A Method and System for Suppressing Baseline Drift in ECG Signals Based on Cascaded Singular Spectrum Analysis

Abstract

This invention relates to a method and system for suppressing baseline drift in electrocardiogram (ECG) signals based on cascaded singular spectrum analysis (SSA). The method includes: acquiring ECG signal data affected by baseline drift, setting the number of layers i = 1; and extracting baseline drift noise n from the ECG signal data affected by baseline drift using SSA with a window length of 2. i With ECG signal components o i Its central electrical signal ECG i Obtained through a formula; calculate baseline drift noise n. i With electrocardiogram (ECG) signal i inner product IP i Determine if the inner product has reached its minimum value. If the inner product has reached its minimum value, output the ECG signal of the current layer. i As a final result of suppressing baseline drift, otherwise, the baseline drift noise n i Using ECG signal data affected by baseline drift and incrementing the layer number i by 1, repeat the above extraction and calculation operations until the inner product reaches its minimum value, then output the ECG signal of the current layer. i .

Description

Technical Field

[0001] This invention relates to the field of electrocardiogram (ECG) signal processing, and more specifically to a method and system for suppressing baseline drift in ECG signals based on cascaded singular spectrum analysis. Background Technology

[0002] In recent years, cardiovascular diseases remain among the most common threats to human health. Electrocardiogram (ECG) signal examination is a crucial clinical method for diagnosing cardiovascular diseases. However, during the acquisition process, baseline drift noise is introduced into the ECG signal due to human respiration and electrode slippage. This type of noise has an extremely low frequency, typically below 1 Hz, but a relatively large amplitude. Therefore, baseline drift not only visually causes overall vertical displacement of the ECG, leading to severe distortion of the trajectory line, but also interferes with the detection of the QRS complex and direct diagnosis of the ECG signal.

[0003] Wavelets and filter banks are traditional methods for suppressing baseline drift interference in electrocardiogram (ECG) signals. This approach transforms the signal into the wavelet domain for analysis and processing. By selecting an appropriate threshold, certain wavelet coefficients representative of the ECG signal are retained, and then an inverse transform is performed to obtain the baseline-drift-suppressed ECG signal. The challenge of this method lies in the need for manual selection of certain parameters, such as the wavelet basis, threshold, and number of decomposition levels. If the parameters are not chosen appropriately, the baseline drift suppression effect will be unsatisfactory.

[0004] In recent years, Empirical Mode Decomposition (EMD) has been widely used to correct baseline drift in electrocardiogram (ECG) signals. This method decomposes the disturbed signal into several intrinsic mode functions (IMFs), each located in a different frequency band. By removing low-frequency modes, the baseline-drift-free ECG signal can be obtained. However, issues such as edge effects, mode aliasing, and how to define which IMFs reside in the low-frequency band remain to be resolved in EMD. Singular Spectrum Analysis (SSA) is a subspace-based signal decomposition method. It obtains eigenvectors of the signal matrix through operations such as embedding and singular value decomposition. The signal matrix is ​​then projected onto the subspace expanded by the eigenvectors, and finally, the signal is returned to a one-dimensional signal using a diagonal averaging method. The decomposition result of SSA consists of L components, where L is a manually chosen window length. The difficulty in suppressing baseline drift using SSA lies in selecting an appropriate window length and defining the effective reconstructed components. Currently, there is still no universally applicable and reasonable parameter selection scheme for this method. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for suppressing baseline drift in electrocardiogram (ECG) signals based on cascaded singular spectrum analysis. This method is an adaptive method for suppressing baseline drift in ECG signals, eliminating the need to rely on empirical parameter selection.

[0006] The technical solution of this invention is a method for suppressing baseline drift in electrocardiogram signals based on cascaded singular spectrum analysis, comprising the following steps:

[0007] S100, acquire ECG signal data affected by baseline drift, set layer number i = 1;

[0008] S200 uses Singular Spectrum Analysis (SSA) with a window length of 2 to extract baseline drift noise n from ECG signal data affected by baseline drift. i With ECG signal components o i Its central electrical signal ECG i Through formula get;

[0009] S300, Calculate baseline drift noise n i With electrocardiogram (ECG) signal i Inner product IP i ;

[0010] S400: Determine if the inner product has reached its minimum value. If the inner product has reached its minimum value, output the ECG signal of the current layer. i As a final result of suppressing baseline drift, otherwise, the baseline drift noise n i Using ECG signal data affected by baseline drift and incrementing the layer number i by 1, repeat steps S200 and S300 until the inner product reaches a minimum value, then output the ECG signal of the current layer. i .

