. This fully utilizes the spatial correlation information of the MCG signals. Furthermore, the deep learning model automatically optimizes the process of distinguishing the characteristic information of MCG signals from different users, making it less susceptible to subjective operator factors and more practical. Furthermore, a convolutional neural network structure specialized for MCG information identification is used, resulting in higher accuracy in the verification results. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0075] Figure 1 This is a flow chart of a method for identifying an individual using a magnetocardiogram signal according to an embodiment of the present invention;

[0076] Figure 2 This is a flow chart of the training steps of a magnetocardiographic signal recognition model of a method for identifying an identity using magnetocardiographic signals according to an embodiment of the present invention;

[0077] Figure 3 This is a schematic diagram of time-frequency matrix conversion of a method for identifying an identity using a magnetocardiogram signal according to an embodiment of the present invention;

[0078] Figure 4 This is a schematic diagram of time-frequency matrix slicing division of a method for identifying identity using magnetocardiographic signals according to an embodiment of the present invention;

[0079] Figure 5 1 is a schematic diagram of sample matrix reorganization of a method for identifying identity using magnetocardiographic signals according to an embodiment of the present invention;

[0080] Figure 6 This is a structural block diagram of a device for identifying identity using a magnetocardiogram signal according to an embodiment of the present invention. DETAILED DESCRIPTION

[0081] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0082] Reference Figure 1 , shows a flow chart of a method for identifying an identity using a cardiogram signal according to an embodiment of the present invention. The method for identifying an identity using a cardiogram signal according to an embodiment of the present invention may include the following steps:

[0083] S101 , receiving a to-be-identified magnetocardiographic signal, wherein the to-be-identified magnetocardiographic signal includes: a corresponding magnetocardiographic signal time series measured by one or more probes.

[0084] In the embodiment of the present invention, the user's magnetocardiographic signal to be identified can be obtained in an environment without magnetic shielding by using a magnetometer. Specific magnetometers include but are not limited to SQUID, lamp-pumped magnetometer, and atomic magnetometer.

[0085] In the embodiment of the present invention, the magnetocardiogram to be identified may be a magnetocardiogram, which includes: a corresponding magnetocardiogram time series measured by one or more probes.

[0086] In this embodiment of the present invention, multiple probes can be used to measure the cardiac magnetometry signals at various locations in front of the chest cavity when the probe sensitivity direction and the geomagnetic field direction are at different angles. The measurement time needs to be long enough to measure the cardiac magnetometry signals of multiple stable cardiac cycles.

[0087] Specifically, each probe can measure the magnetocardiographic signal at each position of the chest cavity, which can be represented as a magnetocardiographic signal time series as the measurement time changes.

[0088] Specifically, in the embodiment of the present invention, the magnetocardiographic signal to be identified is a stable magnetocardiographic signal of the user measured in a resting state.

[0089] S102: performing noise reduction processing on the magnetocardiographic signal time series.

[0090] In the embodiment of the present invention, a conventional noise reduction algorithm may be used to perform noise reduction processing on the magnetocardiographic signal time series to remove high-frequency noise, low-frequency noise and power frequency noise contained in the magnetocardiographic signal time series.

[0091] In the embodiment of the present invention, the noise reduction processing method in the actual application process is the same as the noise reduction processing method of the sample data used in the model training process.

[0092] Specifically, step S102 may include the following sub-steps:

[0093] S1021 , removing high-frequency noise and low-frequency noise from the magnetocardiographic signal time series through digital filtering.

[0094] In the embodiment of the present invention, low-frequency noise of the signal may be removed by sliding average, and high-frequency noise of the signal may be removed by filtering, where the filter may include but is not limited to a Butterworth filter.

[0095] For example, based on the original data, a window with a width of 1 second can be used to remove the sliding average of the signal.

[0096] For example, the signal may be low-pass filtered at 75 Hz using a Butterworth filter.

[0097] S1022: Superimpose a sine wave to reduce the power frequency noise of the magnetocardiographic signal time series.

