A method for suppressing magnetoencephalogram background noise based on spatio-temporal domain projection filtering

Abstract

The present invention relates to a method for suppressing magnetoencephalogram background noise based on spatio-temporal domain projection filtering, belonging to the field of magnetoencephalogram signal denoising processing. Sensors are used to measure multi-channel resting-state noise signals and multi-channel magnetoencephalogram signals; a homogeneous domain sub-projection matrix is solved by using a sensor direction matrix; the multi-channel resting-state noise signals and multi-channel magnetoencephalogram signals are subjected to time-domain and space-domain projection filtering by using the homogeneous domain sub-projection matrix to suppress the background noise in the multi-channel magnetoencephalogram signals and extract and separate the required magnetoencephalogram nerve signals. The present invention suppresses the background noise in the original magnetoencephalogram signals to obtain high-quality magnetoencephalogram data, which has important practical significance for subsequent magnetoencephalogram scientific research and clinical research.

Description

Technical Field

[0001] The present invention belongs to the field of magnetoencephalogram signal denoising processing, and relates to a method for suppressing magnetoencephalogram background noise based on spatio-temporal domain projection filtering, which is applicable to suppressing background noise in magnetoencephalogram signals. Background Art

[0002] Magnetoencephalogram, as a non-invasive magnetic source imaging technology for measuring brain nerve activities, can reflect specific physiological activities of the brain and locate lesions of related neurological diseases by measuring changes in magnetic field intensity. It has higher temporal and spatial resolutions compared to electroencephalogram and some other brain nerve imaging measurement methods.

[0003] In terms of magnetoencephalogram measurement sensors, miniaturized optically pumped magnetometers have currently achieved performance similar to that of superconducting quantum interference devices and are expected to be put into clinical applications in the future. However, since the magnitude of the brain nerve magnetic field is several orders of magnitude lower than that of interference signals, a large amount of background magnetic field noise exists in the signal data measured by the optically pumped magnetometer when measuring magnetoencephalogram. The sources of these noises are diverse, such as nearby moving cars, power lines, motion capture devices, and magnetic particles brought into the magnetic shielding room by the subject himself. If these background noises are not suppressed, the distorted magnetoencephalogram signal data will affect the solution of subsequent magnetoencephalogram forward and inverse problems, functional connectivity, etc. Currently, the research on noise suppression is divided into two categories: hardware and software algorithms. Among them, the hardware denoising methods mainly include methods such as combining reference channels with active magnetic compensation coils. However, on the one hand, the background magnetic field after compensation is still very strong compared to the brain signal, and on the other hand, the hardware conditions of some magnetic shielding rooms are limited and a hardware magnetic compensation system cannot be deployed. Therefore, it is necessary to develop a software algorithm with better noise suppression effect.

[0004] Currently, there are mainly the following several magnetoencephalogram background noise suppression algorithms: (1) Signal space projection algorithm, which projects the measured signal onto a subspace orthogonal to the interference spatial distribution to achieve noise elimination. However, the signal space projection algorithm only uses the noise data of an empty room without subjects and cannot handle the noise brought in by the subjects. (2) Signal space separation algorithm and spatio-temporal signal space separation method. The signal space separation algorithm divides the measured signal into intracerebral nerve signals and extracerebral noise signals based on Maxwell's equations and spherical harmonics. The spatio-temporal signal space separation method further projects the signal separated by the signal space separation algorithm onto the time domain to eliminate the interference sources near the sensors. The algorithm models of this type are relatively complex, and a large number of sensors are required for sampling (currently, the measurement experiments of optically pumped magnetometers cannot meet this requirement); precise registration of the sensor positions is required; and factors such as setting appropriate parameters are needed to construct an accurate spherical harmonic function model. (3) Dual signal subspace projection algorithm, which uses the lead field matrix to perform double projections in the spatial domain and the time domain to obtain the internal nerve signals. This algorithm also requires precise registration of the sensor positions for constructing the lead field matrix. (4) Uniform field correction algorithm, which models the background magnetic field as a low-dimensional signal, constructs a projection subspace through the sensor direction matrix, and separates the nerve signals from external interferences. This algorithm requires a good magnetic shielding effect during the experiment, but sometimes the actual conditions cannot meet the requirements; secondly, the subjects may also bring in certain magnetic field interference sources, resulting in a still high background magnetic field intensity near the sensors, causing poor denoising effect of this method. Summary of the Invention

[0005] The technical problem to be solved by the present invention is: to overcome the problem that the traditional background noise suppression algorithm has poor effect in processing the measurement data of optically pumped magnetometers, and to provide a magnetoencephalogram background noise suppression method based on spatio-temporal domain projection filtering, which has good background noise suppression effect and can obtain high-quality magnetoencephalogram data.