[0011] Furthermore, in the above method of the present invention, step S200 includes the following sub-steps:

[0012] S210, the ECG signal data affected by baseline drift is set as a one-dimensional time signal x = [x1…x2]. N ], where N is the length of the one-dimensional time signal, and the one-dimensional time signal x = [x1…x2] is defined as follows: N Transform into a trajectory matrix:

[0013] The trajectory matrix X is a 2×(N-1) dimensional matrix, and the trajectory matrix has two rows to indicate that the window length L is 2;

[0014] S220, for matrix XX T Perform singular value decomposition: XX T =VΛV TTo obtain the eigenvalues ​​and corresponding eigenvectors, where singular values ​​λ1 and λ2 are stored on the main diagonal of matrix Λ, and λ1 > λ2, and the corresponding eigenvectors v1 and v2 are stored in the first and second columns of matrix V, the trajectory matrix X is transformed into the sum of two primary matrices of rank 1:

[0015] S230, set the matrix X1 corresponding to the large singular value λ1 as the baseline drift matrix, and set the matrix X2 corresponding to the small singular value λ2 as the electrocardiogram signal matrix. The matrix sizes of matrices X1 and X2 are the same as those of the trajectory matrix X, both being 2×(N-1).

[0016] S240, the baseline drift matrix X1 and the ECG signal matrix X2 obtained in sub-step S230 are reconstructed into a one-dimensional time signal of length N using the diagonal averaging method. Specifically, the diagonal averaging method is as follows: for a given matrix G, which is a 2×(N-1) dimensional matrix, the matrix elements are G... i,j Set L * =min(2,N-1) and K * =max(2,N-1), the reconstructed one-dimensional time series g = [g1,g2,L,g N Obtained through the diagonal mean formula:

[0017]

[0018] Furthermore, in the above method of the present invention, step S300 includes the following sub-steps:

[0019] S310, Obtain the baseline drift noise n obtained from the i-th layer in step S200. i and electrocardiogram (ECG) i ;

[0020] S320, using formula IP i =(n i ) T (ECG i Calculate the inner product of the i-th layer.

[0021] Secondly, this invention also discloses a baseline drift suppression system for electrocardiogram signals based on cascaded singular spectrum analysis, comprising:

[0022] The acquisition module is configured to acquire ECG signal data affected by baseline drift, and the layer number i = 1 is set.

[0023] The extraction module is configured to extract baseline drift noise n from ECG signal data affected by baseline drift using Singular Spectrum Analysis (SSA) with a window length of 2. i With ECG signal components o i Its central electrical signal ECGi Through formula get;

[0024] The calculation module is configured to calculate the baseline drift noise n. i With electrocardiogram (ECG) signal i Inner product IP i ;

[0025] The iterative output module is configured to determine whether the inner product has reached a minimum value. If the inner product has reached a minimum value, the ECG signal of the current layer is output. i As a final result of suppressing baseline drift, otherwise, the baseline drift noise n i As the ECG signal data affected by baseline drift, the layer number i is incremented by 1, and the extraction and calculation modules are invoked to perform corresponding extraction and calculation until the inner product reaches a minimum value, at which point the ECG signal of the current layer is output. i .

[0026] Finally, the present invention also proposes a non-transitory computer-readable storage medium storing computer instructions thereon. When these instructions are executed by a processor, any step in the above method is performed.

