Methods, apparatuses, and media for removing spike noise from multi-channel magnetocardiography signals

Abstract

The application belongs to the technical field of magnetocardiogram signal processing, and particularly relates to a method, device and medium for removing impact noise of multi-channel magnetocardiogram signals, comprising the following steps: S10, acquiring multi-channel magnetocardiogram signals collected by a magnetocardiograph; S20, based on the multi-channel magnetocardiogram signals, modeling the spatial relationship between channels by using a graph method, fitting the spatial distribution of the multi-channel magnetocardiogram signals containing noise signals by using a symmetric alpha stable distribution model, and obtaining a graph signal representation of the noisy magnetocardiogram signals; S30, constructing an lp regularization target function based on the graph signal representation, optimizing and solving the target function by using an iteratively reweighted least squares method, and obtaining the denoised magnetocardiogram signals. The method can remove the impact noise of individual channels in the multi-channel magnetocardiogram signals, and improve the signal-to-noise ratio of the magnetocardiogram signals; and the spatial denoising is helpful to improve the signal-to-noise ratio while completely retaining the time domain characteristics of the signals, and has no requirement for the number of time domain samples, and the detection efficiency is higher.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of magnetocardiogram signal processing, and particularly relates to a method for removing impact noise of multi-channel magnetocardiogram signals. BACKGROUND

[0002] Magnetocardiogram detection is a non-contact and non-invasive biological signal detection technology. Magnetocardiogram signals are extremely weak, and are easily affected by environmental noise, instrument noise, human eye movement and respiratory noise in actual detection.

[0003] A Chinese patent application with the application number CN202110202791.6 discloses a magnetocardiogram signal background noise S transform domain removal method and system, and specifically discloses the following contents: collecting noise data points; transforming into a two-dimensional time-frequency matrix by using an S transform method; drawing a two-dimensional time-frequency surface of the maximum value of the matrix; taking a set confidence surface of the maximum value surface as a critical interface for distinguishing noise and magnetocardiogram signals; comparing the relationship between the two-dimensional time-frequency matrix of the magnetocardiogram signals and the critical interface and modifying the two-dimensional time-frequency matrix of the magnetocardiogram signals; and performing inverse S transform on the modified two-dimensional time-frequency matrix of the magnetocardiogram signals to obtain the magnetocardiogram signals after removing the background noise. In the patent application, it is assumed that the noise conforms to a Gaussian distribution model, so impact noise cannot be removed; only single-channel magnetocardiogram signals are investigated, which is quite different from the multi-channel situation in actual magnetocardiogram detection; and up to thousands of signals need to be collected, which will result in too long actual detection time and more complexity, and does not meet the requirements of magnetocardiogram clinical application.

[0004] Therefore, how to remove impact noise in multi-channel magnetocardiogram signals without causing too long magnetocardiogram detection time is a key problem to be solved by those skilled in the art. SUMMARY

[0005] (I) Technical problem to be solved

[0006] In view of the above-mentioned shortcomings and deficiencies of the prior art, the application provides a method, device and medium for removing impact noise of multi-channel magnetocardiogram signals.

[0007] (II) Technical solution

[0008] To achieve the above-mentioned purposes, the application adopts the following technical solution:

[0009] In a first aspect, the application provides a method for removing impact noise of multi-channel magnetocardiogram signals, which comprises the following steps:

[0010] S10, acquiring multi-channel magnetocardiogram signals collected by a magnetocardiogram instrument;

[0011] S20, based on the multi-channel magnetocardiogram signal, modeling the spatial relationship between channels by using a graph method, fitting the spatial distribution of the multi-channel magnetocardiogram signal containing noise signals by using a symmetric alpha stable distribution model, and obtaining a graph signal representation of the noisy magnetocardiogram signal based on the symmetric alpha stable distribution model;

[0012] S30, constructing an lp regularization objective function based on the graph signal representation, and optimizing and solving the objective function by using an iterative reweighted least squares method to obtain a denoised magnetocardiogram signal.

[0013] Optionally, the graph signal representation of the noisy magnetocardiogram signal is a weighted adjacency matrix obtained based on the symmetric alpha stable distribution model.