[0006] The technical solution adopted by the present invention to solve its technical problem is: constructing a low-dimensional homogeneous field magnetic field model with the sensor direction matrix, and performing spatio-temporal domain projection filtering on the measured resting state noise data and magnetoencephalogram signal data to obtain the denoised magnetoencephalogram nerve signals.

[0007] In a first aspect, the present invention provides a magnetoencephalogram background noise suppression method based on spatio-temporal domain projection filtering, including the following steps:

[0008] Step 1: Use sensors to measure multi-channel resting state noise signals and multi-channel magnetoencephalogram signals; the multi-channel resting state noise signals refer to the multi-channel measurement signals obtained when the subject is not stimulated by an experimental paradigm, the subject remains stationary, and there is no evoked response; the multi-channel magnetoencephalogram measurement signals are the multi-channel magnetoencephalogram evoked response signals obtained by stimulating the subject with an experimental paradigm.

[0009] Step 2: Solve the homogeneous domain sub-projection matrix using the sensor direction matrix;

[0010] Step 3: Perform time-domain and space-domain projection filtering on the multi-channel resting-state noise signals and multi-channel magnetoencephalogram signals using the homogeneous domain sub-projection matrix to suppress the background noise in the multi-channel magnetoencephalogram signals and extract and separate the required magnetoencephalogram nerve signals.

[0011] Furthermore, in the above Step 1, an optically pumped magnetometer is used as the sensor.

[0012] Furthermore, in the above Step 2, the specific method for solving the homogeneous domain subspace matrix using the sensor direction matrix is as follows:

[0013] The magnetoencephalogram measurement signal of a single sensor is approximately decomposed into homogeneous components in the three directions of the sensor coordinate axes:

[0014] B = o x b x + o y b y + o z b z (1)

[0015] where B is the multi-channel magnetoencephalogram signal, a coordinate system is established with the center of the head as the origin, o x , o y , o z are the values of the sensor in the three coordinate axes directions of x, y, and z, and b x , b y , b z are the components of the multi-channel magnetoencephalogram signal in the three coordinate axes directions of x, y, and z. Then, the total multi-channel magnetoencephalogram measurement signals of n sensors are expressed as:

[0016]

[0017] The values of the three coordinate axes directions of n sensors are row concatenated into a unit direction matrix N, and the homogeneous field projection matrix M is solved therefrom:

[0018] M = I - NN + (3)

[0019] where I is the identity matrix and N + is the pseudo-inverse matrix of N.

[0020] Furthermore, in the above Step 3, the specific steps for extracting and separating the magnetoencephalogram nerve signals are as follows:

[0021] ① Model the multi-channel magnetoencephalogram measurement signals as follows:

[0022] B = A + H (4)

[0023] H = H f +H c (5)

[0024] B is the multi-channel magnetoencephalogram measurement signal, A is the brain nerve signal, and H is the background noise;

[0025] Among them, the background noise further includes the noise H with a distance from the sensor less than or equal to the head size c and the noise H with a distance from the sensor much greater than the head size f .

[0026] The noise with a distance from the sensor much greater than the head size is zero in the homogeneous domain. Therefore, the final model is:

[0027] MB = MA + MH c (6)

[0028] M is the homogeneous domain sub-projection matrix obtained in step 2.

[0029] For the convenience of writing, MB is denoted as B m , MH c is denoted as H mc .

[0030] ② Perform spatial domain projection filtering on the multi-channel resting state noise signal to preliminarily separate the internal brain signal and the external interference signal;

[0031] The process of spatial domain projection filtering is as follows:

[0032] Perform singular value decomposition on H mc :

[0033]

[0034] N is the number of sensors

[0035] λ1, λ2…λ N are the singular values of the H mc matrix.