[0027] The beneficial effects of this invention are as follows: This invention utilizes a cascaded singularity spectrum analysis method to iteratively decompose electrocardiogram (ECG) signals containing baseline drift interference. In each decomposition layer, components with small singular values ​​are accumulated as an estimate of the ECG signal; components with large singular values ​​are used as an estimate of the baseline drift noise, and the next round of decomposition continues. When the inner product of the estimated ECG signal and the estimated baseline drift noise reaches a minimum value, the decomposition stops, and the ECG signal of the current round is the final ECG signal with baseline drift suppressed. This invention has the following advantages:

[0028] (1) Since the window length of the cascaded singular spectrum analysis is set to 2, the baseline drift suppression caused by the inappropriate selection of window length in the traditional singular spectrum analysis method is not ideal.

[0029] (2) When the inner product of the baseline drift noise and the ECG signal reaches a minimum, the decomposition is automatically stopped. This criterion is adaptive, thus avoiding the inconvenience of manually selecting parameters and the problem of inappropriate parameter selection.

[0030] (3) Improved the accuracy of baseline drift suppression, effectively preserving ECG signal information while removing baseline drift. Attached Figure Description

[0031] Figure 1 The flowchart of the cascaded singular spectrum analysis of the present invention is shown.

[0032] Figure 2The overall flowchart of the ECG signal baseline drift suppression method based on cascaded singular spectrum analysis of the present invention is shown.

[0033] Figure 3 A schematic diagram of the diagonal average method in step 2 of this embodiment of the invention is shown.

[0034] Figure 4 The diagram shows a clean electrocardiogram (a real electrocardiogram from the MIT-BIT arrhythmia database, record number 100) and a disturbed electrocardiogram with added real baseline drift (real baseline drift data from the MIT-BIT noise database, record number bw).

[0035] Figure 5 The baseline drift noise n in an embodiment of the present invention is shown. i and electrocardiogram (ECG) i The inner product curve with respect to the number of layers.

[0036] Figure 6 The diagram shows a comparison of the time-domain waveforms of the ECG signal after baseline drift suppression and the ECG signal affected by interference in an embodiment of the present invention. Detailed Implementation

[0037] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] Reference Figure 1 and Figure 2 This embodiment provides a method for suppressing baseline drift in electrocardiogram signals based on cascaded singular spectrum analysis, which is carried out according to the following steps:

[0039] S100, Acquire ECG signal data. In this embodiment, a real ECG signal (record number 100) from the MIT-BIT arrhythmia database is used as the clean ECG signal. The duration of the clean ECG signal is 10 seconds, with a total of N = 3600 sampling points. To verify the effectiveness of this technical solution, real baseline drift data (record number bw) from the MIT-BIT noise database is used for noise. The duration of the baseline drift noise is also 10 seconds, with a total of N = 3600 sampling points. The clean ECG signal and the baseline drift data are superimposed to obtain the ECG signal affected by baseline drift. The clean ECG signal and the affected ECG signal are as follows: Figure 4 As shown. Initialize the layer number i = 1;

[0040] S200 uses Singular Spectrum Analysis (SSA) with a window length of 2 (L=2) to extract baseline drift noise n from an ECG signal affected by baseline drift. i With an electrocardiogram signal component o i Its central electrical signal ECG i Through formula The process of obtaining it is as follows: Figure 1 As shown;

[0041] S300, Calculate baseline drift noise n i The inner product of ECG and electrocardiogram signals i ;

[0042] S400: Determine if the inner product has reached its minimum value. If the inner product has reached its minimum value, output the current ECG signal. i As a result of ultimately suppressing baseline drift, otherwise, the baseline drift noise n i Using ECG signal data affected by baseline drift and setting the layer number i to increase by 1, repeat steps S200 and S300 until the inner product reaches a minimum. Figure 5 It can be seen that when the decomposition reaches the 2096th level, the inner product reaches its minimum value, which satisfies the IP condition. 2096 <IP 2095 And IP 2096 <IP 2097 At this time, the baseline-suppressed ECG signal is output. 2096 As a final result, the ECG signal after suppressing baseline drift in this embodiment of the invention. 2096 The time-domain waveform comparison with the interfered ECG signal is shown in the figure below. Figure 6 As shown.