[0098] In this embodiment of the present invention, the amplitude of the power-frequency noise in the digitally filtered magnetocardiographic signal time series can be determined by finding the peak. A sine wave with the power-frequency noise amplitude as its amplitude is generated. This sine wave is superimposed on the digitally filtered magnetocardiographic signal time series. The frequency spectrum is scanned within the range of 49.8-50.2 Hz, and the phase is scanned within the range of 0-360°. Based on the superposition effect, the frequency of the superimposed sine wave is fine-tuned to minimize the variance between each point in the superimposed signal.

[0099] In the embodiment of the present invention, in an environment where the mains frequency is unstable, the step of superimposing the sine wave can be divided into small segments for processing, for example, processing the mains frequency noise according to a preset time period.

[0100] Specifically, step S1022 may include the following sub-steps:

[0101] S10221, truncating the mid-edge deformation region of the filtered magnetocardiographic signal time series.

[0102] In the embodiment of the present invention, filtering the MCG signal time series in step S1021 may result in waste segments in the filtered MCG signal time series. Therefore, it is necessary to truncate the edge deformation region of the filtered MCG signal time series.

[0103] S10222: Determine the power frequency noise amplitude by measuring the average absolute value of all maxima and minima in the magnetocardiographic signal time series, and confirm the phase of the power frequency noise by the position of the first maximum value.

[0104] In the embodiment of the present invention, the amplitude and phase of the power frequency noise can be determined through step S10222.

[0105] S10223 , generating a 50 Hz sine wave with the same phase and opposite amplitude as the power frequency noise, and superimposing the 50 Hz sine wave on the magnetocardiographic signal time series to cancel out the power frequency noise.

[0106] In the embodiment of the present invention, an anti-phase sine wave is superimposed on the centripetal magnetic signal time series with a determined amplitude and phase to offset the power frequency noise.

[0107] S10224 , determining the offset of the power frequency noise relative to 50 Hz in the current cycle by frequency sweeping and comparing the signal variance of the superimposed magnetocardiographic signal time series.

[0108] In the embodiment of the present invention, the frequency of the superimposed sine wave can be fine-tuned according to the effect of superposition in S10223 to further accurately determine the frequency of the power frequency noise.

[0109] S10225: Re-determine the amplitude and phase of the power frequency noise for each preset period, and eliminate the power frequency noise according to the corresponding amplitude and phase.

[0110] In an embodiment of the present invention, in view of the unstable frequency of power frequency noise, the amplitude and phase of the power frequency noise can be re-determined at the beginning of each preset cycle. For example, the amplitude and phase of the power frequency noise can be re-determined every 5 seconds, and the power frequency noise can be eliminated.

[0111] S103 , converting the denoised magnetocardiographic signal time series into a time-frequency matrix to obtain one or more time-frequency matrices.

[0112] In the embodiment of the present invention, the time series can be converted into a time-frequency matrix through wavelet transformation.

[0113] Specifically, the time series of the magnetocardiographic signal is a function of the magnetic field strength with respect to time. The strength of each frequency component of the magnetocardiographic signal at a certain moment can be determined by wavelet transform.

[0114] Specifically, wavelet transform can be implemented by directly calling macro packages in languages ​​such as Python and MATLAB.

[0115] S104 , determining a time slicing mode and a position combination mode according to the matrix reorganization mode used in training the magnetocardiographic signal recognition model.

[0116] S105 , reorganizing the one or more time-frequency matrices into a target matrix according to a time slicing mode and a position combination mode.

[0117] In this embodiment of the present invention, the time slicing mode is used to indicate the time division interval to divide the time-frequency matrix into multiple time-frequency matrix slices. The position combination mode is used to indicate the combination mode of the time-frequency matrices corresponding to the time series of magnetocardiographic signals obtained by probes at different positions, so that the time-frequency matrix slices at multiple adjacent positions are combined into a target matrix according to the same time division interval.

