A composite empirical mode decomposition respiration extraction method and system based on electrocardiosignal

Abstract

The application provides a composite empirical mode decomposition respiration extraction method and system based on an electrocardiosignal, and the method comprises the following steps: simultaneously performing ensemble empirical mode decomposition on a Gaussian white noise and an original electrocardiosignal of a tested person collected, obtaining each intrinsic mode function component, selecting an IMF frequency band range of respiration by using a Fourier transform technology, subtracting each IMF component in the corresponding central frequency band range to eliminate white noise residues in the original EEMD decomposition, and obtaining a new IMF; calculating the correlation of the new IMF, the IMF obtained from the original electrocardiosignal and a measured respiration signal respectively, determining the optimal amplitude noise coefficient alpha of the algorithm according to the maximum principle of the correlation coefficient increment, and using the optimal amplitude noise coefficient alpha to reconstruct the respiration signal. The respiration extraction method has the characteristics of simple operation, high accuracy and strong robustness.

Description

Technical Field

[0001] This invention relates to the field of respiratory signal extraction technology, specifically to a method and system for respiratory extraction based on composite empirical mode decomposition of electrocardiogram signals. Background Technology

[0002] Breathing is a vital physiological parameter, commonly associated with heart disease, sleep apnea syndrome, and anxiety. A coupling effect between respiration and heart rate has been demonstrated. However, monitoring respiration requires bulky equipment, which may interfere with natural breathing. Obtaining respiratory signals from single-lead electrocardiogram (ECG) signals can effectively reduce monitoring costs, eliminate the limitations of dedicated respiratory monitoring devices, improve user comfort, and is more suitable for outpatient and home monitoring.

[0003] EMD (Empirical Mode Decomposition) algorithms can be used to analyze nonlinear and non-stationary signal sequences, exhibiting high signal-to-noise ratios and good time-frequency focusing. However, inherent limitations of the algorithm obscure the physical meaning of individual intrinsic mode functions, leading to severe mode aliasing in the time-frequency distribution. EEMD (Empirical Empirical Mode Decomposition) represents a significant improvement over EMD. By performing multiple EMD decompositions with superimposed Gaussian white noise, it leverages the statistical characteristic of Gaussian white noise's uniform frequency distribution, effectively addressing the mode aliasing problem through noise-assisted analysis. However, in the EEMD decomposition algorithm, the setting of the noise amplitude coefficient still relies on human experience. Although the impact of noise on the results decreases with increasing overall average iteration count, the algorithm's time cost also increases accordingly, and the residual white noise cannot be ignored. This paper proposes a respiratory extraction method based on composite empirical mode decomposition to derive EDR (ECG-derived respiration) signals from single-lead ECGs. This method is adaptive in selecting the optimal amplitude noise coefficient and, to some extent, eliminates residual white noise in the reconstructed signal, providing a new solution for EDR signal extraction. Summary of the Invention

[0004] To overcome the shortcomings of the prior art, this invention first proposes a respiratory extraction method and system based on composite empirical mode decomposition. Based on the original electrocardiogram signal and Gaussian white noise collected from the subject, this invention can quickly and accurately calculate the subject's electrocardiogram-derived respiration.

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

[0006] A method for respiratory extraction based on composite empirical mode decomposition of electrocardiogram signals includes the following steps:

[0007] Step S1: Acquire Gaussian white noise signal and collect raw ECG signal. Perform ensemble empirical mode decomposition on the collected Gaussian white noise signal and the collected raw ECG signal respectively to obtain each IMF component ① and IMF component ② of the Gaussian white noise signal and the collected raw ECG signal respectively.

[0008] Step S2: Calculate and screen the respiratory frequency range in IMF component ① and IMF component ② using Fourier transform technology. Then, subtract each IMF component ① and IMF component ② from the corresponding center frequency range to eliminate the white noise residue in the original ECG set empirical mode decomposition and obtain the new IMF component ③.

[0009] Step S3: Calculate the correlation between the new IMF component ③, the IMF component ② obtained from the original ECG signal, and the measured respiratory signal. Compare the magnitude of the increment of the correlation coefficient between the two. Determine the optimal amplitude noise coefficient α of the algorithm based on the principle of maximizing the increment of the correlation coefficient. Select each IMF component within the respiratory frequency band and sum them to calculate the reconstructed respiratory signal.