[0043] Furthermore, the method for performing L=2 SSA decomposition on the disturbed data described in step S200 is performed according to the following sub-steps:

[0044] S210: Embedded

[0045] Suppose an electrocardiogram signal affected by baseline drift is x = [x1…x2]. N The length of the matrix is ​​N = 3600. Convert it into a trajectory matrix:

[0046] Where X is a 2×3599 dimensional matrix with two rows, representing a window length L of 2.

[0047] S220: Singular Value Decomposition

[0048] For matrix XX T Perform singular value decomposition: XX T =VΛV TWe can obtain the eigenvalues ​​and their corresponding eigenvectors, where the singular values ​​λ1 and λ2 are stored on the main diagonal of matrix Λ, and λ1 > λ2, while the corresponding eigenvectors v1 and v2 are stored in the first and second columns of matrix V. Therefore, the trajectory matrix X can be transformed into the sum of two primary matrices of rank 1:

[0049] S230: Grouping

[0050] The matrix X1 corresponding to the large singular value λ1 is the baseline drift matrix, and the matrix X2 corresponding to the small singular value λ2 is the electrocardiogram signal matrix. The matrix sizes of X1 and X2 are the same as X, both being 2×(N-1).

[0051] S240: Restructuring

[0052] The baseline drift matrix X1 and the electrocardiogram signal matrix X2 obtained in step S230 are reconstructed into one-dimensional time signals of length N using the diagonal averaging method. Specifically, the diagonal averaging method is as follows: given G as a 2×3599 dimensional matrix, the matrix elements are G... i,j L * =min(2,N-1) and K * =max(2,N-1), i.e., L * =2 and K * =3599, then the reconstructed one-dimensional time series g = [g1, g2, ..., g N This can be obtained through the diagonal mean formula:

[0053]

[0054] That is, through such Figure 3 The diagonal averaging method shown converts the matrix forms of X1 and X2 back into one-dimensional time signals, where X1 is converted into the baseline drift noise n of the i-th layer. i X2 is restored to the signal component of the i-th layer. i , all o i The ECG signal of the i-th layer can be obtained by summing the signals. i .

[0055] Furthermore, step 300 of the present invention, calculating the inner product of baseline drift noise and electrocardiogram signal, includes the following steps:

[0056] Step 310: Obtain the baseline drift noise n obtained from the i-th layer in step 200. i and electrocardiogram (ECG) i ;

[0057] Step 320: Using formula IP i =(n i ) T (ECGi Calculate the inner product of the i-th layer.

[0058] In this embodiment of the invention, the inner product IP of the i-th layer i The curve regarding layer number i is as follows: Figure 5 As shown.

[0059] According to one or more embodiments of the present invention, a baseline drift suppression system for electrocardiogram signals based on cascaded singular spectrum analysis is also proposed, comprising:

[0060] The acquisition module is configured to acquire ECG signal data affected by baseline drift, and the layer number i = 1 is set.

[0061] The extraction module is configured to extract baseline drift noise n from ECG signal data affected by baseline drift using Singular Spectrum Analysis (SSA) with a window length of 2. i With ECG signal components o i Its central electrical signal ECG i Through formula get;

[0062] The calculation module is configured to calculate the baseline drift noise n. i With electrocardiogram (ECG) signal i Inner product IP i ;

[0063] The iterative output module is configured to determine whether the inner product has reached a minimum value. If the inner product has reached a minimum value, the ECG signal of the current layer is output. i As a result of ultimately suppressing baseline drift, otherwise, the baseline drift noise n i As the ECG signal data affected by baseline drift, the layer number i is incremented by 1, and the extraction and calculation modules are called to perform the corresponding extraction and calculation until the inner product reaches a minimum value, at which point the ECG signal of the current layer is output. i .