[0118] S106: Input the target matrix into a pre-trained magnetocardiographic signal recognition model to determine the identity information corresponding to the user's magnetocardiographic signal.

[0119] In an embodiment of the present invention, the training samples of the magnetocardiographic signal recognition model are a sample matrix and its identity information, and the sample matrix is ​​obtained based on the sample magnetocardiographic signal time series after noise reduction processing and the matrix reorganization pattern.

[0120] In this embodiment of the present invention, the magnetocardiographic signal recognition model is obtained by training a preset convolutional neural network using a sample matrix as training samples and the identity information corresponding to the sample matrix as labels. In this embodiment of the present invention, the method for determining the sample matrix corresponds to a matrix reorganization mode (including a time slicing mode and a position combination mode). Therefore, when using the trained magnetocardiographic signal recognition model for identity recognition, it is necessary to first reorganize the time-frequency matrix to be recognized using the matrix reorganization mode to obtain a target matrix. This target matrix is ​​then recognized to predict the corresponding identity.

[0121] In this embodiment of the present invention, a deep learning model (a MCG signal recognition model) is used to identify the identity of a person using a cardiogram (MC) signal. The MCG signal is converted into a target matrix for identification. This fully utilizes the spatial correlation information of the MCG signal. Furthermore, the deep learning model automatically optimizes the process of distinguishing the characteristic information of the MCG signal for different users, making it less susceptible to subjective operator factors and more practical. Furthermore, a convolutional neural network structure specialized for identifying MCG information is used, resulting in higher accuracy in the verification results.

[0122] Reference Figure 2 , shows a flow chart of the training and verification steps of a central magnetic signal recognition model of a method for identifying an identity by a cardiogram signal according to an embodiment of the present invention. The training and verification steps of the central magnetic signal recognition model of the method for identifying an identity by a cardiogram signal according to an embodiment of the present invention may specifically include:

[0123] S201, the sample matrix data in the training set is used to obtain a predicted value through a convolutional neural network.

[0124] In the embodiment of the present invention, the network structure of the convolutional neural network can be determined first, so that it has a higher ability to distinguish small individual feature differences and extract correlation information between different channels.

[0125] In an embodiment of the present invention, a convolutional neural network may be trained using a sample matrix to obtain a magnetocardiographic signal recognition model.

[0126] In the embodiment of the present invention, the sample matrix is ​​determined by the following steps:

[0127] S1, acquiring sample magnetocardiographic signals, where the sample magnetocardiographic signals include a plurality of sample magnetocardiographic signal time series.

[0128] In the embodiment of the present invention, a sample magnetocardiographic signal time series may correspond to a sample subject, and thus the sample magnetocardiographic signal may include sample magnetocardiographic signal time series of multiple sample subjects. Thus, the magnetocardiographic signal recognition model obtained by training with the sample magnetocardiographic signals can be used to identify the identities of these sample subjects.

[0129] S2, performing noise reduction processing on the multiple sample magnetocardiographic signal time series respectively.

[0130] S3, converting the sample magnetocardiographic signal time series after noise reduction processing into a sample time-frequency matrix.

[0131] For example, Figure 3 As shown, the 7*7 probe is used to measure the Figure 3 The MCG signals at various locations in front of the chest are shown. The probe at each location measures a time series, which in this embodiment of the present invention is treated as a MCG signal time series. In this example, a wavelet transform is used to convert a MCG signal time series into a time-frequency matrix.

[0132] In the embodiment of the present invention, the magnetocardiographic signal time series may also be converted into a time-frequency matrix through short-time Fourier transform.

[0133] In this embodiment of the present invention, the regions at the beginning and end of the sample MCG signal time series, where significant distortion is caused by filtering, can be truncated. For example, 5 seconds of data before and after can be removed. In this embodiment of the present invention, the MCG signal data is segmented into small segments containing no fewer than three cardiac cycles. The length of each segment can be determined based on the GPU performance of the operating platform. For example, a segment length of 3 seconds can be used.