[0036] Arrange the singular values λ1, λ2…λ N in descending order, and select the eigenvectors of H mc corresponding to the first p singular values in the spatial domain:

[0037] F P = [f1,…,f P (8)

[0038] Obtain the projection operator of H mc :

[0039] Then, the multi-channel magnetoencephalogram signals are subjected to spatial projection filtering through equations (9) and (10) to obtain the preliminary internal brain signal B in and the external interference signal B out :

[0040]

[0041]

[0042] I is the identity matrix;

[0043] ③ Perform temporal projection filtering on the internal brain signal and the external interference signal;

[0044] For the internal brain signal B in and the external interference signal B out perform singular value decomposition:

[0045]

[0046]

[0047] γ1, γ2…γ N are the singular values of B in ; σ1, σ2…σ N are the singular values of B out ;

[0048] Arrange the singular values γ1, γ2…γ N , σ1, σ2…σ N in descending order, and select the eigenvectors E in in the time domain corresponding to the first q singular values and the eigenvectors D q in the time domain corresponding to the first k singular values of B out : k :

[0049] E q = [e1, …, e q (13)

[0050] D k = [d1, …·, d k (14)

[0051] Perform singular value decomposition on their common intersection :

[0052]

[0053] The singular value is the angle cos(θ ) q between the subspace of the common intersection

[0054] Since the interference signal and the nerve signal should be uncorrelated in time, only the near-orthogonal singular values are retained from Equation (16),

[0055] retaining the eigenvectors Φ corresponding to the first r singular values : r :

[0056] cos(θ1) = cos(θ2) = … = cos(θ r ) ≈ 0.98 (16)

[0057] Φ r = [φ1, …, φ r (17)

[0058] φ1, φ2 …, φ r are the first r columns of the right matrix of the singular value decomposition.

[0059] Therefore, the time-domain projection filtering operator P for the internal brain signal B in and the external interference signal B out is obtained from Equation (18): proj :

[0060] P proj = D r Φ r (18)

[0061] where D r is the eigenvector corresponding to Φ r ,

[0062] (4) Finally, Equation (19) uses P proj to achieve the purpose of suppressing background noise;

[0063]

[0064] where represents the denoised magnetoencephalogram nerve signal.

[0065] The principle of the present invention is as follows: using the direction matrix of the magnetoencephalogram measurement sensor to solve the background magnetic field as a low-dimensional homogeneous field to obtain a homogeneous field projection operator; substituting the resting-state noise data matrix and the magnetoencephalogram signal data matrix to obtain the corresponding homogeneous field projection; performing a spatial-domain projection on the noise homogeneous field projection matrix to obtain a preliminarily separated internal brain signal and external interference signal. Finally, performing a time-domain projection on the internal brain signal and the external interference signal to obtain the denoised magnetoencephalogram nerve signal.

[0066] In a second aspect, the present invention provides an electronic device, including a processor, a communication interface, a memory, and a communication bus. Among them, the processor, the communication interface, and the memory complete communication with each other through the communication bus;

[0067] The memory is used to store a computer program;

[0068] The processor is used to execute the computer program stored on the memory, and when executed, it implements the method of the foregoing method for suppressing magnetoencephalogram background noise based on spatio-temporal domain projection filtering.

[0069] In a third aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the foregoing method for suppressing magnetoencephalogram background noise based on spatio-temporal domain projection filtering.

[0070] The advantages of the present invention compared with the prior art are as follows:

[0071] (1) In the traditional signal space projection algorithm SSP, the noise data is the noise data of an empty room without a subject collected. The noise data collected by the present invention is multi-channel resting state noise signal data, which enables the effective removal of the interference sources brought into the shielding room by the subject and the spontaneous neural activity noise in the subject's brain that is irrelevant to the magnetoencephalogram evoked experiment task during the denoising process.

[0072] (2) In the traditional signal space separation SSS series algorithms, the head model and the lead field matrix in the double signal subspace projection algorithm DSSP both require a large number of sampling sensors and accurate sensor registration information. The present invention uses the sensor direction matrix to model the magnetic field around the head as a low-dimensional homogeneous field, enabling better measurement results to be obtained with fewer optically pumped magnetometer sensors.

