Joint electrocardiosignal noise reduction method based on WOA-VMD

Abstract

The invention discloses a joint electrocardiosignal noise reduction method based on WOA-VMD. The joint electrocardiosignal noise reduction method comprises the following steps that (1) electrocardiosignals are collected and preprocessed; (2) optimizing the number K of modes and a penalty parameter alpha in the VMD algorithm by using the optimized WOA algorithm; (3) de-noising the electrocardiosignal by adopting a VMD algorithm optimized by IWOA and an improved wavelet threshold; (4) selecting a signal-to-noise ratio, a root mean square error and an autocorrelation coefficient as indexes for measuring a denoising effect; (5) outputting the denoised high-quality electrocardiosignal; the diagnosis value of the electrocardiosignals is improved, and more accurate diagnosis and treatment schemes are provided for heart disease patients.

Description

Technical Field

[0001] The present invention relates to the technical field of electrocardiogram signal processing, and in particular to a combined electrocardiogram signal denoising method based on WOA-VMD. Background Art

[0002] Traditional digital filtering methods process ECG signals in the frequency domain, but due to the frequency band aliasing between ECG signals and noise signals, many useful signals will be filtered out. In the wavelet threshold method, its local time-frequency analysis capability is strong, but the difficulty lies in the threshold selection. If the threshold is set too low, noise will be generated, while if the threshold is set too high, the ECG signal will be damaged. The selection of wavelet basis will also affect the denoising effect. The EMD algorithm can recursively decompose the noisy signal into a series of intrinsic mode function (IMF) components, but there are usually mode aliasing and false components between the noise and noise-free IMF components, which makes the denoising effect limited. In 2014, Dragomiretskiy et al. proposed variational mode decomposition (VMD). Based on completely non-recursive decomposition, the signal can be decomposed into a series of IMF components and the center frequency of each mode component can be determined. Therefore, the mode aliasing phenomenon in the EMD algorithm can be effectively solved. However, each mode component decomposed by VMD contains both signal and noise. If any component is removed arbitrarily, the accuracy of the reconstructed ECG signal will be affected. In recent years, many researchers have used VMD to denoise low-frequency noise baseline interference in ECG signals, while leaving high-frequency noise electromyographic interference alone. In addition, the selection of the number of modes K and the penalty parameter α in the VMD decomposition process directly affects the decomposition effect of the signal. In summary, while removing ECG signal noise, the existing technology has problems such as information loss, insufficient adaptability, complex calculations, difficulty in parameter selection, and incomplete noise removal. Summary of the invention

[0003] Purpose of the invention: The purpose of the present invention is to provide a joint ECG signal denoising method based on WOA-VMD, which solves the problems existing in the background technology by calculating the energy entropy of the IMF component, the cosine similarity with the noisy signal and other indicators, and dynamically adjusting the WOA algorithm and denoising operation according to these indicators.

[0004] Technical solution: The WOA-VMD-based combined ECG signal denoising method of the present invention comprises the following steps:

[0005] (1) Collecting ECG signals and preprocessing them;

[0006] (2) The optimized WOA algorithm is used to optimize the number of modes K and the penalty parameter α in the VMD algorithm;

[0007] (3) Using IWOA optimized VMD algorithm and improved wavelet threshold to denoise ECG signals;

[0008] (4) Select signal-to-noise ratio, root mean square error, and autocorrelation coefficient as indicators to measure the denoising effect;

[0009] (5) Output high-quality ECG signal after denoising.

[0010] Furthermore, in step (1), the ECG signal is collected as follows: using an ECG monitoring device, the device includes multiple electrodes, the electrodes are placed on the patient's chest and limbs, the ECG monitoring device is connected to a data collector via wires or wirelessly, the data collector has a built-in high-precision analog-to-digital converter to collect ECG signals, the collected signals include P waves, QRS complexes and T waves, and the timestamp and electrode position information are recorded at the same time.

[0011] Furthermore, in step (1), the preprocessing is specifically as follows: the collected ECG signal is passed through a bandpass filter adaptive filtering algorithm to remove the baseline drift; wherein the passband of the filter is set between 0.5 Hz and 150 Hz to remove the baseline drift below 0.5 Hz and the high-frequency noise above 150 Hz.

[0012] Furthermore, in step (2), an improved WOA is obtained by changing the search path and introducing a chaotic mechanism; wherein the search path is changed by introducing a Fermat spiral curve.

