Surface diaphragmatic electromyography denoising method based on blind source separation and stripping strategy

Abstract

The application discloses a surface diaphragm electromyography denoising method based on blind source separation and stripping strategy, comprising the following steps: 1, collecting multi-channel electrophysiological signals containing diaphragm electromyography on the surface of the thoracic skin; 2, performing a blind source separation algorithm to obtain a plurality of source signals; 3, clustering and selecting the source signals to obtain a reference signal; 4, based on the reference signal, using a time-domain constraint FastICA algorithm to obtain more accurate respiratory and electrocardio signal components; 5, stripping all components from the initial signal, then performing steps 2-5 on the residual signal until the termination condition is reached; 6, accumulating the respiratory signal components in the result set to obtain a pure and complete surface diaphragm electromyography signal. The application denoises the surface diaphragm electromyography collected non-invasively but containing high-intensity noise, obtains a pure diaphragm electromyography signal, and exhibits the neural instructions related to the respiration of a subject, so as to provide accurate control information for air supply of a clinical breathing machine.

Description

Technical Field

[0001] This invention belongs to the field of electromyographic signal processing technology, specifically involving a surface diaphragm electromyographic noise reduction method based on blind source separation and stripping strategy, which enables the identification of the subject's respiratory rhythm and assessment of respiratory status. Background Technology

[0002] In clinical practice, a key aspect of mechanical ventilation is the accurate identification and quantification of a patient's spontaneous breathing capacity. Traditional methods use flow and pressure measurements to achieve this. However, the pressure generated by inspiratory muscle contraction must overcome airway resistance and alveolar positive end-expiratory pressure (PEEPi) before triggering a ventilator-induced flow or pressure change in the airway. Therefore, there is a delay between inspiratory muscle contraction and the generation of a flow or pressure change in the airway, resulting in a lag between the timing of mechanical ventilation and the patient's actual inhalation. In contrast, the electromyographic signal (EMGdi) generated by the diaphragm during respiration appears earlier than flow and pressure signals and is less affected by factors such as lung and airway diseases. As one of the main respiratory muscle groups, the diaphragm is responsible for 60%-80% of respiratory movements. Therefore, measuring and processing the diaphragm's EMG signal can be used to trigger mechanical ventilation and can also serve as one of the criteria for assessing the strength of the diaphragm's EMG signal.

[0003] In recent years, a type of ventilator with neurally adjusted ventilation assist (NAVA) has achieved this goal by utilizing esophageal diaphragmatic electromyography (EMGdi) signals. However, collecting EMGdi signals through the esophagus is difficult to implement, requiring specific acquisition conditions and a suitable environment, which is inconvenient for patients and also somewhat invasive, potentially causing discomfort (such as bleeding or vomiting). Another method involves obtaining EMGdi signals through surface electrodes. However, the diaphragmatic EMG signal must pass through a series of body tissues to reach the surface, resulting in very weak amplitude signals. Furthermore, surface electrodes are often affected by various internal and external interferences and noises, such as ECG interference. Extreme motion artifacts, power line interference, and other noises cause weak, useful information to be overwhelmed by a large amount of noise. Traditional single-channel methods typically rely on signal morphology for noise reduction. When the amplitude of the useful components in the signal is too small and completely overwhelmed by noise, single-channel methods such as wavelet transform and thresholding become unsuitable. If traditional blind source separation methods are used for source signal separation and noise reduction, the algorithm often converges locally to the large amplitude ECG components, ignoring weak respiratory source signals. Current mainstream noise reduction techniques cannot achieve ideal noise reduction. Therefore, despite the many drawbacks of esophageal sampling, most NAVA ventilators on the market still primarily use esophageal EMG sampling. Summary of the Invention

[0004] To overcome the shortcomings of existing technologies, this invention proposes a surface diaphragm electromyography (EMG) denoising method based on a blind source separation and stripping strategy. The aim is to denoise weak surface diaphragm EMG signals from environments containing high-intensity noise, obtain pure multi-channel diaphragm physiological signals, and thus accurately extract the subject's respiratory rhythm and determine their respiratory status.