[0014] Optionally, the step of solving the weighted adjacency matrix comprises:

[0015] The fractional low order moment (FLOM) of the symmetric alpha stable distribution model is calculated by the following formula:

[0016]

[0017] wherein, is the element of the i-th row and the j-th column of the fractional low order moment, α is the distribution characteristic index, x i，k represents the k-th element of the vector x i represents the vector x j is the dispersion coefficient under the characteristic index α, |·| p represents the p-vector of the matrix, and K is the sampling number;

[0018] The fractional low order moment is taken as the covariance matrix;

[0019] The covariance matrix is symmetrically normalized to obtain a weighted adjacency matrix, comprising:

[0020] The covariance matrix C FLOM is normalized by the following formula:

[0021]

[0022] The normalized covariance matrix is symmetrized by the following formula:

[0023]

[0024] The normalized C symm is divided by its maximum eigenvalue λ max to obtain a weighted adjacency matrix:

[0025] A=C symm / |λ max |​

[0026] wherein A is a weighted adjacency matrix.

[0027] Optionally, in the process of solving the fractional lower order moment, a method of minimizing the covariance standard deviation of the fractional lower order moment is used to obtain the optimal estimation value of p corresponding to the partial value of alpha, and then the value of p is set as a function of alpha by linear interpolation, so as to obtain the optimal value of p through the value of alpha.

[0028] Optionally, when the symmetric alpha stable distribution model is used to fit the spatial distribution of the multi-channel magnetocardiogram signal containing noise signals, a maximum likelihood method is used to estimate the distribution parameters of the noisy magnetocardiogram signal.

[0029] Optionally, S30 comprises:

[0030] S31, taking p as a to-be-optimized parameter, constructing an lp regularization objective function based on the weighted adjacency matrix, the objective function being:

[0031]

[0032] wherein b is a regularization parameter, A is an adjacency matrix, is the denoised magnetocardiogram signal, and N is the number of channels.

[0033] S32, based on the objective function, obtaining the multi-channel magnetocardiogram signal by an iterative reweighted least squares method;

[0034] S33, based on the obtained multi-channel magnetocardiogram signal, obtaining a noise-free effective signal by gradient descent in each channel;

[0035] S34, based on the obtained noise-free effective signal, iteratively performing steps S31-S32 until a preset iteration number is reached, to obtain the denoised magnetocardiogram signal.

[0036] Optionally, S32 comprises:

[0037] Let the initial value of iteration be S (0) The relationship with the original noisy data X is:

[0038] S (0) =X+∈

[0039] wherein ∈ is a constant much smaller than 1, used to prevent numerical errors;

[0040] Let the iteration round be t, first perform least squares solution to calculate the residual r of each column r=S(k,:) (t) -X(k,:)(k=1,...,K), wherein the updated value of S is:

[0041] S(k,:)(t) = ((D T D) + b(I-A) H (I-A) -1 (D T D)X(k, : ) T

[0042] wherein I is a unit matrix, and D is a diagonal matrix obtained by residual transform:

[0043] D = diag(|r + ∈| (p-2) )

[0044] The least square solution S based on iterative reweighting is calculated (t) as the guessed initial value of the channel gradient descent method.

[0045] Optionally, S10 comprises:

[0046] Obtaining original multi-channel magnetocardiogram signals collected by a magnetocardiograph;

[0047] Removing damaged channel signals from the original multi-channel magnetocardiogram signals to obtain multi-channel magnetocardiogram signals;

[0048] Scaling the multi-channel magnetocardiogram signals to make the intensity of the impact noise meet a preset intensity range.

[0049] In a second aspect, the embodiments of the present application provide an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the computer program is executed by the processor, the steps of the method for removing impact noise of multi-channel magnetocardiogram signals according to any one of the first aspect are implemented.

[0050] In a third aspect, the embodiments of the present application provide a computer readable storage medium, and the computer readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for removing impact noise of multi-channel magnetocardiogram signals according to any one of the first aspect are implemented.

[0051] (III) Beneficial Effects

[0052] The beneficial effects of the present application are: the present application provides a method, device and medium for removing impact noise of multi-channel magnetocardiogram signals, wherein the method comprises: S10, acquiring multi-channel magnetocardiogram signals collected by a magnetocardiograph; S20, based on the multi-channel magnetocardiogram signals, modeling the spatial relationship between channels by using a graph method, fitting the spatial distribution of the multi-channel magnetocardiogram signals containing noise signals by using a symmetric alpha stable distribution model, and obtaining a graph signal representation of the noisy magnetocardiogram signals; S30, constructing an lp regularization objective function based on the graph signal representation, and solving the objective function by using an iteratively reweighted least squares method to obtain a denoised magnetocardiogram signal. The method of the present application can remove impact noise of individual channels in the multi-channel magnetocardiogram signals, and improve the signal-to-noise ratio of the multi-channel magnetocardiogram signals; by using spatial denoising, the signal time domain characteristics can be completely retained while improving the signal-to-noise ratio, and the number of time domain samples is not required, and the detection efficiency is higher. BRIEF DESCRIPTION OF DRAWINGS