[0134] Specifically, in an embodiment of the present invention, the waveform signal of the sample magnetocardiographic signal time series can be converted into a time-frequency matrix based on the Symlet 6 wavelet basis. The specific number of scales of the wavelet transform can be determined by the GPU performance of the operating platform.

[0135] It is understandable that when the trained MCG signal recognition model is actually used for MCG signal recognition, the noise reduction and conversion methods for the MCG signal time series to be recognized should be the same as those for the sample signal time series.

[0136] S4, dividing each sample time-frequency matrix into a plurality of sample time-frequency matrix slices according to the time division interval.

[0137] Because a large number of training samples are required during model training, in practical applications, a sufficiently long period of magnetocardiographic signals from the sample body can be collected and then sliced ​​into shorter signals to obtain a large amount of data. Specifically, each sample time-frequency matrix can be divided into multiple sample time-frequency matrix slices according to preset time intervals. In embodiments of the present invention, the preset time intervals can be determined based on actual needs, for example, 5 seconds. It will be understood that during the verification and actual application of a model trained using such shorter signals, only signals of the same length as the shorter signals are required to obtain results.

[0138] For example, Figure 4 As shown, taking a sample time-frequency matrix as an example, each sample time-frequency matrix can be divided into a series of small time-frequency matrix slices as sample time-frequency matrix slices. Specifically, the time division interval can be determined according to a preset time length.

[0139] S5, according to the same time division interval, combining the time-frequency matrix slices of multiple adjacent positions into a sample matrix; determining the time division interval as a time slicing mode, determining the multiple adjacent positions as a position combination mode, and determining the time slicing mode and the position combination mode as a matrix reorganization mode.

[0140] In an embodiment of the present invention, the time-frequency matrix converted from the time series of the magnetocardiographic signals at different locations can be reorganized into different channels of the deep learning dataset according to the spatial position relationship to train the convolutional neural network.

[0141] In the embodiment of the present invention, the number of channels depends on the number of test probes and the performance of the operating platform.

[0142] For example, Figure 5 As shown, the time-frequency matrix slices at four adjacent positions are combined into a sample matrix to obtain sample matrix data.

[0143] Furthermore, in an embodiment of the present invention, sample magnetocardiographic signals of different users may be reorganized into a four-channel sample matrix, and user names may be used as labels for calibration to obtain sample data.

[0144] S6, dividing the multiple sample matrices into a training set and a test set, and each sample matrix is ​​labeled with corresponding identity information.

[0145] S202, quantify the error between the predicted value and the true value through the cross entropy loss function.

[0146] S203, training errors through back propagation algorithm to optimize network parameters.

[0147] S204 , determining whether the training error is less than an error threshold, and if so, using the next set of sample matrix data to perform model training until a magnetocardiographic signal recognition model with an error less than the error threshold is obtained through training.

[0148] S205 , the sample matrix in the test set is calculated using the trained magnetocardiographic signal recognition model, and its output value is compared to see whether it meets the verification threshold of a certain identity identification information. If so, it is determined to be the identity identification information.

[0149] S206 : Compare the determined identity information with the identity information corresponding to the sample matrix to verify the accuracy of the magnetocardiographic signal recognition model.

[0150] In the embodiment of the present invention, specifically, pytorch can be used to build a deep learning platform.

[0151] In an embodiment of the present invention, 20% of the calibrated sample data can be randomly selected as a test set, and the rest is a training set. Both the training set and the test set are in random order. This data set is trained using an NVIDIA TESLA K20c graphics card. After 40 generations of training, the effectiveness of the model is verified after each generation of training is completed, and the model with the highest verification accuracy is selected and saved as the magnetocardiographic signal recognition model.

[0152] Specifically, the network structure used in the embodiment of the present invention is a structure combining DenseBlock and SE-Module, specifically including: three DenseBlocks with a width of 12, a convolutional layer with a convolution kernel of 3*3 and an SE-Module inserted before each DenseBlock, and a fully connected layer inserted at the end.