[0005] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0006] The surface diaphragm electromyography denoising method based on blind source separation and stripping strategy of the present invention is characterized by the following steps:

[0007] Step 1: Acquire surface diaphragm electromyography (EMG) signals and perform noise reduction using frequency domain analysis to obtain multi-channel surface diaphragm EMG signals X = [x1, x2, ..., x...]. i ,…,x n ] T , and serve as the initial signal, where x i This represents the surface diaphragm electromyography signal of the i-th channel after filtering out low-frequency and high-frequency noise, where T represents transpose;

[0008] Step 2: Execute the u-th blind source separation algorithm to obtain several source signals under the u-th separation:

[0009] Step 2.1: Initialization:

[0010] Set the total number of algorithm executions to u. max The current execution count is u, and u is initialized to 1; define and initialize the residual signal under the current u-th execution of the blind source separation algorithm.

[0011] Initialize the number of respiratory signal component strippings a = 0 and the number of electrocardiogram signal component strippings b = 0;

[0012] Define and initialize the set of respiratory signal sources γ and its corresponding unmixed vector set ε as empty sets;

[0013] Step 2.2: Use equation (1) to analyze the residual signal after the u-th separation. The signal is expanded to obtain the extended residual signal after the u-th execution of the blind source separation algorithm.

[0014]

[0015] In equation (1), K is the delay factor; This represents the residual signal of the i-th channel during the u-th separation. The delayed signal after a delay of K+1; t is the current time; T represents the transpose;

[0016] Step 2.3: Calculate the residual signal of the i-th channel in the u-th iteration. source signal Thus, the residual signal in the u-th separation is obtained. source signal in, This represents the i-th unmixed vector under the u-th execution of the blind source separation algorithm;

[0017] Step 3: Use a clustering algorithm to separate the source signal Y from the u-th cluster. u The signals were categorized and selected separately as respiratory-related signals and electrocardiogram-related signals as reference signals for the u-th separation.

[0018] Step 3.1, for Perform clustering to obtain F clusters under the u-th separation;

[0019] Step 3.2: Calculate the similarity between the average sequence of each cluster under the u-th separation and the acquired prior respiratory signal and prior ECG signal, and determine whether there is a cluster whose average sequence matches the prior respiratory signal template or the prior ECG signal template; if there is a cluster whose average sequence matches the prior respiratory signal template, assign a+1 to a and then proceed to step 3.3; if there is a cluster whose average sequence matches the prior ECG signal template, assign b+1 to b and then proceed to step 3.4; if neither exists, proceed to step six.

[0020] Step 3.3: Select the source signal with the highest signal-to-noise ratio from the clusters that match the prior respiratory signal template as the respiratory-related source signal for the a-th stripping under the u-th separation. Then, proceed to step four;

[0021] Step 3.4: Select the source signal with the highest signal-to-noise ratio from the cluster that matches the prior ECG signal template as the ECG-related source signal for the b-th stripping under the u-th separation. Then, proceed to step four;

[0022] Step 4: The ECG-related source signals or respiratory-related signals r, as the reference signal under the u-th separation u Thus utilizing the reference signal r u In the extended residual signal The FastICA algorithm with time-domain constraints is executed, resulting in the ECG signal component stripped in the b-th separation under the u-th separation. Or the respiratory signal component of the a-th stripping

[0023] Step 5: or From residual signal The residual signal after the (u+1)th separation is obtained by stripping. Assign u+1 to u, return to step 2.2 and execute sequentially until the maximum number of iterations u is reached. max until;

[0024] Step 6: After summing all respiratory signal components in set γ, a pure and complete diaphragmatic electromyographic signal F is obtained.

[0025] The surface diaphragm electromyography denoising method based on blind source separation and stripping strategy described in this invention is also characterized in that step 1 includes:

[0026] Step 1.1: Acquire multi-channel surface electromyographic signals of the diaphragm containing noise and interference using an array of electrode pads with rows θ and columns ρ. C = [c1, c2, ..., c i ,…,c n ] T , where c i The signal represents the surface diaphragm electromyography signal of the i-th channel containing noise and interference, where n represents the total number of channels, and n = θ × ρ;

[0027] Step 1.2: Process the multichannel surface diaphragm electromyography signal C = [c1, c2, ..., c] containing noise and interference. i ,…,c n ] T Performing a discrete Fourier transform yields the multi-channel surface diaphragm electromyography signal X = [x1, x2, ..., x...] after filtering out low-frequency and high-frequency noise. i ,…,x n ] T .