[0053] The present application is described with the help of the following drawings:

[0054] Figure 1 A flowchart of a method for removing impact noise of multi-channel magnetocardiogram signals in an embodiment of the present application;

[0055] Figure 2 A flowchart of a method for removing impact noise of multi-channel magnetocardiogram signals in another embodiment of the present application;

[0056] Figure 3 A result example diagram of a method for removing impact noise of multi-channel magnetocardiogram signals in another embodiment of the present application;

[0057] Figure 4 An architecture diagram of an electronic device in another embodiment of the present application. DETAILED DESCRIPTION

[0058] In order to better explain the present application and facilitate understanding, the present application is described in detail by specific embodiments in combination with the drawings. It can be understood that the specific embodiments described below are only used to explain the related application, and are not a limitation on the application. In addition, it should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict; in order to facilitate description, only the parts related to the application are shown in the drawings.

[0059] The present application provides a method, device and medium for removing impact noise of multi-channel magnetocardiogram signals, which are described in detail by embodiments.

[0060] Embodiment one

[0061] Figure 1A flowchart of a method for removing impulse noise from multi-channel magnetocardiogram signals in an embodiment of the present application is shown in FIG. 1. Figure 1 The method for removing impulse noise from multi-channel magnetocardiogram signals in the embodiment includes the following steps.

[0062] S10, acquiring multi-channel magnetocardiogram signals collected by a magnetocardiograph.

[0063] S20, based on the multi-channel magnetocardiogram signals, modeling the spatial relationship between channels using a graph method, fitting the spatial distribution of the multi-channel magnetocardiogram signals containing noise signals using a symmetric alpha stable distribution model, and obtaining a graph signal representation of the noisy magnetocardiogram signals.

[0064] S30, constructing an lp regularization objective function based on the graph signal representation, and optimizing and solving the objective function by an iteratively reweighted least squares method to obtain a denoised magnetocardiogram signal.

[0065] The method for removing impulse noise from multi-channel magnetocardiogram signals in the embodiment can remove impulse noise in individual channels of the multi-channel magnetocardiogram signals by using a symmetric alpha stable distribution model to describe the distribution of noisy magnetocardiogram signals, thereby improving the signal-to-noise ratio of the magnetocardiogram signals. The use of spatial denoising helps to improve the signal-to-noise ratio while preserving the time-domain characteristics of the signals completely, and has no requirement for the number of time-domain samples, thereby improving the detection efficiency.

[0066] To better understand the present application, the steps in the embodiment are described below.

[0067] The embodiment S10 can include the following steps.

[0068] Acquiring original multi-channel magnetocardiogram signals collected by a magnetocardiograph.

[0069] Removing damaged channel signals from the original multi-channel magnetocardiogram signals to obtain multi-channel magnetocardiogram signals.

[0070] Scaling the multi-channel magnetocardiogram signals to make the intensity of the impulse noise meet a preset intensity range.

[0071] In the embodiment S20, the adjacency matrix of the noisy graph signal is solved based on the symmetric alpha stable distribution model, which is used as a representation of the node connection relationship in the graph signal representation.

[0072] The embodiment S30 can include the following steps.

[0073] S31, taking p as an optimization parameter, constructing an lp regularization objective function based on the weighted adjacency matrix, and the objective function is as follows:

[0074]

[0075] wherein b is a regularization parameter, and A is an adjacency matrix. N is the number of channels, and the denoised magnetocardiogram signal is obtained.

[0076] S32, based on the target function, the multi-channel magnetocardiogram signal is obtained by iterative reweighted least squares method.

[0077] S33, based on the obtained multi-channel magnetocardiogram signal, the noise-free effective signal is obtained by gradient descent in each channel.

[0078] S34, based on the obtained noise-free effective signal, steps S31-S32 are iteratively executed until a preset iteration number is reached, and the denoised magnetocardiogram signal is obtained.

[0079] Embodiment two

[0080] The execution subject of the embodiment can be a data acquisition and processing module of a magnetocardiograph, and the control system can include a memory and a processor. In other embodiments, the execution subject can also be other electronic devices that can achieve the same or similar functions, and the embodiment does not limit this.