[0073] (3) The present invention models the noise as including the noise H closer to the sensor c and the noise H farther away f , improves the noise model of the uniform field correction algorithm, performs spatial domain filtering on the measured resting state noise data, obtains the projection operator of the noise through steps such as singular value decomposition, so as to separate the internal signal and the noise signal. And on this basis, taking advantage of the characteristic that the magnetoencephalogram signal and the noise signal are uncorrelated in the time domain, further perform time domain filtering on the preliminarily separated signal, and finally obtain high-quality magnetoencephalogram neural signals with good background noise suppression effect after denoising through singular value decomposition and threshold screening. It overcomes the disadvantages of the traditional uniform field correction algorithm that has poor noise suppression effect in an environment with poor magnetic shielding effect and cannot suppress the interference sources around the sensors. Description of the Drawings

[0074] Figure 1Flowchart of the method for suppressing magnetoencephalogram background noise based on spatio-temporal domain projection filtering in the present invention;

[0075] Figure 2 Schematic diagram for separating internal signals and external interference signals using spatial domain projection in the present invention;

[0076] Figure 3 Schematic diagram for separating and denoising signals using time domain projection in the present invention. Detailed implementation manners

[0077] The present invention will be described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and not to limit the present invention.

[0078] Figure 1 The flowchart of the method for suppressing magnetoencephalogram background noise based on spatio-temporal domain projection filtering is given. The specific implementation process:

[0079] 1. Use sensors to measure multi-channel resting state noise signals and multi-channel magnetoencephalogram measurement signals; the multi-channel resting state noise signals refer to the measurement signals obtained when the subject is not stimulated by an experimental paradigm, the subject remains stationary, and there is no evoked response; the multi-channel magnetoencephalogram measurement signals refer to the magnetoencephalogram evoked response signals obtained by stimulating the subject using an experimental paradigm. The sensor in the present invention is an optically pumped magnetometer, and the optically pumped magnetometer is an extremely weak magnetic measurement sensor using the spin-exchange relaxation-free effect.

[0080] 2. Solve the homogeneous domain subspace matrix using the sensor direction matrix;

[0081] The magnetoencephalogram measurement signal of a single sensor is approximately decomposed into homogeneous components in the three directions of the sensor coordinate axes:

[0082] B = o x b x + o y b y + o z b z (20)

[0083] where B is the multi-channel magnetoencephalogram measurement signal, a coordinate system is established with the center of the head as the origin, o x , o y , o z are the values of the sensor in the three coordinate axes directions of xyz, and b x , b y , b z are the components of the multi-channel magnetoencephalogram signal in the three coordinate axes directions of xyz. Then the total multi-channel magnetoencephalogram measurement of n sensors is expressed as:

[0084]

[0085] The numerical directional values of n sensors in the x, y, and z coordinate axes are serially concatenated into a unit direction matrix N. And based on this, the homogeneous field projection matrix M is solved as follows:

[0086] M = I - NN + (22)

[0087] where I is the identity matrix, and N + is the pseudo-inverse matrix of N.

[0088] 3. Perform time-domain and spatial-domain projection filtering on the magnetoencephalogram measurement signal and the resting-state noise signal to suppress the background noise in the measurement signal and extract and separate the required magnetoencephalogram nerve signal. The specific process is as follows:

[0089] ① Model the multi-channel magnetoencephalogram measurement signal:

[0090] B = A + H (23)

[0091] H = H f + H c (24)

[0092] where B is the multi-channel magnetoencephalogram measurement signal, A is the brain nerve signal, and H is the background noise;

[0093] where the background noise further includes noise H c with a distance from the sensor less than or equal to the head size and noise H f .

[0094] Use the homogeneous field projection matrix M obtained in step 2 to project the measurement signal into the homogeneous subspace:

[0095] MB = MA + MH (25)

[0096] Since the noise with a distance from the sensor much greater than the head size is zero in the homogeneous domain, the projection of the measurement signal in the homogeneous subspace is updated from equations (23) and (24) to:

[0097] MB = MA + MH c (26)

[0098] For the convenience of writing, the measurement signal projection matrix MB will be denoted as B m subsequently, and the noise signal projection matrix MH c will be denoted as H mc .

[0099] ② Perform spatial-domain projection on the interference signal projection to preliminarily separate the internal brain signal and the external interference signal.