[0013] Furthermore, in step (3), the VMD algorithm is optimized using the perturbation mechanism and the IWOA of the complex path, the parameter combination K and α is determined, and the energy entropy is selected as the fitness function.

[0014] Furthermore, in step (3), the key to wavelet threshold denoising lies in the selection of threshold and threshold function, where a fixed threshold is selected to determine the threshold of IMF:

[0015]

[0016] Among them, x i is the threshold of the ith IMF function, and N is the signal length.

[0017] Select the soft threshold function as the threshold function:

[0018]

[0019] Among them, w j,k is the wavelet coefficient, and λ is the threshold.

[0020] Furthermore, the formula of step (5) is as follows:

[0021]

[0022] Among them, s i is the original signal, y i To reconstruct the signal.

[0023] The present invention provides a combined ECG signal denoising system based on WOA-VMD, comprising:

[0024] Preprocessing module: used to collect ECG signals and perform preprocessing;

[0025] WOA module: used to optimize the number of modes K and penalty parameter α in the VMD algorithm using the optimized WOA algorithm;

[0026] IWOA module: used to denoise ECG signals using the IWOA-optimized VMD algorithm and improved wavelet threshold;

[0027] Measurement module: used to select signal-to-noise ratio, root mean square error, and autocorrelation coefficient as indicators for measuring denoising effect;

[0028] Output module: used to output high-quality ECG signals after denoising.

[0029] An electronic device described in the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is loaded into the processor, a combined ECG signal denoising method based on WOA-VMD is implemented according to any one of the items.

[0030] A storage medium described in the present invention stores a computer program, and when the computer program is executed by a processor, it implements any one of the WOA-VMD-based combined ECG signal denoising methods described.

[0031] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: It combines optimized VMD decomposition and wavelet threshold. The Fermat curve is used to replace the classic logarithmic spiral curve to improve the whale optimization algorithm (WOA). The number of modes K and the penalty parameter α in VMD are optimized by the improved WOA to obtain a series of intrinsic mode function (IMF) components. The main component and the noise component are determined by correlation coefficient analysis, and then the noise component is denoised by the wavelet threshold method. Finally, the dominant mode and the denoised component are reconstructed to obtain the denoised electrocardiogram (ECG). This method improves the accuracy and efficiency of ECG signal processing through advanced signal processing technology and algorithm optimization, ensuring the provision of high-quality ECG signals under different monitoring conditions. This will enhance the diagnostic value of ECG signals and provide more accurate diagnosis and treatment plans for patients with heart disease. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1It is a flow chart of the present invention. DETAILED DESCRIPTION

[0033] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0034] like Figure 1 As shown, an embodiment of the present invention provides a joint ECG signal denoising method based on WOA-VMD, comprising the following steps:

[0035] (1) Collecting ECG signals and preprocessing them; Collecting ECG signals is as follows: using an ECG monitoring device, the device includes multiple electrodes, the electrodes are placed on the patient's chest and limbs, the ECG monitoring device is connected to a data collector via a wire or wirelessly, the data collector has a built-in high-precision analog-to-digital converter to collect ECG signals, and the collected signals include P waves, QRS complexes and T waves, while recording timestamps and electrode position information. Preprocessing is as follows: The collected ECG signals are passed through a bandpass filter adaptive filtering algorithm to remove baseline drift; wherein the passband of the filter is set between 0.5Hz and 150Hz to remove baseline drift below 0.5Hz and high-frequency noise above 150Hz.

[0036] (2) The optimized WOA algorithm is used to optimize the mode number K and penalty parameter α in the VMD algorithm; specifically, the improved WOA is obtained by changing the search path and introducing a chaotic mechanism; wherein the search path is changed by introducing a Fermat spiral curve. The following steps are included:

[0037] (21) Introducing chaotic sequences. In order to enhance the local search capability of WOA, a set of more favorable search positions X can be generated by chaotic sequences after each iteration. * (t), but generally will not directly enter the next iteration, but use the randomness and ergodicity of the chaotic mechanism to thoroughly search for X * (t) The surrounding positions can be used to increase the probability of finding a better solution, which can generate new optimal search agents in subsequent iterations, promote the algorithm to continuously evolve towards a better solution, and effectively avoid the algorithm from falling into a local optimum.