[0010] Furthermore, the specific implementation of set empirical mode decomposition in step S1 includes the following sub-steps:

[0011] Step S11: Set the original electrocardiogram signal as y(t), the Gaussian white noise signal as g(t), the preset total average number of times as m, and the noise amplitude coefficient as α;

[0012] Step S12: The white noise sequence with a standard normal distribution added for the i-th time is n. i (t), calculate the noisy signal of the i-th experiment: y i (t)=y(t)+αn i (t), g i (t)=g(t)+αn i (t);

[0013] Step S13: For y i (t), g i If (t) performs the i-th EMD process, then:

[0014]

[0015] Where, x ij (t), k ij (t) represents the average value of the j-th IMF component obtained from the EMD decomposition of the original electrocardiogram signal and Gaussian white noise, respectively; r i (t), l i (t) is the average value of its corresponding residual term.

[0016] Furthermore, the specific implementation of the composite noise reduction process in step S2 includes the following sub-steps:

[0017] Step S21: Subtract the IMF components obtained in steps S12 and S13 within their respective frequency bands to obtain a new IMF component W. ij (t)=x ij (t)-k ij (t) and the average of the new residuals T ij (t)=r i (t)-l i (t);

[0018] Step S22: Repeat steps S13 and S21 until i = m, and calculate the average value of the j-th IMF component obtained after m EMD decompositions. and the average of the residuals

[0019]

[0020] Step S23: The final result after denoising using Gaussian white noise is expressed as follows:

[0021]

[0022] Furthermore, in step S21, the respiratory frequency band ranges in IMF component ① and IMF component ② are first screened, and the corresponding frequency bands are extracted from the screened IMF component ① and IMF component ② and subtracted from each other, while the remaining frequency bands are not subtracted.

[0023] Further, step S3 specifically includes: selecting a respiratory frequency range, and using the IMF component in the frequency band as the component for reconstructing the respiration.

[0024] Using Pearson correlation coefficient The index is used to measure the correlation between the IMF component and the measured respiration, where Cov(X·Y) is the covariance of X and Y, Var[X] is the variance of X, and Var[Y] is the variance of Y.

[0025] Furthermore, the EEMD decomposition components x of the electrocardiogram signal are compared separately. ij (t), composite empirical mode decomposition component W ij The increment of the correlation coefficient between (t) and the measured respiratory signal is used to determine the optimal amplitude noise coefficient α of the algorithm based on the principle of maximizing the increment of the correlation coefficient, which is then used to reconstruct the respiratory signal.

[0026] The present invention also provides a composite empirical mode decomposition respiratory extraction system based on electrocardiogram signals, comprising:

[0027] The signal acquisition module is used for real-time, long-term, and accurate acquisition of electrocardiogram signals from a person in a resting state.

[0028] The signal processing module is used to perform ensemble empirical mode decomposition on the signal transmitted by the signal acquisition module and the Gaussian white noise signal to obtain the IMF components① and ② of the Gaussian white noise signal and the acquired raw electrocardiogram signal, respectively.

[0029] The composite noise reduction module is used to subtract each IMF component ① and IMF component ② from their corresponding center frequency bands to eliminate the white noise residue in the original ECG set empirical mode decomposition and obtain a new IMF component ③.

[0030] The respiratory reconstruction module is used to calculate the correlation between the new IMF component ③, the IMF component ② obtained from the original electrocardiogram signal and the measured respiratory signal, respectively. It compares the magnitude of the increment of the correlation coefficient between the two, determines the optimal amplitude noise figure α of the algorithm based on the principle of maximizing the increment of the correlation coefficient, selects each IMF component in the respiratory frequency band to sum, and calculates the reconstructed respiratory signal.

[0031] Furthermore, in the signal acquisition module, the electrode pads and electrode vest, mainly made of flexible dry electrodes, are worn and fitted to the user's upper left chest cavity to achieve real-time, long-range, and accurate acquisition of electrocardiogram signals in the human body at rest, and the acquired data is directly sent to the signal processing module.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] The algorithm for deriving respiratory signals from electrocardiogram (ECG) signals is adaptive and can obtain good EDR signals without requiring preprocessing of ECG data. The EDR signals obtained by this invention have a high similarity to respiratory signals synchronously recorded by commercial instruments. Compared with the EEMD algorithm, the EDR extraction method proposed in this paper has higher accuracy and robustness, and can be used in various applications such as sleep apnea detection and home-based respiratory monitoring. Attached Figure Description

[0034] Figure 1 This is a schematic diagram of the composite empirical mode decomposition respiration extraction process in an embodiment of the present invention;

[0035] Figure 2 The following are the time-domain plot and histogram of Gaussian white noise in an embodiment of the present invention, wherein (a) is the time-domain plot u(n) of Gaussian white noise; and (b) is the histogram of u(n).