[0100] As Figure 2 shown, for the noise signal projection matrix Hmc Perform singular value decomposition:

[0101]

[0102] N is the number of sensors, λ1, λ2…λ N are the singular values of the H mc matrix. Arrange the singular values λ1, λ2…λ N in descending order, and select the eigenvectors of H mc corresponding to the top p singular values in the spatial domain:

[0103] F P = [f1,…,f P (28)

[0104] Thus, the projection operator of H mc is obtained Then, project the measured signal through equations (29) and (30) to obtain the preliminary internal brain signal B in and the external interference signal B out .

[0105]

[0106]

[0107] I is the identity matrix.

[0108] ③ Perform time-domain projection filtering on the internal brain signal and the external interference signal to obtain the denoised magnetoencephalogram neural signal.

[0109] As Figure 3 shown, perform singular value decomposition on the internal brain signal B in and the external interference signal B out :

[0110]

[0111]

[0112] γ1, γ2…γ N are the singular values of B in , and σ1, σ2…σ N are the singular values of B out . Arrange the singular values γ1, γ2…γ N , σ1, σ2…σ N in descending order, and select the eigenvectors E in of B q corresponding to the top q singular values in the time domain and the eigenvectors D out of B k corresponding to the top k singular values in the time domain:

[0113] E q = [e1, …, e q (33)

[0114] D k = [d1, ····, d k (34)

[0115] For the common intersection of the two perform singular value decomposition:

[0116]

[0117] where the singular values are the angles cos(θ ) between the subspace of the common intersection q and the magnetoencephalogram neural signals. Since the interference signals and neural signals should be uncorrelated in time, only the r singular values that are nearly orthogonal are retained from Equation (36), and Φ r is the corresponding eigenvector:

[0118] cos(θ1) = cos(θ2) = … = cos(θ r ) ≈ 0.98 (36)

[0119] Φ r = [φ1, …, φ r (37)

[0120] φ1, φ2…, φ r are the first r columns of the right matrix of the singular value decomposition.

[0121] Therefore, the time-domain projection operator P proj of the interference signal can be obtained from Equation (38):

[0122] P proj = D r Φ r (38)

[0123] where D r is the eigenvector corresponding to Φ r . Finally, the neural signal is reconstructed by applying the projection operator from Equation (39) to achieve the purpose of suppressing background noise.

[0124]

[0125] where represents the denoised magnetoencephalogram neural signal.

[0126] To verify the advantages of the method of the present invention in suppressing the background noise of magnetoencephalogram compared with traditional algorithms, simulation experiments were carried out. A current dipole was set in the brain head model, which was equivalent to the brain nerve source, and its signal was equivalent to the magnetoencephalogram nerve signal; and a set of multi-channel resting state noise signals was actually measured; the two signals were superimposed to simulate the multi-channel magnetoencephalogram measurement signal. The simulation multi-channel magnetoencephalogram measurement signal was denoised by the method of the present invention and several traditional algorithms respectively, and the results are as follows:

[0127] The signal-to-noise ratio results of the signal after denoising are shown in Table 1. The signal-to-noise ratio refers to the ratio of the signal to the noise. In this simulation experiment, it refers to the ratio of the current dipole signal to the difference between the signal after denoising and the current dipole signal. Therefore, the larger the signal-to-noise ratio, the better the denoising effect, and the unit is decibel (dB).

[0128] Table 1

[0129]

[0130] The root mean square error of the signal after denoising is shown in Table 2. The root mean square error is the square root of the difference between the signal after denoising and the current dipole signal, and is used to measure the deviation between the two signals. Therefore, the smaller the root mean square error, the better the denoising effect.

[0131] Table 2

[0132]

[0133] In summary, the method for suppressing the background noise of magnetoencephalogram signals proposed by the present invention has significantly better noise suppression effect than existing classical algorithms.

[0134] The content not described in detail in the specification of the present invention belongs to the prior art well-known to those skilled in the art.

[0135] The above embodiments are provided only for the purpose of describing the present invention, and are not intended to limit the scope of the present invention. The scope of the present invention is defined by the appended claims. All equivalent substitutions and modifications made without departing from the spirit and principle of the present invention shall be covered within the scope of the present invention.