[0028] Step 5 includes:

[0029] Step 5.1: Solve using the least squares method. or The original signal x corresponding to the i-th channel i The transformation matrix estimated for the a-th time Or the transformation matrix estimated in the bth iteration. The respiratory signal component was estimated using the a-th time. The waveform corresponding to the i-th channel of the original signal X or ECG signal components The waveform corresponding to the i-th channel of the original signal X Where, x i Let X be the original signal of the i-th channel in the original signal X;

[0030] Step 5.2: Convert the waveform or From residual signal Subtract from the i-th channel to obtain the residual signal of the i-th channel after the (u+1)-th separation. Or the residual signal of the i-th channel after the (u+1)-th separation. Thus, the residual signal after the (u+1)th separation is obtained. and respiratory signal components Place it into the respiratory signal source set γ.

[0031] The present invention provides an electronic device, including a memory and a processor, wherein the memory is used to store a program that supports the processor in executing the surface diaphragm electromyography denoising method based on a blind source separation and stripping strategy, and the processor is configured to execute the program stored in the memory.

[0032] The present invention discloses a computer-readable storage medium on which a computer program is stored, wherein the computer program, when executed by a processor, performs the steps of the surface diaphragm electromyography denoising method based on a blind source separation and stripping strategy.

[0033] Compared with existing technologies, the beneficial effects of this invention are reflected in:

[0034] 1. Compared with traditional ventilators, this invention reduces human-machine asynchrony and directly reflects the subject's breathing intention. Compared with the method of measuring diaphragmatic electromyography through the esophagus, it uses a convenient and non-invasive electromyography acquisition method to obtain high-density sEMGdi signals with rich spatiotemporal information, simplifying the signal acquisition steps and reducing the pain of the subject during the acquisition process.

[0035] 2. This invention applies an optimized blind source separation method to high-density surface diaphragm electromyography analysis, removing interference and noise from the acquired signals while fully preserving the information in the sEMGdi signals to obtain relevant electrophysiological information of the subjects.

[0036] 3. This invention uses an automated method to automatically select respiratory-related source signals, and uses the time-domain constrained FastICA algorithm to detect and optimize the denoising and extraction results, which improves the reliability of the preprocessing process. It also separates the accurately extracted components, avoiding local convergence of the algorithm due to excessive noise, and improving the extraction effect of weak signals in environments containing high-intensity noise. Attached Figure Description

[0037] Figure 1 This is a flowchart of the present invention;

[0038] Figure 2 This is a schematic diagram of the electrode configuration used in this invention;

[0039] Figure 3 This is a schematic diagram showing the electrode positions of the present invention;

[0040] Figure 4 This is a schematic diagram of the original 32-channel filtered signal acquired by this invention.

[0041] Figure 5 This invention automatically selects an electrocardiogram-related component and an automatically selected respiratory-related component.

[0042] Figure 6 This is a schematic diagram of the pure diaphragm electromyography signal and the extracted diaphragm electrical activity curve (EAdi) extracted in this invention. Detailed Implementation

[0043] In this embodiment, a surface diaphragm electromyography denoising method based on a blind source separation and stripping strategy is described, such as... Figure 1 As shown, the procedure is as follows:

[0044] Step 1: Acquire surface diaphragm electromyography signals and perform noise reduction using frequency domain analysis;

[0045] Step 1.1: Acquire multi-channel electromyographic signals of the surface diaphragm containing noise and interference using an array of electrode pads with rows θ and columns ρ. C = [c1, c2, ..., c i ,…,c n ] T , where c i Let n represent the surface diaphragm electromyography signal containing noise and interference in the i-th channel, and n represent the total number of channels, and n = θ × ρ;

[0046] In this embodiment, the data acquisition device uses a 32-channel high-density surface electromyography (EMG) acquisition device, and the schematic diagram of the electrode pads used is shown below. Figure 2 As shown. The array of electrode pads is attached to the right anterior axillary line where it intersects with the 7th-9th ribs. Additionally, a common reference electrode is attached to the upper edge of the sternum, as shown. Figure 3 As shown. The system sampling frequency is 1000Hz, the electrodes are arranged in 4 rows × 8 columns, i.e., θ=4, ρ=8, n=32, and the acquisition time is 40 seconds.