[0036] Figure 3 This is a detailed diagram of the EEMD decomposition of Gaussian white noise in an embodiment of the present invention;

[0037] Figure 4This is a detailed diagram of the ECG composite empirical mode decomposition in an embodiment of the present invention;

[0038] Figure 5 This is a graph showing the incremental correlation coefficient of IMF under different noise figures α in the embodiments of the present invention;

[0039] Figure 6 This is an example diagram comparing the cycle of respiration extracted by the algorithm and the cycle of reference respiration in an embodiment of the present invention;

[0040] Figure 7 This is an example diagram comparing the breaths extracted by different algorithms with the original breaths in an embodiment of the present invention. Detailed Implementation

[0041] To make the technical problems, technical solutions, and beneficial effects of the embodiments of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining the present invention and are not intended to limit the present invention.

[0042] It should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", and "outer" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the embodiments of the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the present invention.

[0043] In the description of this invention, unless otherwise stated, the term "connection" should be interpreted broadly, and may refer to a fixed connection, a detachable connection, or an integral connection. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0044] The implementation process of the present invention will be further described in detail below with reference to specific accompanying drawings and examples.

[0045] Example 1

[0046] Based on the shortcomings in the background technology, such as Figure 1 As shown, this embodiment provides a method for respiratory extraction based on composite empirical mode decomposition of electrocardiogram signals, comprising the following steps:

[0047] Step S1: Acquire Gaussian white noise signal and collect raw ECG signal. Perform ensemble empirical mode decomposition on the collected Gaussian white noise signal and the collected raw ECG signal respectively to obtain each IMF component ① and IMF component ② of the Gaussian white noise signal and the collected raw ECG signal respectively.

[0048] Step S2: Calculate and screen the respiratory frequency range in IMF component ① and IMF component ② using Fourier transform technology. Then, subtract each IMF component ① and IMF component ② from the corresponding center frequency range to eliminate the white noise residue in the original ECG set empirical mode decomposition and obtain the new IMF component ③.

[0049] Step S3: Calculate the correlation between the new IMF component ③, the IMF component ② obtained from the original ECG signal, and the measured respiratory signal. Compare the magnitude of the increment of the correlation coefficient between the two. Determine the optimal amplitude noise coefficient α of the algorithm based on the principle of maximizing the increment of the correlation coefficient. Select each IMF component within the respiratory frequency band and sum them to calculate the reconstructed respiratory signal.

[0050] The specific implementation of set empirical mode decomposition in step S1 includes the following sub-steps:

[0051] Step S11: Set the original electrocardiogram signal as y(t), the Gaussian white noise signal as g(t), the preset total average number of times as m, and the noise amplitude coefficient as α;

[0052] Step S12: The white noise sequence with a standard normal distribution added for the i-th time is n. i (t), calculate the noisy signal of the i-th experiment: y i (t)=y(t)+αn i (t), g i (t)=g(t)+αn i (t);

[0053] Step S13: For y i (t), g i If (t) performs the i-th EMD process, then:

[0054]

[0055] Where, x ij (t), k ij (t) represents the average value of the j-th IMF component obtained from the EMD decomposition of the original electrocardiogram signal and Gaussian white noise, respectively; r i (t), l i (t) is the average value of its corresponding residual term.

[0056] The specific implementation of the composite noise reduction process in step S2 includes the following sub-steps:

[0057] Step S21: Subtract the IMF components obtained in steps S12 and S13 within their respective frequency bands to obtain a new IMF component W. ij (t)=x ij (t)-k ij(t) and the average of the new residuals T ij (t)=r i (t)-l i (t);

[0058] Step S22: Repeat steps S13 and S21 until i = m, and calculate the average value of the j-th IMF component obtained after m EMD decompositions. and the average of the residuals

[0059] Step S23: The final result after denoising using Gaussian white noise is expressed as follows:

[0060]

[0061] Furthermore, in step S21, the respiratory frequency band ranges in IMF component ① and IMF component ② are first screened, and the corresponding frequency bands are extracted from the screened IMF component ① and IMF component ② and subtracted from each other, while the remaining frequency bands are not subtracted.

[0062] like Figure 2 As shown, (a) is the Gaussian white noise time-domain plot u(n), with the horizontal axis representing the number of points N used, the same as the original ECG, and the vertical axis representing the amplitude uniformly distributed between 0 and 1; (b) is the histogram of u(n), with the horizontal axis representing the range of u(n) and the vertical axis representing the frequency.