[0038] Steps to introduce chaos mechanism:

[0039] a Select Logistic mapping to generate a chaotic sequence. Its expression is:

[0040] x n+1 = r·x n (1-x n ) (1)

[0041] where r is a parameter that controls the characteristics of the chaotic mechanism and is usually between 3.75 and 4; n and xn+1 Represent the values ​​of the nth and n+1th iterations respectively.

[0042] b updates the search position through the chaos mechanism and then generates the search position X * (t), a chaotic mechanism is introduced to generate a new candidate search position X new The new position calculation formula is:

[0043] x new =X * (t)+α·(X * (t)-X best )+C·chaos(t) (2)

[0044] Among them, X best is the current most likely position, α is the perception coefficient, C is a randomly generated weight factor, which is taken as 0.5 in this paper (usually between 0 and 1), and chaos(t) is the chaos value generated at the current time. (3) Optimizing the search agent, through such an update method, the search agent will effectively search within the range of chaos generation, and its goal is to find a new optimal solution in the neighborhood.

[0045] (22) Change the search path. The Fermat spiral curve is introduced. Its formation process is to set an origin as the center and a point in the polar coordinate system along a uniformly rotating and gradually expanding trajectory. The mathematical model of its search path is shown in (3). For the original algorithm, the logarithmic spiral curve will have sharp corners or mutations, which will affect the performance of the algorithm. The Fermat spiral curve is selected. Its smoothness allows continuity to be maintained during the optimization process, thereby improving the stability of the algorithm and reducing oscillations or instability. The formula is as follows:

[0046]

[0047] Where r is the radial distance, θ is the polar angle, and a is a constant that determines the size of the spiral and how fast it expands.

[0048] By changing the search path and introducing the chaotic mechanism, the improved WOA is obtained. In the improved WOA algorithm, the initialization stage uses the Logistic chaotic map and the Fermat spiral curve to coordinate, so that the search agent is widely distributed and has regularity and diversity, which is conducive to locating the approximate location of the better area; in the early iteration, the Fermat spiral curve guides the orderly exploration, and the chaotic mechanism fine-tunes the path to help break out of the fixed mode and explore potential areas; in the mid-term iteration, the algorithm balances the exploration of new areas and the deep search of the better areas, the spiral curve expands the range to prevent local optimality, and the chaotic mechanism leads to a jump search; in the late stage of chaos, the algorithm locks the target area, the chaotic mechanism shrinks the range to limit the moving area, and changes the direction according to the sequence, and searches carefully near the target to find the precise optimal solution. The step size of the search agent is redefined: the maximum number of iterations is set to Nmax , define a stage judgment function stage(n), where n is the number of iterations. mid (N mid is the number of mid-term iterations), stage(n) = 1; when N mid ≤n≤N late (N late is the number of later iterations), stage(n)=2; when n≥N late When , stage(n) = 3. The comprehensive formula of the step length s of the search agent is:

[0049]

[0050] where k 10 is the scale factor for initialization and early iteration stages, b0 is the corresponding scaling factor; k 1m b is the proportional coefficient of the mid-term selection stage; m is the corresponding scaling factor; k 1l is the proportional coefficient of the late stage of chaos, b l is the corresponding scaling factor; x is the value related to the chaotic factor generation (the value generated by the Logistic chaotic map), x min Used to adjust the range of chaos factors, a is the parameter of the Fermat spiral curve, θ is the polar angle, and Δθ is the small change in the polar angle. This comprehensive formula can dynamically adjust the step size of the search agent according to the iteration stage to adapt to the search needs at different stages.

[0051] From the above formula, we can see that the step size is getting smaller and smaller, and it is the minimum value at the end of the iteration.

[0052] In order to verify the optimality of the search path in WOA, three other classic curves (logarithmic curve, rose curve, and Fermat curve) were selected for comparison, and different test functions were used for experiments. The test functions of test 1, 2, and 3 are shown in formulas (5), (6), and (7):

[0053]

[0054] Under the same number of iterations, the convergence degree and convergence speed of the WOA algorithm using the Fermat spiral curve as the search path are better than the logarithmic spiral curve and the other two curves in the original algorithm. In particular, the convergence degree of the WOA algorithm using the Fermat spiral curve is close to 0, which is much better than the logarithmic spiral curve.

[0055] WOA mainly includes three stages: capturing prey, attacking with bubble net, and searching for prey. The specific steps of the improved algorithm are as follows.

[0056] (S1) Capture prey. In the initial stage, the WOA algorithm assumes that the current best candidate is the target prey or the best target.