[0047] Step 1.2: Analyze the electromyographic signals of the diaphragm muscle containing noise and interference, C = [c1, c2, ..., c]. i ,…,c n ] T Performing a discrete Fourier transform yields the surface diaphragm electromyography signal X = [x1, x2, ..., x] after filtering out low-frequency and high-frequency noise. i ,…,x n ] T , as the initial signal, where xi This represents the surface diaphragm electromyography signal of the i-th channel after filtering out low-frequency and high-frequency noise, where T represents transpose;

[0048] In this embodiment, since the energy of the sEMGdi signal is mainly distributed in the 10-250Hz frequency band, a sixth-order Butterworth bandpass filter of 10-250Hz is used to remove possible low-frequency and high-frequency noise, and a 50Hz notch filter is used to eliminate power frequency interference. The signal waveform after filtering is as follows. Figure 4 As shown.

[0049] Step 1.3: Set the total number of algorithm executions to u. max The current execution count is u, and u is initialized to 1; define and initialize the residual signal under the current u-th execution of the blind source separation algorithm.

[0050] Initialize the number of respiratory signal component strippings a = 0 and the number of electrocardiogram signal component strippings b = 0;

[0051] Define and initialize the set of respiratory signal sources γ and its corresponding unmixed vector set ε as an empty set;

[0052] In this embodiment, u max Set to 20;

[0053] Step 2: Execute the blind source separation algorithm to obtain several source signals under the u-th separation; in this embodiment, the FastICA algorithm in the blind source separation algorithm is used for independent component analysis;

[0054] Step 2.1: Use equation (1) to analyze the residual signal after the u-th separation. The signal is expanded to obtain the extended residual signal after the u-th execution of the blind source separation algorithm.

[0055]

[0056] In equation (1), K is the delay factor, and in this embodiment, K is set to 5; This represents the residual signal of the i-th channel in the u-th iteration. The delayed signal after a delay of K+1; t is the current time; T represents the transpose;

[0057] In this embodiment, since the more accurate ECG or respiratory source signal separated each time needs to be stripped from the residual signal, the residual signal is stored at the beginning of each round of algorithm execution so that it can be stripped from the ECG or respiratory source signal after it is finally separated.

[0058] Step 2.2: Calculate the residual signal source signal Thus, the residual signal is obtained. source signal in, This represents the i-th unmixed vector under the u-th execution of the blind source separation algorithm;

[0059] Step 2.3: Calculate the source signal of the residual signal and in, w represents the i-th unmixed vector i The corresponding source signal;

[0060] Step 3: Use a clustering algorithm to separate the source signal Y from the u-th cluster. u The signals were categorized and selected separately as respiratory-related signals and electrocardiogram-related signals as reference signals for the u-th separation.

[0061] Step 3.1, for Perform clustering to obtain F clusters under the u-th separation;

[0062] Step 3.2: Calculate the similarity between the average sequence of each cluster under the u-th separation and the acquired prior respiratory signal and prior ECG signal, and determine whether there is a cluster whose average sequence matches the prior respiratory signal template or the prior ECG signal template. If there is a cluster whose average sequence matches the prior respiratory signal template, assign a+1 to a and proceed to step 3.3. If there is a cluster whose average sequence matches the prior ECG signal template, assign b+1 to b and proceed to step 3.4. If neither exists, proceed to step six. Step 3.2 provides the selection logic for choosing reference signals. Considering that the ECG-related source signals and respiratory-related source signals to be selected have obvious prior characteristics in signal morphology, a prior template matching method is used to select the corresponding signals based on prior knowledge. After each ECG-related or respiratory-related cluster is selected, the subsequent constraint algorithm and stripping algorithm are executed, and the next iteration begins, until no respiratory or ECG-related source signals can be found.

[0063] Step 3.3: Select the source signal with the highest signal-to-noise ratio from the clusters that match the prior respiratory signal template as the respiratory-related source signal for the a-th stripping under the u-th separation. Then proceed to step four;

[0064] Step 3.4: Select the source signal with the highest signal-to-noise ratio from the cluster that matches the prior ECG signal template as the ECG-related source signal for the b-th stripping under the u-th separation. Then proceed to step four;

[0065] In this embodiment, considering that the actual respiratory curve is a sinusoidal undulating curve and the actual respiratory rate is generally once every 2-5 seconds, the selected prior template respiratory signal is y = sin(0.3t); the selected prior template ECG signal is a segment of a clean multi-channel surface ECG signal. Since we need to automatically select source signals with similar morphology but loose requirements on the rate of signal change from several clusters obtained by clustering, a template-based dynamic time warping method is used to automatically select the cluster with the lowest warping cost, thereby obtaining ECG-related source signals or respiratory-related source signals based on the prior template. Figure 5 It consists of an automatically selected electrocardiogram-related component and an automatically selected respiratory-related component.