[0081] Figure 2 The flowchart of the method for removing impact noise of multi-channel magnetocardiogram signal in another embodiment of the application is shown. Based on embodiment one, the specific implementation process of the embodiment is described in detail. The following will be combined with Figure 2 The steps of the method of the embodiment are described in detail.

[0082] S1, obtaining the original multi-channel magnetocardiogram signal collected by the magnetocardiograph.

[0083] The magnetocardiograph is used to record the magnetocardiogram and record the magnetic field in the cardiac cycle. It captures the magnetic field intensity generated by the electrical activity of the human heart in different regions above the heart through a multi-channel magnetic field sensor array arranged above the patient's chest, and the weak signal obtained is converted into raw data for imaging by the data acquisition and processing module; invalid data and artifacts are removed, and finally two-dimensional and three-dimensional cardiac magnetic field reconstruction is performed based on specific algorithms by imaging software, and the required indicators and data for related diagnosis are output.

[0084] In the embodiment, the magnetocardiogram signal of each channel is collected by one magnetic field sensor in the magnetic field sensor array. In step S1, the original multi-channel magnetocardiogram signal is also preprocessed, specifically, the obviously damaged channel signal is removed from the magnetocardiogram signal collected by the multi-channel magnetic field sensor array, and the other channel signals are used as the multi-channel magnetocardiogram signal, denoted as X=[x1,..., xn]∈Rn. N K×N , wherein x i ∈R K ​Xi represents the original magnetocardiogram signal of the i-th channel, R represents a matrix, K represents the number of signal samples, and N represents the number of channels of the multi-channel magnetocardiogram signal.

[0085] Preferably, the amplitude of the noise distribution model established in the method is between [0, 1], and the order of magnitude of the magnetocardiogram signal is 10 -10 T is about 1, so in order to improve the denoising effect, the original magnetocardiogram signal can be scaled by a proper ratio, so that the intensity value of the impact noise remains about 1, and after denoising, it is restored to the same ratio.

[0086] S2, model the spatial relationship between channels using a graph method.

[0087] In this embodiment, S2 specifically includes:

[0088] S21, the signal components of the multi-channel magnetocardiogram signal X are represented as:

[0089] X = S + W (1)

[0090] Wherein, S ∈ R K×N represents a noise-free effective signal matrix, W ∈ R K×N represents a noise signal matrix.

[0091] S22, a graph signal representation of the noise signal matrix is established to obtain a noise graph signal.

[0092] In this embodiment, the number of channels of the multi-channel magnetocardiogram signal is N, and the graph signal representation corresponding to the magnetocardiogram signal X is:

[0093] G = (V, E) (2)

[0094] Wherein, V and E represent the node set and edge set of the graph, respectively.

[0095] Here, is the set of vertices, and E ∈ C N×N represents the connection between the vertices, and on this basis, with the original noisy signal, the method of this embodiment further describes the relationship between the channels by calculating the corresponding adjacency matrix.

[0096] S23, solve the adjacency matrix of the graph signal.

[0097] In order to represent the mutual dependence relationship between different nodes (channels), in this embodiment, the association weight between each point in the graph is encoded as an adjacency matrix A ∈ R N×N .

[0098] This embodiment uses a weighted adjacency matrix, which not only encodes the existence of the connection between two nodes, but also encodes the strength of the connection, so that more complex relationships between nodes in the graph can be captured.

[0099] Except for the case of Cauchy distribution when the characteristic exponent α = 1 and Gaussian distribution when α = 2, there is no closed-form expression for the symmetric α-stable distribution. In this embodiment, the analysis is performed under the condition of the location parameter δ = 0. Since the remaining distribution functions lack analytical solutions except for the Gaussian distribution and the Cauchy distribution, the method of the present application proposes a statistical weighted adjacency matrix based on fractional lower-order moments to describe different distribution noises.

[0100] S231, fitting the distribution of the noise map signal using the symmetric α-stable distribution model, and estimating the distribution parameters of the source signal using the maximum likelihood method.

[0101] Here, the source signal is a multi-channel magnetocardiogram signal containing effective signals and impulse noises.

[0102] This embodiment assumes that the noise matrix W conforms to the independent and identically distributed symmetric α-stable model. The characteristic expression of the symmetric α-stable model family is:

[0103] φ(t) = exp(iδt - γ α |t| α ) (3)

[0104] where α is the characteristic exponent, 0 < α ≤ 2, δ ∈ R is the location parameter, and γ is the dispersion.