[0153] In the embodiment of the present invention, the trained magnetocardiographic signal recognition model can be stored and transferred using a storage medium after being saved.

[0154] In the embodiment of the present invention, the trained magnetocardiographic signal recognition model can identify multiple users corresponding to the sample matrix to confirm the personal identity recognition information of the input target matrix.

[0155] Based on the same inventive concept, the embodiment of the present invention provides a device for identifying identity using a cardiogenic signal. Figure 6 , Figure 6 : is a schematic diagram of a device for identifying an individual using a magnetocardiogram signal according to an embodiment of the present invention, the device comprising:

[0156] The receiving module 601 is configured to receive a to-be-identified magnetocardiographic signal, wherein the to-be-identified magnetocardiographic signal includes: a corresponding magnetocardiographic signal time series measured by one or more probes;

[0157] A noise reduction processing module 602 is used to perform noise reduction processing on the magnetocardiographic signal time series;

[0158] A conversion module 603 is configured to convert the denoised magnetocardiographic signal time series into a time-frequency matrix through wavelet transformation to obtain one or more time-frequency matrices;

[0159] A first determining module 604 is configured to determine a time slicing mode and a position combination mode according to a matrix reorganization mode used in training a magnetocardiographic signal recognition model;

[0160] a reorganization module 605 , configured to reorganize the one or more time-frequency matrices into a target matrix according to a time slicing pattern and a position combination pattern;

[0161] The second determination module 606 is configured to input the target matrix into a pre-trained magnetocardiographic signal recognition model to determine the identity information corresponding to the user's magnetocardiographic signal;

[0162] The training samples of the magnetocardiographic signal recognition model are a sample matrix and its identity information, and the sample matrix is ​​obtained based on the sample magnetocardiographic signal time series after noise reduction processing and the matrix reorganization pattern.

[0163] Optionally, the noise reduction processing module 602 specifically includes:

[0164] A digital filtering submodule, configured to remove high-frequency noise and low-frequency noise from the magnetocardiographic signal time series through digital filtering;

[0165] The superposition submodule is used to suppress the power frequency noise of the magnetocardiographic signal time series by superimposing a sine wave.

[0166] Optionally, the superposition submodule is specifically configured to:

[0167] The mid-edge deformation region of the filtered magnetocardiographic signal time series is truncated;

[0168] The power frequency noise amplitude is determined by measuring the average absolute value of all maxima and minima in the magnetocardiographic signal time series, and the phase of the power frequency noise is determined by the position of the first maximum value.

[0169] A 50 Hz sine wave with the same phase and opposite amplitude is generated and superimposed on the magnetocardiographic signal time series to cancel out the power frequency noise.

[0170] Optionally, the superposition submodule is further configured to:

[0171] By sweeping the frequency and comparing the signal variance of the superimposed magnetocardiographic signal time series, the offset of the power frequency noise relative to 50Hz in the current cycle is determined;

[0172] The amplitude and phase of the power frequency noise are re-determined for each preset period, and the power frequency noise is eliminated according to the corresponding amplitude and phase.

[0173] Optionally, the training step of the magnetocardiographic signal recognition model includes:

[0174] The sample matrix data in the training set is passed through the convolutional neural network to obtain the predicted value;

[0175] The error between the predicted value and the true value is quantified by the cross entropy loss function;

[0176] Optimize network parameters through training error through back propagation algorithm;

[0177] It is determined whether the training error is less than an error threshold. If so, the next set of sample matrix data is used for model training until a magnetocardiographic signal recognition model with an error less than the error threshold is obtained through training.

[0178] Optionally, the sample matrix data is determined by the following steps:

[0179] Acquire sample magnetocardiographic signals, wherein the sample magnetocardiographic signals include a plurality of sample magnetocardiographic signal time series;

[0180] performing noise reduction processing on the multiple sample magnetocardiographic signal time series respectively;

[0181] Convert the sample magnetocardiographic signal time series after noise reduction into a sample time-frequency matrix;

[0182] Divide each sample time-frequency matrix into multiple sample time-frequency matrix slices according to the time division interval;

[0183] Combining sample time-frequency matrix slices of a plurality of adjacent positions into a sample matrix according to the same time division interval; determining the time division interval as a time slicing mode, determining the plurality of adjacent positions as a position combination mode, and determining the time slicing mode and the position combination mode as a matrix reorganization mode;

[0184] Multiple sample matrices are divided into training sets and test sets, and each sample matrix is ​​labeled with corresponding identity information.