[0066] Step 4: Use the time-constrained FastICA algorithm to extract the more accurate source signals corresponding to the above-mentioned respiratory and ECG related source signals, namely the respiratory and ECG signal components. The time-constrained FastICA algorithm is used to test the reliability of the decomposed respiratory source signal. It is generally believed that a reliable signal source converges to an output that is as similar as possible to the reference source signal. This determines the reliability of the source signal to be tested. At the same time, it corrects possible errors or omissions by updating the demixing vector.

[0067] ECG-related source signals or respiratory-related signals r, respectively, serves as the reference signal for the u-th separation. u Thus utilizing the reference signal r u In the extended residual signal The FastICA algorithm with time-domain constraints is executed, resulting in the ECG signal component stripped in the b-th separation under the u-th separation. Or the respiratory signal component of the a-th stripping

[0068] Step 5: or From residual signal In this embodiment, the stripping step removes the more accurate ECG and respiratory signal components selected in each step. This firstly avoids the algorithm's tendency to converge to larger ECG components, leading to local convergence and failure to extract weak diaphragmatic electromyography (EMG) signals. Secondly, it ensures that the extracted diaphragmatic EMG components are previously undetected, allowing for effective identification of all respiratory-related source signals, no matter how weak. The waveform estimated using the least squares method takes into account the different morphologies of the extracted source signals in each original signal channel, resulting in greater accuracy during stripping.

[0069] Step 5.1: Solve using the least squares method. or The original signal x corresponding to the i-th channel i The transformation matrix estimated for the a-th time Or the transformation matrix estimated in the bth iteration. The respiratory signal component was estimated using the a-th time. The waveform corresponding to the i-th channel of the original signal X or ECG signal components The waveform corresponding to the i-th channel of the original signal X Where, x i Let X be the original signal of the i-th channel in the original signal X;

[0070] Step 5.2: Convert the waveform or From residual signal Subtract from the i-th channel to obtain the residual signal of the i-th channel after the (u+1)-th separation. Or the residual signal of the i-th channel after the (u+1)-th separation. Thus, the residual signal after the (u+1)th separation is obtained. and respiratory signal components Place it into the respiratory signal source set γ.

[0071] Step 5.3: Assign u+1 to u, return to step 2.1 and execute sequentially until the maximum number of iterations u is reached. max until;

[0072] Step 6: After summing all respiratory signal components in set γ, a pure and complete diaphragmatic electromyographic signal F is obtained.

[0073] Since the respiratory (diaphragmatic electromyography) component identified by each round of the algorithm may not be a complete signal, but only a component of the original signal, the complete and pure respiratory source signal can be obtained by accumulating all the separated respiratory-related components.

[0074] The pure diaphragm electromyography signal obtained in this embodiment and the diaphragm electroactivity curve obtained based on the pure diaphragm electromyography signal are as follows: Figure 6 As shown, the actual number of respiratory source signals separated, a, is 4 in this embodiment.

[0075] In this embodiment, an electronic device includes a memory and a processor. The memory stores a program that supports the processor in executing the above-described method, and the processor is configured to execute the program stored in the memory.

[0076] In this embodiment, a computer-readable storage medium stores a computer program, which is executed by a processor to perform the steps of the above method.