[0105] α is used to represent the sharpness of the impulse signal, and the smaller α is, the higher the sharpness of the represented distribution. δ represents the center point position of the distribution. γ determines the degree to which the stable distribution random variable deviates from its mean or median, and the larger the value of γ, the more concentrated the distribution.

[0106] In this embodiment, the distribution parameters of the source signal are α and γ, and independent parameter estimation is performed for the multi-channel parameters at each time. Assuming that the signal sampling length is K, the parameter arrays α ∈ R 1×K and γ ∈ R 1×K are obtained.

[0107] It should be noted that the maximum likelihood method is a known prior art to those skilled in the art and will not be described in detail in this application. For the estimation of the distribution parameters of the source signal, in addition to using the maximum likelihood estimation method, any existing estimation method such as the least squares method can also be used, and the estimation method is not specifically limited in this embodiment.

[0108] S232, calculating the fractional lower-order moment of the symmetric α-stable distribution model.

[0109] Most of the symmetric α-stable distribution models lack second-order moments, but all have fractional lower-order moments. For a distribution X ~ f α (γ, δ = 0), the fractional lower-order moment can be expressed as:

[0110] E{|X| p}=(C(p,α)·γ) p ,0<p<α (4)

[0111] where

[0112]

[0113] where, f α (γ, δ = 0) is the symmetric alpha stable distribution with parameters γ and δ taking 0, p is the p parameter in lp norm.

[0114] The method introduces the covariates, which plays a similar role to the covariance in the Gaussian case. For the same symmetric alpha stable distribution of random variables X, Y (that is, the alpha parameters of the variable distribution are the same), the dispersion is γ X , γ Y , and the covariates of X and Y are defined as:

[0115]

[0116] where, the calculation symbol:

[0117] z ＜a> = |z| a sign(z)z∈R, a≥0,

[0118] z ＜a> =[|z| a sign(z1), …, |z N | a sign(z k )]z∈R K

[0119] In the discrete case, the fractional lower moments of two independent vectors x i , x j ∈R K of the same symmetric alpha stable distribution are expressed as:

[0120]

[0121] where, is the element in the i-th row and j-th column of the fractional lower moments, α is the characteristic exponent of the distribution, x i，k ∈R represents the k-th element of the vector x i , represents the dispersion coefficient of the vector xj under the characteristic exponent α, |·| p represents the p vector of the matrix, and K is the sampling number.

[0122] α and can be estimated according to the method in S231.

[0123] S233、According to the score low-order moment, a covariance matrix is obtained, the covariance matrix is symmetrically normalized, and a weighted adjacency matrix is obtained.

[0124] Since the score low-order moment contains the correlation information between elements, the score low-order moment is taken as the covariance matrix C FLOM . The covariance matrix C FLOM is converted into a weighted adjacency matrix.

[0125] The covariance matrix C FLOM is normalized:

[0126]

[0127] Then, symmetry is performed:

[0128]

[0129] Finally, by normalizing C symm , dividing by its maximum eigenvalue λ max , the final weighted adjacency matrix is obtained:

[0130] A = C symm / | λ max | (10)

[0131] In the process of solving the weighted adjacency matrix A, the parameter p needs to be optimized and selected. P depends on the characteristic index a. In this embodiment, the method of minimizing the covariance standard deviation of the score low-order moment is adopted to obtain the optimal estimation value of p corresponding to part of a values. Then, by linear interpolation, the value of p is set as a function of a, so as to obtain the optimal value of p through a value.

[0132] By solving the optimal value of the parameter p, non-Gaussian noise under different distributions can be solved in a targeted manner, and good effects can be obtained for noise in different environments.

[0133] S3, constructing an lp regularization objective function based on the weighted adjacency matrix, and solving the denoising signal by an iterative reweighted least squares method.

[0134] In this embodiment, S3 includes the following steps:

[0135] S31, taking p as a to-be-optimized parameter, and constructing an lp regularization objective function based on the weighted adjacency matrix;

[0136] In this embodiment, the denoising problem is represented as a regularization lp parameter optimization problem, and the to-be-optimized objective function is:

[0137]

[0138] where b is a regularization parameter, A is an adjacency matrix, is the denoised magnetocardiogram signal.

[0139] The regularization parameter can control the smoothness of signal denoising, and the value range of p is 0

[0140] The lp norm of the matrix is defined as:

[0141]

[0142] S32, the multi-channel magnetocardiogram signal is solved by the iterative reweighted least squares method;

[0143] S33, based on the multi-channel magnetocardiogram signal solved, the noise-free effective signal is solved by gradient descent in each channel;

[0144] S34, based on the noise-free effective signal solved, steps S31-S32 are iteratively executed until a preset iteration number is reached, and the denoised magnetocardiogram signal is obtained.