[0064] Furthermore, in the above system, the extraction module also includes the following sub-modules:

[0065] The embedded module is configured to set ECG signal data affected by baseline drift as a one-dimensional time signal x = [x1…x2]. N ], where N is the length of the one-dimensional time signal, and the one-dimensional time signal x = [x1…x N Transform into a trajectory matrix: The trajectory matrix X is a 2×(N-1) dimensional matrix, and the trajectory matrix has two rows to indicate that the window length L is 2;

[0066] The singular value decomposition module is configured to process matrix XX. T Perform singular value decomposition: XXT =VΛV T To obtain the eigenvalues ​​and corresponding eigenvectors, where singular values ​​λ1 and λ2 are stored on the main diagonal of matrix Λ, and λ1 > λ2, and the corresponding eigenvectors v1 and v2 are stored in the first and second columns of matrix V, the trajectory matrix X is transformed into the sum of two primary matrices of rank 1:

[0067] The grouping module is configured to set the matrix X1 corresponding to the large singular value λ1 as the baseline drift matrix and the matrix X2 corresponding to the small singular value λ2 as the electrocardiogram signal matrix. The matrix sizes of X1 and X2 are the same as those of the trajectory matrix X, which are both 2×(N-1).

[0068] The reconstruction module is configured to reconstruct a one-dimensional time signal of length N from the baseline drift matrix X1 and the electrocardiogram signal matrix X2 obtained in sub-step S230 using a diagonal averaging method. Specifically, the diagonal averaging method is used when the given matrix G is a 2×(N-1) dimensional matrix with matrix elements G. i,j Set L * =min(2,N-1) and K * =max(2,N-1), the reconstructed one-dimensional time series g = [g1,g2,L,g N Obtained through the diagonal mean formula:

[0069]

[0070] Furthermore, in the above system, the calculation module is also configured to acquire the baseline drift noise n obtained from the extraction module at the i-th layer. i and electrocardiogram (ECG) i ; and using formula IP i =(n i ) T (ECG i Calculate the inner product of the i-th layer.

[0071] It should be recognized that embodiments of the present invention can be implemented or carried out by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable storage medium. The method can be implemented using standard programming techniques—including a non-transitory computer-readable storage medium configured with a computer program, wherein such a storage medium causes the computer to operate in a specific and predefined manner—according to the methods and drawings described in the specific embodiments. Each program can be implemented in a high-level procedural or object-oriented programming language to communicate with the computer system. However, if desired, the program can be implemented in assembly or machine language. In any case, the language can be a compiled or interpreted language. Furthermore, for this purpose, the program can run on a programmed application-specific integrated circuit (ASIC).

[0072] Furthermore, the method can be implemented in any suitable type of computing platform, including but not limited to personal computers, minicomputers, mainframes, workstations, networked or distributed computing environments, standalone or integrated computer platforms, or in communication with charged particle tools or other imaging devices, etc. Aspects of the invention can be implemented as machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into a computing platform, such as a hard disk, optical read and / or write storage medium, RAM, ROM, etc., such that it is readable by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the processes described herein. Furthermore, the machine-readable code, or portions thereof, can be transmitted via wired or wireless networks. The applications described herein include these and other different types of non-transitory computer-readable storage media when such media comprises instructions or programs that implement the steps described above in conjunction with a microprocessor or other data processor. When programmed according to the methods and techniques described in the invention, the invention also includes the computer itself.

[0073] While the disclosed technology allows for various modifications and alternative constructions, certain embodiments have been shown in the accompanying drawings and described in detail above. However, it should be understood that the invention is not intended to be limited to the one or more specific forms disclosed; rather, it is intended to cover all modifications, alternative constructions, and equivalents that fall within the scope and definition of the invention as defined in the appended claims.

Claims

1. A method for suppressing baseline drift in electrocardiogram signals based on cascaded singular spectrum analysis, characterized in that, Includes the following steps: S100, acquire ECG signal data affected by baseline drift, and set the layer number. ; S200 uses Singular Spectrum Analysis (SSA) with a window length of 2 to extract baseline drift noise from ECG signal data affected by baseline drift. With ECG signal components Its central electrical signal Through formula get; S300, calculate baseline drift noise With electrocardiogram signal inner product ; S400: Determine if the inner product has reached its minimum value. If the inner product has reached its minimum value, output the ECG signal of the current layer. As a final result of suppressing baseline drift, otherwise, the baseline drift noise... As ECG signal data affected by baseline drift and layer number The value is set to increase by 1, and steps S200 and S300 are repeated until the inner product reaches a minimum value. The ECG signal of the current layer is then output. .