Claims

1. A surface diaphragm electromyography denoising method based on blind source separation and stripping strategy, characterized in that, The procedure is as follows: Step 1: Acquire surface diaphragm electromyography (EMG) signals and perform noise reduction using frequency domain analysis to obtain multi-channel surface diaphragm EMG signals X = [x1, x2, ..., x...]. i , ..., x n ] T , and serve as the initial signal, where x i This represents the surface diaphragm electromyography signal of the i-th channel after filtering out low-frequency and high-frequency noise, where T represents transpose; Step 2: Execute the u-th blind source separation algorithm to obtain several source signals under the u-th separation: Step 2.1: Initialization: Set the total number of algorithm executions to u. max The current execution count is u, and u is initialized to 1; define and initialize the residual signal under the current u-th execution of the blind source separation algorithm. Initialize the number of respiratory signal component strippings a = 0 and the number of electrocardiogram signal component strippings b = 0; Define and initialize the set of respiratory signal sources γ and its corresponding unmixed vector set ε as empty sets; Step 2.2: Use equation (1) to analyze the residual signal after the u-th separation. The signal is expanded to obtain the extended residual signal after the u-th execution of the blind source separation algorithm. In equation (1), K is the delay factor; This represents the residual signal of the i-th channel during the u-th separation. The delayed signal after a delay of K+1; t is the current time; T represents the transpose; Step 2.3: Calculate the residual signal of the i-th channel in the u-th iteration. source signal Thus, the residual signal in the u-th separation is obtained. source signal in, This represents the i-th unmixed vector under the u-th execution of the blind source separation algorithm; Step 3: Use a clustering algorithm to separate the source signal Y from the u-th cluster. u The signals were categorized and selected separately as respiratory-related signals and electrocardiogram-related signals as reference signals for the u-th separation. Step 3.1, for Perform clustering to obtain F clusters under the u-th separation; Step 3.2: Calculate the similarity between the average sequence of each cluster under the u-th separation and the acquired prior respiratory signal and prior ECG signal, and determine whether there is a cluster whose average sequence matches the prior respiratory signal template or the prior ECG signal template; if there is a cluster whose average sequence matches the prior respiratory signal template, assign a+1 to a and then proceed to step 3.3; if there is a cluster whose average sequence matches the prior ECG signal template, assign b+1 to b and then proceed to step 3.4; if neither exists, proceed to step six. Step 3.3: Select the source signal with the highest signal-to-noise ratio from the clusters that match the prior respiratory signal template as the respiratory-related source signal for the a-th stripping under the u-th separation. Then, proceed to step four; Step 3.4: Select the source signal with the highest signal-to-noise ratio from the cluster that matches the prior ECG signal template as the ECG-related source signal for the b-th stripping under the u-th separation. Then, proceed to step four; Step 4: The ECG-related source signals or respiratory-related signals r, as the reference signal under the u-th separation u Thus utilizing the reference signal r u In the extended residual signal The FastICA algorithm with time-domain constraints is executed, resulting in the ECG signal component stripped in the b-th separation under the u-th separation. Or the respiratory signal component of the a-th stripping Step 5: or From residual signal The residual signal after the (u+1)th separation is obtained by stripping. Assign u+1 to u, return to step 2.2 and execute sequentially until the maximum number of iterations u is reached. max until; Step 6: After summing all respiratory signal components in set γ, a pure and complete diaphragmatic electromyographic signal F is obtained.

2. The surface diaphragm electromyography denoising method based on blind source separation and stripping strategy according to claim 1, characterized in that, Step 1 includes: Step 1.1: Acquire multi-channel surface electromyographic signals of the diaphragm containing noise and interference using an array of electrode pads with rows θ and columns ρ. The signals are C = [c1, c2, ..., c...]. i c n ] T , where c i The signal represents the surface diaphragm electromyography signal of the i-th channel containing noise and interference, where n represents the total number of channels, and n = θ × ρ; Step 1.2: Process the multichannel surface diaphragm electromyography signal C = [c1, c2, ..., c] containing noise and interference. i c n ] T Performing a discrete Fourier transform yields the multi-channel surface diaphragm electromyography signal X = [x1, x2, ..., x...] after filtering out low-frequency and high-frequency noise. i , ..., x n ] T .

3. The surface diaphragm electromyography denoising method based on blind source separation and stripping strategy according to claim 2, characterized in that, Step 5 includes: Step 5.1: Solve using the least squares method. or The original signal x corresponding to the i-th channel i The transformation matrix estimated for the a-th time Or the transformation matrix estimated in the bth iteration. The respiratory signal component was estimated using the a-th time. The waveform corresponding to the i-th channel of the original signal X or ECG signal components The waveform corresponding to the i-th channel of the original signal X Where, x i Let X be the original signal of the i-th channel in the original signal X; Step 5.2: Convert the waveform or From residual signal Subtract from the i-th channel to obtain the residual signal of the i-th channel after the (u+1)-th separation. Or the residual signal of the i-th channel after the (u+1)-th separation. Thus, the residual signal after the (u+1)th separation is obtained. and respiratory signal components Place it into the respiratory signal source set γ.

4. An electronic device, comprising a memory and a processor, characterized in that, The memory is used to store a program that supports the processor in executing the surface diaphragm electromyography denoising method based on the blind source separation and stripping strategy according to any one of claims 1-3, and the processor is configured to execute the program stored in the memory.

5. A computer-readable storage medium storing a computer program thereon, characterized in that, The computer program, when run by the processor, executes the steps of the surface diaphragm electromyography denoising method based on the blind source separation and stripping strategy described in any of claims 1-3.

Images