[0145] For the solution of , this embodiment proposes a hybrid optimization scheme, which first solves the magnetocardiogram signal sample based on the iterative reweighted least squares (IRLS), and then uses a gradient descent-based solution on different channels to solve the non-convex problem caused by p optimization in the regularization expression. When the maximum iteration number T is reached, the two optimization steps are terminated simultaneously.

[0146] Let the initial value of iteration be S (0) , and the relationship with the original noisy data X is:

[0147] S (0) = X + ∈ (13)

[0148] Where ∈ is a constant much smaller than 1, used to prevent numerical errors.

[0149] Let the iteration round be t, first calculate the least squares solution to calculate the residual of each column r = S(k, :) (t) - X(k, :) (k = 1,..., K), where the updated value of S is:

[0150] S(k, :) (t) = ((D T D) + b(I-A) H (I-A)) -1 (D T D) X(k, :) (14)

[0151] where I is an identity matrix, and D is a diagonal matrix obtained by residual transform:

[0152] D = diag(|r + ∈| (p-2) ) (15)

[0153] Here, p takes the p estimate corresponding to the kth value in the array α obtained in step S231.

[0154] The least square solution S based on iterative re-weighting is calculated (t) as the initial guess of the channel gradient descent method. The update expression of the channel gradient descent method is:

[0155]

[0156] where l is the regularization rate, is the objective function i.e. the denoised signal to be solved, and the specific algebraic calculation expression is:

[0157]

[0158] where and The elements of and correspond to the n channels in the magnetocardiogram signal, and are diagonal matrices calculated by the following expressions, respectively:

[0159]

[0160]

[0161] The above iteration process is repeated until the maximum number of iterations T is reached, and the final denoised magnetocardiogram signal is given by

[0162] Figure 3 is a result example diagram of removing impact noise from a multi-channel magnetocardiogram signal in another embodiment of the present application, Figure 3 shows the denoising results of a group of spatial domain impact noise containing magnetocardiogram signals by the method of the present embodiment, Figure 3 (a) in the figure is the signal curve after denoising, Figure 3 (b) in the figure is the original signal curve, both horizontal axes are time sampling points, the sampling interval is 0.05s, and both vertical axes are sensor signal values, i.e. magnetic induction intensity, unit: T Tesla; Figure 3Fig. 3 is a diagram showing the change of signal-to-noise ratio before and after denoising, where the horizontal axis represents time, the vertical axis represents signal-to-noise ratio in dB, the thin line represents the change of signal-to-noise ratio before processing, and the thick line represents the change of signal-to-noise ratio after processing. The magnetocardiogram signal in the diagram has 32 channels, the time domain sampling length K = 100, the regularization learning rate l is set to 0.01, and the signal intensity is scaled to [-10, 10] by preprocessing. The signal-to-noise ratio of each sampling point is shown in Fig. 3(c). It can be seen from the diagram that the signal-to-noise ratio after processing is significantly improved, indicating that the method of the embodiment has a good effect on removing spatial impact signals while not affecting the effective signals. Figure 3 As can be seen from the comparison between (a) and (b), the impact noise of individual channels is effectively removed without affecting the overall trend of the time domain signal. Figure 3 As can be seen from (c), the spatial signal-to-noise ratio of each sampling point is shown, and it can be seen that the signal-to-noise ratio after processing is significantly improved, indicating that the method of the embodiment has a good effect on removing spatial impact signals while not affecting the effective signals.

[0163] The method of the embodiment models the spatial relationship between channels by the graph method. First, the symmetric alpha stable distribution model is used to fit the distribution of the noisy magnetocardiogram signal. Then, the fractional lower order moment of the signal distribution is obtained by using the covariates of the noisy signal, and the optimization parameter p is introduced according to the characteristic exponent a. After symmetric normalization transformation, the weighted adjacency matrix is obtained. Finally, based on the adjacency matrix, the original signal is optimized by regularization to obtain the denoised signal. This method can remove the impact noise of individual channels in the multi-channel magnetocardiogram signal sampling, has a better effect on removing non-Gaussian distributed noise, improves the signal-to-noise ratio and quality of the multi-channel magnetocardiogram signal, and helps to improve the signal-to-noise ratio while retaining the time domain characteristics of the signal. Moreover, the method has no requirement for the number of time domain samples and has high detection efficiency.