[0057] After defining the best agent, other agents will update their positions to it. The mathematical model can be described as:

[0058] D=|C·X * (t)-X(t)| (8)

[0059] X(t+1)=X * (t)-A·D (9)

[0060] A=2a·ra (10)

[0061] C=2r (11)

[0062] Where D is the distance vector between the best search agent and the normal agent, t is the number of iterations, A and C are coefficient vectors, and X * (t) is the known best vector, X(t) is the position of other search agents, and r is a random vector in [0,1].

[0063] (S2) Bubble net attack (development stage). There are two mechanisms for humpback whale predation: the contraction and encirclement mechanism and the spiral update position. In the spiral update position, a Fermat spiral equation is created between them. In polar coordinates, the Fermat spiral curve equation is (where a is a constant and θ is the polar angle). If the position is expressed in polar coordinates, let the polar diameter corresponding to the position at the current time t be r(t) and the polar angle be θ(t). After one iteration (to time t+1), the polar angle increases by Δθ (Δθ can be set according to the specific iteration procedure and is related to the search progress factor). Then the polar diameter at time t+1 is

[0064] At this time, X(t+1) is expressed in polar coordinates as (r(t+1),θ(t)+Δθ), that is,

[0065] It is worth mentioning that humpback whales swim around their prey, gradually contracting into circles and spirals. To facilitate the establishment of the model, it is assumed that there is a 50% probability of choosing between the contraction and surrounding mechanism and the spiral update position to update the position of the whale during the optimization process. Converting the Fermat spiral curve from polar coordinates to Cartesian coordinates, we have x = rcosθ, y = sinθ, for the Fermat spiral curve spiral Available

[0066] The mathematical model is as follows:

[0067]

[0068] Where ρ is a random number in [0,1], with a value of 0.5; Δθ is Random value in the range.

[0069] (S3) Searching for prey. In this stage, in order to make the search range wider, the search agents are randomly dispersed, and the positions of the search agents are replaced by the best candidate agents randomly selected, and the Logistic chaotic map is added. The mathematical description model is:

[0070] D=|C·X rand -X(t)| (13)

[0071] X(t+1)=X rand -A·D+γ·y n (14)

[0072] Among them, X rand is a random position vector (random whale) selected from the current population. In each iteration, a Logistic chaotic map is executed to thoroughly search the position near the agent. In the next iteration, a new best search agent will be generated to update the position. When |A|≥1, a random agent is selected as the next reference. When |A|<1, the current best search agent is selected. {y n} is a Logistic chaotic sequence, let the initial value y0∈(0,1), through y n+1 = r·y n (1-y n ) Iterate to get the sequence; γ is a scaling factor used to control the influence of the Logistic chaotic map on the position update.

[0073] The present invention uses energy entropy as a fitness function to measure the energy distribution characteristics of the time series after decomposition. Energy entropy can reflect the concentration of signal energy in different frequency components or subsequences. The smaller the entropy value, the more concentrated and orderly the energy distribution is, indicating that each subsequence obtained after VMD decomposition is more regular at the energy level, and thus contains more effective information. During the operation of the algorithm, by calculating the energy entropy of the VMD decomposition results under different parameter combinations and selecting the parameter combination corresponding to the minimum energy entropy, the optimal parameter combination can be obtained.

[0074] (3) The IWOA-optimized VMD algorithm and improved wavelet threshold are used to denoise the ECG signal; the details are as follows:

[0075] Before VMD decomposition, it is necessary to set the appropriate number of modes K and penalty parameter α. If the K value is too large, it will be over-decomposed, resulting in one component being included in multiple components, resulting in spectral aliasing; if the K value is too small, it will be under-decomposed, resulting in multiple components being included in one component. The same is true for α. If α is too large, the bandwidth limit will become narrower, resulting in loss of frequency band information; otherwise, redundant components will be retained. Therefore, this paper uses IWOA with a perturbation mechanism and complex path to optimize the VMD algorithm, determine the parameter combination K and α, and select energy entropy as the fitness function.

[0076] The key to wavelet threshold denoising lies in the selection of threshold and threshold function. A fixed threshold is selected to determine the threshold of IMF:

[0077]

[0078] Among them, x i is the threshold of the ith IMF function, and N is the signal length.