[0185] Optionally, the verification step of the magnetocardiographic signal recognition model includes:

[0186] The sample matrix in the test set is calculated using the trained magnetocardiographic signal recognition model, and its output value is compared to see if it meets the verification threshold of a certain identity information. If so, it is determined to be the identity information;

[0187] The determined identity identification information is compared with the identity identification information corresponding to the sample matrix to verify the accuracy of the magnetocardiographic signal recognition model.

[0188] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0189] Based on the same inventive concept, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the method for identifying identity by magnetocardiographic signals described in any of the above embodiments are implemented.

[0190] Based on the same inventive concept, an embodiment of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the method for identifying identity by magnetic cardiogram signals described in any of the above embodiments are implemented.

[0191] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0192] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, apparatus, or computer program products. Thus, embodiments of the present invention may take the form of a fully hardware embodiment, a fully software embodiment, or an embodiment combining software and hardware. Furthermore, embodiments of the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0193] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as the combination of processes and / or blocks in the flowcharts and / or block diagrams, 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 magnetic cardiogram signal identification terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable magnetic cardiogram signal identification terminal device generate instructions for implementing the process in the flowcharts and / or block diagrams. 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.

[0194] These computer program instructions can also be stored in a computer readable memory that can guide a computer or other programmable cardiogenic signal identification terminal device to work in a specific manner, so that the instructions stored in the computer readable memory produce a product including an instruction device, which is implemented in the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0195] These computer program instructions can also be loaded onto a computer or other programmable magnetic cardiogram signal identification terminal device, so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions executed on the computer or other programmable terminal device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0196] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they become aware of the basic creative concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present invention.

[0197] Finally, it should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or terminal device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or terminal device that includes the element.

[0198] The above is a detailed introduction to the method, device, electronic device and storage medium for identifying cardiogenic signals provided by the present invention. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the ideas of the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for identifying an individual using a cardiogenic signal, characterized in that: The method comprises: receiving a to-be-identified magnetocardiographic signal, wherein the to-be-identified magnetocardiographic signal comprises: a corresponding magnetocardiographic signal time series measured by one or more probes; performing noise reduction processing on the magnetocardiographic signal time series; Converting the denoised magnetocardiographic signal time series into a time-frequency matrix to obtain one or more time-frequency matrices; Determine the time slice mode and position combination mode according to the matrix reorganization mode used in the training of the magnetocardiographic signal recognition model; reorganizing the one or more time-frequency matrices into a target matrix according to a time slicing pattern and a position combination pattern; Inputting the target matrix into a pre-trained magnetocardiographic signal recognition model to determine the identity information corresponding to the user's magnetocardiographic signal; The training samples of the magnetocardiographic signal recognition model are a sample matrix and its identity information, and the sample matrix is ​​obtained based on the sample magnetocardiographic signal time series after noise reduction processing and the matrix reorganization pattern.

2. The method for identifying identity using a cardiogenic signal according to claim 1, wherein: The noise reduction process is performed on the magnetocardiographic signal time series, comprising: Removing high-frequency noise and low-frequency noise from the magnetocardiographic signal time series by digital filtering; The power frequency noise of the magnetocardiographic signal time series is suppressed by superimposing a sine wave.

3. The method for identifying identity using a cardiogenic signal according to claim 2, wherein: The power frequency noise of the magnetocardiographic signal time series is suppressed by superimposing a sine wave, including: The mid-edge deformation region of the filtered magnetocardiographic signal time series is truncated; The power frequency noise amplitude is determined by measuring the average absolute value of all maxima and minima in the magnetocardiographic signal time series, and the phase of the power frequency noise is determined by the position of the first maximum value. A 50 Hz sine wave with the same phase and opposite amplitude is generated and superimposed on the magnetocardiographic signal time series to cancel out the power frequency noise.