[0164] Embodiment Three

[0165] The second aspect of the present application provides an electronic device through embodiment three, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the computer program implements the steps of the method for removing impact noise of multi-channel magnetocardiogram signals according to any one of the above embodiments when executed by the processor.

[0166] Figure 3 The architecture of the electronic device in another embodiment of the present application is shown in the schematic diagram.

[0167] Figure 3 The electronic device shown can include at least one processor 101, at least one memory 102, at least one network interface 104, and other user interfaces 103. The various components in the electronic device are coupled together through a bus system 105. It can be understood that the bus system 105 is used to realize the connection and communication between the components. In addition to the data bus, the bus system 105 also includes a power bus, a control bus, and a status signal bus. However, for the sake of clarity, only the data bus is shown in the figure. Figure 3The various buses are shown as bus system 105.

[0168] The user interface 103 can include a display, a keyboard, or a pointing device (e.g., a mouse, a trackball, or a touchpad).

[0169] It can be understood that the memory 102 in the embodiment can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. The non-volatile memory can be a Read-Only Memory (ROM), a Programmable ROM (PROM), an Erasable PROM (EPROM), an Electrically EPROM (EEPROM), or a flash memory. The volatile memory can be a Random Access Memory (RAM) used as an external cache. By way of example, and not limitation, many forms of RAM are available, such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced SDRAM (ESDRAM), Sync Link DRAM (SLDRAM), and Direct Rambus RAM (DRRAM). The memory 62 described herein is intended to include, without being limited to, these and any other suitable types of memory.

[0170] In some embodiments, the memory 102 stores the following elements, executable units or data structures, or a subset of them, or an extended set of them: an operating system 1021 and an application program 1022.

[0171] The operating system 1021 contains various system programs, such as a framework layer, a core library layer, a driver layer, etc., for implementing various basic services and processing hardware-based tasks. The application program 1022 contains various application programs for implementing various application services. The program for implementing the method of the embodiment of the application can be included in the application program 1022.

[0172] In the embodiments of the present application, the processor 101 executes the programs or instructions stored in the memory 102, specifically, the programs or instructions stored in the application program 1022, to implement the method steps of the first aspect.

[0173] The method disclosed in the embodiments of the present application can be applied to or implemented by the processor 101. The processor 101 can be an integrated circuit chip having a signal processing capability. In the implementation process, the steps of the above method can be completed by hardware integrated logic circuit or software form instructions in the processor 101. The processor 101 mentioned above can be a general processor, a digital signal processor, an application specific integrated circuit, a ready programmable gate array or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component. The disclosed methods, steps and logic block diagrams in the embodiments of the present application can be implemented or executed. The general processor can be a microprocessor or any conventional processor. The steps of the method disclosed in combination with the embodiments of the present application can be directly embodied as a hardware coding processor for execution, or a combination of hardware and software units in the coding processor for execution. The software unit can be located in a random access memory, a flash memory, a read only memory, a programmable read only memory or an electrically erasable programmable memory, a register or other mature storage medium in the art. The storage medium is located in the memory 102, and the processor 101 reads the information in the memory 102 and completes the steps of the above method in combination with the hardware.

[0174] In addition, in combination with the method for removing impact noise of multi-channel cardiac magnetic signal in the above embodiments, the embodiments of the present application can provide a computer readable storage medium, and the computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement any one of the methods for removing impact noise of multi-channel cardiac magnetic signal in the above method embodiments.

[0175] It should be noted that in the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The application can be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. The use of the word "at least" followed by a list of elements does not exclude the presence of one or more additional elements not listed. The word "first", "second", "third", etc. does not imply any order. These words are used to name different elements.

[0176] In addition, it should be noted that, in the description of this specification, the description of the terms "one embodiment", "some embodiments", "embodiment", "example", "specific example" or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.

[0177] 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 after learning the basic creative concept. Therefore, the claims should be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0178] Obviously, those skilled in the art may make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if such modifications and variations fall within the scope of the claims and their equivalents, the present invention shall also include such modifications and variations.