[0079] Selection of threshold function. After determining the threshold, a threshold function wavelet coefficient is needed for processing. Common threshold functions include hard threshold function and soft threshold function: Since the wavelet coefficients obtained by the hard threshold function are discontinuous, additional oscillations will be generated during reconstruction, making the result after signal processing uneven. The soft threshold function has better overall continuity and will not produce additional oscillations, so the result after signal processing is relatively smooth. Select the soft threshold function as the threshold function:

[0080]

[0081] Among them, w j,k is the wavelet coefficient, and λ is the threshold.

[0082] The process of denoising the ECG signal is as follows:

[0083] (A1) Input a noisy ECG signal, initialize the WOA algorithm parameters (population size, number of iterations, spatial dimension) and the value range of the decomposition parameters K and α in the VMD algorithm, and initialize an empty denoised signal container.

[0084] (A2) VMD decomposition of the noisy ECG signal is performed according to the search agent parameter combination generated by the current WOA algorithm to obtain a set of IMF components.

[0085] (A3) Calculate the energy entropy of each IMF component, use the energy entropy as an evaluation index, and select the IMF component with the smallest energy entropy as the current optimal IMF component.

[0086] (A4) 0.5 is used as the threshold for distinguishing modal components (it provides a simple and clear judgment standard, which can balance the search range and accuracy in the WOA-based optimization process, and has been experimentally verified to effectively balance denoising and retaining signal features in ECG denoising processing). When p<0.5: calculate the cosine similarity between the current optimal IMF component and the noisy ECG signal. According to the similarity index, adjust the parameters of the shrinking and surrounding mechanism in the WOA algorithm to make the search agent more inclined to search in areas with high similarity. Determine the value of the updated position A. When |A|≥1: make a large-scale random adjustment to the search agent position, and at the same time, combine the energy information of the current optimal IMF component to weight the adjustment range to ensure that while enhancing the global search capability, it does not deviate too much from the current better area. The new search agent position = original position + random factor × A × current optimal IMF component energy-related weight × (search space upper limit - search space lower limit). When |A|<1: fine-tune the search proxy position, mainly perform local search optimization near the current optimal IMF component, and the new search proxy position = original position + small random factor × A × current optimal IMF component energy-related weight × (current optimal IMF component eigenvalue range). After updating the search proxy position, perform VMD decomposition on the noisy ECG signal again according to the new parameter combination, and repeat the operations in the p<0.5 part of steps (3) to (4) until the internal loop termination condition is met (such as reaching a certain number of internal iterations or the energy entropy of the current optimal IMF component is no longer significantly reduced). When p≥0.5: perform wavelet transform on the current optimal IMF component to obtain wavelet coefficients. Determine the wavelet threshold according to a threshold selection method based on statistical characteristics. Use the determined threshold to process the wavelet coefficients to obtain denoised wavelet coefficients, and then perform inverse wavelet transform to obtain denoised IMF components.

[0087] (A5) The denoised IMF component is stored in a denoised signal container.

[0088] (A6) Repeat steps (A2)-(A5) until the set number of iterations is reached.

[0089] (A7) All denoised IMF components in the denoised signal container are reconstructed to obtain the final denoised ECG signal.

[0090] The modal number K and penalty parameter α are determined by optimizing the VMD algorithm with WOA with perturbation mechanism and complex path. First, energy entropy is selected as the fitness function because it can reflect the energy distribution characteristics of the signal. The smaller the entropy, the more orderly the energy is concentrated, the more regular the subsequence is, and the more effective information it contains. In the operation, the WOA algorithm parameters and the value range of VMD decomposition parameters K and α are initialized within the set range. Then, the noisy ECG signal is decomposed by VMD according to the WOA search agent parameter combination to obtain the IMF component, and the energy entropy of each component is calculated and the parameter corresponding to the smallest one is selected as the current optimal. Subsequently, the search agent position and parameter combination are continuously iterated and updated until the termination conditions such as reaching the set number of iterations or the energy entropy of the optimal IMF component is no longer significantly reduced are met. The modal number K and penalty parameter α finally determined are what are required, so as to ensure the VMD decomposition effect and improve the quality of ECG signal denoising.

[0091] (4) The signal-to-noise ratio, root mean square error, and autocorrelation coefficient are selected as indicators to measure the denoising effect; the formula is as follows:

[0092]

[0093] Among them, s i is the original signal, y i is the reconstructed signal. The larger the SNR and the smaller the RMSE, the better the denoising effect of this method; the larger the AC, the smaller the deviation between the reconstructed signal and the original signal.