4. The method for identifying identity using magnetocardiographic signals according to claim 3, wherein: The method further comprises: suppressing the power frequency noise of the magnetocardiographic signal time series by superimposing a sine wave; By sweeping the frequency and comparing the signal variance of the superimposed magnetocardiographic signal time series, the offset of the power frequency noise relative to 50Hz in the current cycle is determined; The amplitude and phase of the power frequency noise are re-determined for each preset period, and the power frequency noise is eliminated according to the corresponding amplitude and phase.

5. The method for identifying identity using magnetocardiographic signals according to claim 1, wherein: The training steps of the magnetocardiographic signal recognition model include: The sample matrix in the training set is passed through the convolutional neural network to obtain the predicted value; The error between the predicted value and the true value is quantified by the cross entropy loss function; Optimize network parameters through training error through back propagation algorithm; It is determined whether the training error is less than an error threshold. If so, the next set of sample matrix data is used for model training until a magnetocardiographic signal recognition model with an error less than the error threshold is obtained through training.

6. The method for identifying identity using magnetocardiographic signals according to claim 5, wherein: The sample matrix is ​​determined by the following steps: Acquire sample magnetocardiographic signals, wherein the sample magnetocardiographic signals include a plurality of sample magnetocardiographic signal time series; performing noise reduction processing on the multiple sample magnetocardiographic signal time series respectively; Convert the sample magnetocardiographic signal time series after noise reduction into a sample time-frequency matrix; Divide each sample time-frequency matrix into multiple sample time-frequency matrix slices according to the time division interval; Combining sample time-frequency matrix slices of a plurality of adjacent positions into a sample matrix according to the same time division interval; determining the time division interval as a time slicing mode, determining the plurality of adjacent positions as a position combination mode, and determining the time slicing mode and the position combination mode as a matrix reorganization mode; Multiple sample matrices are divided into training sets and test sets, and each sample matrix is ​​labeled with corresponding identity information.

7. The method for identifying identity using magnetocardiographic signals according to claim 6, wherein: The verification step of the magnetocardiographic signal recognition model includes: The sample matrix in the test set is calculated using the trained magnetocardiographic signal recognition model, and its output value is compared to see if it meets the verification threshold of a certain identity information. If so, it is determined to be the identity information; The determined identity identification information is compared with the identity identification information corresponding to the sample matrix to verify the accuracy of the magnetocardiographic signal recognition model.

8. A device for identifying identity using a magnetic cardiogram signal, characterized in that: The device comprises: A receiving module is configured to receive a to-be-identified magnetocardiographic signal, wherein the to-be-identified magnetocardiographic signal comprises: a corresponding magnetocardiographic signal time series measured by one or more probes; a noise reduction processing module, configured to perform noise reduction processing on the magnetocardiographic signal time series; a conversion module, configured to convert the denoised magnetocardiographic signal time series into a time-frequency matrix by wavelet transform, thereby obtaining one or more time-frequency matrices; A first determination module is used to determine a time slicing mode and a position combination mode according to a matrix reorganization mode used in training a magnetocardiographic signal recognition model; a reorganization module, configured to reorganize the one or more time-frequency matrices into a target matrix according to a time slicing pattern and a position combination pattern; A second determination module is configured to input the target matrix into a pre-trained magnetocardiographic signal recognition model to determine the identity information corresponding to the user's magnetocardiographic signal; The training samples of the magnetocardiographic signal recognition model are a sample matrix and its identity information, and the sample matrix is ​​obtained based on the sample magnetocardiographic signal time series after noise reduction processing and the matrix reorganization pattern.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method for identifying the identity of a cardiogenic signal according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for identifying an identity by a magnetocardiographic signal are implemented.

Images