Claims

1. A method for removing impulse noise from multi-channel magnetocardiographic signals, characterized in that: The method includes: S10, obtaining multi-channel magnetocardiograph signals collected by a magnetocardiograph, Specifically, Obtaining original multi-channel magnetic cardiogram signals collected by a magnetocardiograph; removing damaged channel signals from the original multi-channel magnetocardiographic signal to obtain a multi-channel magnetocardiographic signal; Scaling the multi-channel magnetocardiographic signal in a scaling range of [-10, 10] so that the intensity of the impact noise meets a preset intensity range; S20, based on the multi-channel MCG signal, using a graph method to model the spatial relationship between channels, using a symmetric α-stable distribution model to fit the spatial distribution of the multi-channel MCG signal containing noise signals, and solving the symmetric α-stable distribution model to obtain a graph signal representation of the noisy MCG signal, The graph signal representation of the noisy magnetocardiographic signal is a weighted adjacency matrix obtained by solving the symmetric α-stable distribution model. When the symmetric α-stable distribution model is used to fit the spatial distribution of multi-channel MCG signals containing noise signals, the maximum likelihood method is used to estimate the distribution parameters of the noisy MCG signals. S30 , constructing an lp regularized objective function based on the graph signal representation, optimizing and solving the objective function through iterative reweighted least squares method to obtain a denoised magnetocardiographic signal.

2. The method for removing impulse noise from multi-channel magnetocardiographic signals according to claim 1, characterized in that: The steps of solving the weighted adjacency matrix include: The fractional low-order moment FLOM of the symmetric α-stable distribution model is calculated by the following formula: Among them, x i is the original magnetic cardiogram signal of the i-th channel, x j is the original magnetic cardiogram signal of the jth channel, x j，k is the vector x j The kth element of is the element of the i-th row and j-th column of the fractional low-order moment, α is the distribution characteristic index, x i，k ∈R represents the vector x i The kth element of Represents vector x j The dispersion coefficient under the characteristic index α, |·| p represents the p vector of the matrix, K is the number of sampling times; The fractional lower-order moments are used as covariance matrices; The covariance matrix is ​​symmetric-normalized to obtain a weighted adjacency matrix, including: The covariance matrix C is calculated by the following formula FLOM Normalize: The normalized covariance matrix is ​​symmetrized by the following formula: By normalizing C symm , divided by its maximum eigenvalue λ max , and get the weighted adjacency matrix: A=C symm / |l max | Where A is the weighted adjacency matrix.

3. The method for removing impulse noise from multi-channel magnetocardiographic signals according to claim 2, characterized in that: In the process of solving the fractional low-order moments, the method of minimizing the covariate standard deviation of the fractional low-order moments is adopted to find the optimal estimated value of p corresponding to some α values. Then, through linear interpolation, the value of p is set as a function of α, and the optimal value of p is obtained through the α value.

4. The method for removing impulse noise from multi-channel magnetocardiographic signals according to claim 1, characterized in that: The S30 includes: S31, taking p as the parameter to be optimized, constructing the objective function of lp regularization based on the weighted adjacency matrix, the objective function is: Where S∈R K×N is the noise-free effective signal matrix, X is the original noisy data, p is the parameter to be optimized, b is the regularization parameter, A is the adjacency matrix, is the denoised magnetic cardiogram signal, N is the number of channels; S32. Based on the objective function, obtain a multi-channel magnetocardiographic signal by using an iterative reweighted least squares method; S33, based on the obtained multi-channel magnetocardiographic signal, obtain a noise-free valid signal by performing gradient descent on each channel; S34 , based on the noise-free valid signal obtained by solution, iteratively execute steps S31 - S32 until a preset number of iterations is reached to obtain a denoised magnetocardiographic signal.

5. The method for removing impulse noise from multi-channel magnetocardiographic signals according to claim 4, characterized in that: S32 includes: Let the initial value of the iteration be S (0) , and the relationship with the original noisy data X is: S (0) =X+∈ Where, ∈ is a constant much smaller than 1 to prevent numerical errors; Assume that the number of iterations is t, and first perform the least squares solution to calculate the residual r = S(k,:) for each column. (t) -X(k,:) (k=1,...,K), where the updated value of S is: S(k,:) (t) =((D T D)+b(I-A) H (I-A)) -1 (D T D)X(k,:) T Where A is the adjacency matrix, b is the regularization parameter, p is the parameter to be optimized, K is the number of samples, X(k,:) is the k-th row vector of the original noisy signal data X, and X(k,:) T is the transposed matrix of X(k,:), I is the identity matrix, and D is the diagonal matrix obtained by the residual transformation: D=diag(|r+∈| (p-2) ) The least squares solution S based on iterative reweighting is calculated (t) , as the guessed initial value of the channel gradient descent method.

6. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the steps of the method for removing impulse noise from multi-channel magnetic cardiogram signals as described in any one of claims 1 to 5 are implemented.

7. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for removing impulse noise from multi-channel magnetocardiographic signals according to any one of claims 1 to 5.

Images