[0094] Table 1 Comparison of denoising results of three methods

[0095] method SNR / dB RMSE Wavelet Thresholding 14.5755 0.0934 VMD 32.1596 0.0213 Methods 32.6635 0.0131

[0096] As can be seen in Table 1, the WOA-optimized VMD algorithm and the improved wavelet threshold joint denoising method proposed in this paper show higher SNR, AC and lower RMSE after denoising compared with the other two methods. Therefore, the denoising effect of the algorithm in this paper is the best, and it has good adaptability to different types of ECG signals.

[0097] (5) The signal output module outputs high-quality ECG signals after denoising for subsequent analysis and diagnosis.

[0098] The signal is transmitted to the diagnostic system through the signal output module. During this process, the signal may be further processed, such as resampling or quantization, to adapt to different analysis tools. The output signal will be used for further analysis, such as automatic heart rhythm recognition, heart rate variability analysis, or heart disease diagnosis. These analysis results will be used for clinical decision support to help doctors make more accurate diagnoses.

Claims

1. A joint ECG signal denoising method based on WOA-VMD, characterized in that: The following steps are involved: (1) Collecting ECG signals and preprocessing them; (2) The optimized WOA algorithm is used to optimize the number of modes K and the penalty parameter α in the VMD algorithm; (3) Using IWOA optimized VMD algorithm and improved wavelet threshold to denoise ECG signals; (4) Select signal-to-noise ratio, root mean square error, and autocorrelation coefficient as indicators to measure the denoising effect; (5) Output high-quality ECG signal after denoising.

2. The WOA-VMD-based combined ECG signal denoising method according to claim 1, characterized in that: In step (1), the ECG signal is collected as follows: using an ECG monitoring device, the device includes multiple electrodes, the electrodes are placed on the patient's chest and limbs, the ECG monitoring device is connected to a data collector via wires or wirelessly, the data collector has a built-in high-precision analog-to-digital converter to collect ECG signals, the collected signals include P waves, QRS complexes and T waves, and the timestamp and electrode position information are recorded at the same time.

3. The OCT image classification method based on self-supervised learning according to claim 1, characterized in that: In step (1), the preprocessing is as follows: the collected ECG signal is passed through a bandpass filter adaptive filtering algorithm to remove the baseline drift; wherein the passband of the filter is set between 0.5 Hz and 150 Hz to remove the baseline drift below 0.5 Hz and the high-frequency noise above 150 Hz.

4. The WOA-VMD-based combined ECG signal denoising method according to claim 1, characterized in that: It is characterized in that In step (2), an improved WOA is obtained by changing the search path and introducing a chaotic mechanism; wherein the search path is changed by introducing a Fermat spiral curve.

5. The WOA-VMD-based combined ECG signal denoising method according to claim 1, characterized in that: It is characterized in that In step (3), the VMD algorithm is optimized using the perturbation mechanism and the IWOA of complex paths, the parameter combination K and α is determined, and the energy entropy is selected as the fitness function.

6. The WOA-VMD-based combined ECG signal denoising method according to claim 1, characterized in that: It is characterized in that In step (3), the key to wavelet threshold denoising lies in the selection of threshold and threshold function, where a fixed threshold is selected to determine the threshold of IMF: Among them, x i is the threshold of the ith IMF function, and N is the signal length. Select the soft threshold function as the threshold function: Among them, w j,k is the wavelet coefficient, and λ is the threshold.

7. The WOA-VMD-based combined ECG signal denoising method according to claim 1, characterized in that: The formula for step (5) is as follows: Among them, s i is the original signal, y i To reconstruct the signal.

8. A combined ECG signal denoising system based on WOA-VMD, characterized in that: include: Preprocessing module: used to collect ECG signals and perform preprocessing; WOA module: used to optimize the number of modes K and penalty parameter α in the VMD algorithm using the optimized WOA algorithm; IWOA module: used to denoise ECG signals using the IWOA-optimized VMD algorithm and improved wavelet threshold; Measurement module: used to select signal-to-noise ratio, root mean square error, and autocorrelation coefficient as indicators for measuring denoising effect; Output module: used to output high-quality ECG signals after denoising.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the computer program is loaded into a processor, the WOA-VMD-based combined ECG signal denoising method according to any one of claims 1 to 7 is implemented.

10. A storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, a WOA-VMD-based combined ECG signal denoising method according to any one of claims 1 to 7 is implemented.

Images