2. The method for suppressing baseline drift in electrocardiogram signals according to claim 1, characterized in that, Step S200 includes: S210, the ECG signal data affected by baseline drift is set as a one-dimensional time signal. ,in The length of the one-dimensional time signal is given by the length of the one-dimensional time signal. Transform into trajectory matrix : , The trajectory matrix for dimensional matrix, and the trajectory matrix It has two rows to indicate that the window length L is 2; S220, for the matrix Perform singular value decomposition: To obtain the eigenvalues ​​and their corresponding eigenvectors, including the singular values. and Stored in matrix On the main diagonal, and The corresponding feature vector and Stored in The first and second columns of the matrix will contain the trajectory matrix. This can be transformed into the addition of two primary matrices of rank 1: ; S230, large singularity The corresponding matrix Set as the baseline drift matrix, and use small singular values The corresponding matrix Set as an electrocardiogram signal matrix, where the matrix and Matrix size and trajectory matrix Consistent, all are ; S240, the baseline drift matrix obtained in sub-step S230 is... and ECG signal matrix Using the diagonal averaging method, they are reconstructed into lengths of... The one-dimensional time signal, wherein the diagonal averaging method specifically refers to: for a given matrix for A 3D matrix with matrix elements of 12000. ,set up and Reconstructed one-dimensional time series Obtained using the diagonal mean formula: 。 3. The method for suppressing baseline drift in electrocardiogram signals according to claim 1, characterized in that, Step S300 includes: S310, obtain the first step in step S200. Baseline drift noise obtained from the layer and electrocardiogram signal ; S320, using the formula Calculate the first The inner product of the layer.

4. A system for suppressing baseline drift in electrocardiogram signals based on cascaded singular spectrum analysis, characterized in that, include: The acquisition module is configured to acquire ECG signal data affected by baseline drift, and the number of layers is set. ; The extraction module is configured to extract baseline drift noise from ECG signal data affected by baseline drift using Singular Spectrum Analysis (SSA) with a window length of 2. With ECG signal components Its central electrical signal Through formula get; The calculation module is configured to calculate baseline drift noise. With electrocardiogram signal inner product ; The iterative output module is configured to determine whether the inner product has reached a minimum value. If the inner product has reached a minimum value, the ECG signal of the current layer is output. As a final result of suppressing baseline drift, otherwise, the baseline drift noise... As ECG signal data affected by baseline drift and layer number The value is set to increase by 1, and the extraction module and the calculation module are called to perform the corresponding extraction and calculation until the inner product reaches a minimum value, at which point the ECG signal of the current layer is output. .

5. The ECG signal baseline drift suppression system according to claim 4, wherein the extraction module comprises the following sub-modules: The embedded module is configured to set the ECG signal data affected by baseline drift as a one-dimensional time signal. ,in The length of the one-dimensional time signal is given by the length of the one-dimensional time signal. Transform into trajectory matrix : The trajectory matrix for dimensional matrix, and the trajectory matrix It has two rows to indicate that the window length L is 2; The singular value decomposition module is configured to process matrices. Perform singular value decomposition: To obtain the eigenvalues ​​and their corresponding eigenvectors, including the singular values. and Stored in matrix On the main diagonal, and The corresponding feature vector and Stored in The first and second columns of the matrix will contain the trajectory matrix. This can be transformed into the addition of two primary matrices of rank 1: ; The grouping module is configured to group large singular values. The corresponding matrix Set as the baseline drift matrix, and use small singular values The corresponding matrix Set as an electrocardiogram signal matrix, where the matrix and Matrix size and trajectory matrix Consistent, all are ; The reconstruction module is configured to convert the baseline drift matrix obtained from the grouping module. and ECG signal matrix Using the diagonal averaging method, they are reconstructed into lengths of... The one-dimensional time signal, wherein the diagonal averaging method specifically refers to: for a given matrix for A 3D matrix with matrix elements of 12000. ,set up and Reconstructed one-dimensional time series Obtained using the diagonal mean formula: 。 6. A non-transitory computer-readable storage medium storing computer instructions thereon, characterized in that... When the computer instructions are executed by the processor, they implement the steps of the method as described in any one of claims 1 to 3.

Images