Claims

1. A method for suppressing magnetoencephalogram background noise based on spatio-temporal domain projection filtering, characterized in that: It includes the following steps: Step 1: Use sensors to measure multi-channel resting-state noise signals and multi-channel magnetoencephalogram (MEG) signals. The multi-channel resting-state noise signals refer to multi-channel measurement signals obtained when the subject is not stimulated by an experimental paradigm, the subject remains stationary, and there is no evoked response. The multi-channel MEG signals are multi-channel MEG evoked response signals obtained by stimulating the subject with an experimental paradigm. Step 2: Use the sensor direction matrix to solve the homogeneous domain sub-projection matrix. Step 3: Perform time-domain and space-domain projection filtering on the multi-channel resting-state noise signals and multi-channel MEG signals using the homogeneous domain sub-projection matrix to suppress the background noise in the multi-channel MEG signals and extract and separate the desired MEG nerve signals.

2. The method for suppressing magnetoencephalogram background noise based on spatio-temporal domain projection filtering according to claim 1, characterized in that: In the said Step 1, an optically pumped magnetometer is used as the sensor.

3. The method for suppressing magnetoencephalogram background noise based on spatio-temporal domain projection filtering according to claim 1, characterized in that: In the said Step 2, the specific process of using the sensor direction matrix to solve the homogeneous domain sub-space matrix is as follows: The MEG signal of a single sensor is decomposed into homogeneous components in three directions of the sensor coordinate axes: （1） Among them is the multi-channel magnetoencephalogram signal. A coordinate system is established with the center of the head as the origin. , , are the values of the sensor in the x, y, and z axis directions. , , are the components of the multi-channel magnetoencephalogram signal in the x, y, and z axis directions. Then the total multi-channel magnetoencephalogram signal of n sensors is expressed as: （2） Concatenate the values of the three coordinate axes (x, y, and z) of n sensors in series to form a unit direction matrix , and solve the homogeneous field projection matrix M based on this: （3） where I is the identity matrix, is the pseudo-inverse matrix of.

4. The method for suppressing magnetoencephalogram background noise based on spatio-temporal domain projection filtering according to claim 1, characterized in that: In the said Step 3, the specific steps of extracting and separating the MEG nerve signals are as follows: ① Model the multi-channel MEG signals as follows: （4） （5） For multi-channel magnetoencephalogram measurement signals, A is the brain nerve signal, and is the background noise; Among them, the background noise further includes noise with a distance from the sensor less than or equal to the head size and noise with a distance from the sensor much greater than the head size ; where the noise with a distance from the sensor much larger than the head size is zero in the homogeneous domain, and the final model is: （6） M is the homogeneous domain sub-projection matrix obtained in Step 2; For the convenience of writing, is denoted as , is denoted as ; ② Perform space-domain projection filtering on the multi-channel resting-state noise signals to preliminarily separate the signals inside the brain and the external interference signals; The process of space-domain projection filtering is: Perform singular value decomposition on: （7） N is the number of sensors; is singular value of matrix; Arrange the singular values in descending order, and select the corresponding eigenvectors in the spatial domain for the first p singular values: （8） Obtain projection operator ; Then, the multi-channel magnetoencephalogram signals are subjected to spatial domain projection filtering through equations (9) and (10) to obtain preliminary internal brain signals and external interference signals : （9） （10） I is the identity matrix; ③ Perform time-domain projection filtering on the signals inside the brain and the external interference signals; Internal signals of the brain and external interference signals are subjected to singular value decomposition: （11） （12） be the singular value of, be the singular value of; Arrange the singular values , in descending order, and select the eigenvectors in the time domain corresponding to the first q singular values and the eigenvectors in the time domain corresponding to the first k singular values : （13） （14） For their common intersection perform singular value decomposition: （15） where the singular value is the common intersection between the subspace and the magnetoencephalogram neural signal ; Retain the eigenvectors corresponding to the first r singular values of : （16） For the first r columns of the right matrix of singular value decomposition; Internal brain signals and external interference signals time-domain projection filtering operator is obtained from Equation (17): （17） In the formula is the corresponding eigenvector ④Finally, by using Equation (18) and , the purpose of suppressing background noise is achieved; （18） In the formula represents the magnetoencephalogram nerve signal after denoising.

5. An electronic device, characterized in that: It includes a processor, a communication interface, a memory, and a communication bus. Among them, the processor, the communication interface, and the memory complete mutual communication through the communication bus; The memory is used to store computer programs; The processor is used to execute the computer programs stored on the memory, and when executed, it implements the method described in any one of claims 1-4.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that: The computer program, when executed by the processor, implements the method described in any one of claims 1-4.

Images