[0063] In the above embodiments, each IMF component ① is as follows Figure 3 As shown, the decomposition yields 12 IMF components, and each IMF component ② is as follows. Figure 4 As shown, the composite empirical mode decomposition yields 13 IMF components, which have multi-resolution characteristics and can reflect more scale information of the original electrocardiogram signal.

[0064] In this embodiment, the effectiveness of the respiratory extraction algorithm of the present invention is tested using actual measured breathing. At the same time, it is compared with the EMD and EEMD decomposition algorithms to conduct respiratory extraction experiments and verify the superiority of the method of the present invention.

[0065] Furthermore, the correlation between the new IMF component ③, the IMF component ② obtained from the original electrocardiogram signal and the measured respiratory signal is calculated respectively, and the optimal amplitude noise figure α of the algorithm is determined by the principle of maximizing the correlation coefficient increment.

[0066] Specifically, multiple IMF components are obtained from the original signal by composite empirical mode decomposition. This paper utilizes the Pearson correlation coefficient. The index is used to measure the correlation between the IMF component and measured respiration. The statistical significance of the IMF component is assessed using a p-value. A p-value less than 0.05 is considered statistically significant, and the correlation R is considered significant.

[0067] The IMF frequency band range for reconstructing respiration is selected. Since respiration and heartbeat occur in different frequency bands, respiration can be separated from the electrocardiogram signal through filtering or decomposition methods. Under normal circumstances, a person's respiratory rate is generally 0.1-0.5 Hz, while the heartbeat rate is generally 0.8-2 Hz. Considering that the patient may have shortness of breath or apnea, this embodiment selects a respiratory rate range of 0.07-0.75 Hz, using the IMF components in the frequency band as the components for reconstructing respiration. Table 1 below shows the maximum center frequency of each IMF obtained by different decomposition methods.

[0068] Table 1: Maximum center frequency of IMF within the respiratory band

[0069]

[0070] Determine the optimal amplitude noise figure for the algorithm. Compare the EEMD decomposition components x of the ECG signal. ij (t), Figure 5 The composite empirical mode decomposition component W is shown. ij The increment of the correlation coefficient between (t) and the measured respiratory signal is represented by the horizontal axis, which represents the IMF component in the respiratory frequency band, and the vertical axis represents the increment of the correlation coefficient. The optimal amplitude noise coefficient α of the algorithm is determined by the principle of maximizing the increment of the correlation coefficient, and is used to reconstruct the respiratory signal.

[0071] Table 2: Comparison of P-values ​​for different IMF components and primitive respiration

[0072]

[0073] Table 3: Comparison of R values ​​for different IMF components and primitive respiration

[0074]

[0075] To quantify the error between the breath extraction results and the original breath in this invention, this embodiment uses the mean square error (MSE), root mean square error (RMSE), and mean absolute error (MAE) to evaluate the accuracy of the algorithm.

[0076] Specifically, in Represents the predicted value, y i These are actual measured values.

[0077] The detected respiratory time in the extracted respiratory signal is compared with the corresponding reference respiratory signal time. A reference respiratory time window of 2 seconds is determined for each extracted breath. Each respiratory peak or trough is marked to define the respiratory beat. The specific steps are as follows:

[0078] Step 1: Represent the breathing sequence as: V = [v1, v2, ..., v n ].

[0079] Step 2: Calculate the first-order difference vector of V:

[0080] Diff v (i)=V(i+l)-V(i), i∈1,2,…,N-1

[0081] Step 3: Perform a sign function operation on the difference components, Trend = sign(Diff). v ), that is, traversing the Diff v If Diff v If (i)>0, then take 1; if Diff v If (i) < 0, then take -1; otherwise take 0.

[0082]

[0083] Step 4: Traverse the Trend vector from the tail and perform the following operations:

[0084]

[0085] Step 5: Perform a first-order difference operation on the Trend vector, as in Step 2, to obtain:

[0086] R = diff(Trend)

[0087] Step 6: Traverse the obtained difference vector R. If R(i) = -2, then i+1 is a peak position of the projection vector V, and the corresponding peak value is V(i+1); if R(i) = 2, then i+1 is a trough position of the projection vector V, and the corresponding trough value is V(i+1).

[0088] Figure 6 This shows an example comparing the cycles of extracted and reference breaths. Within a 2-second time window, one alternation of a peak and trough is defined as a respiratory cycle, which is essentially equal to the measured number of breaths. Figure 7 Example image comparing breaths extracted by different algorithms with the original breaths.

[0089] Example 2

[0090] This embodiment provides a composite empirical mode decomposition respiratory extraction system based on electrocardiogram signals, including:

[0091] The signal acquisition module is used to acquire ECG signals in real time, over long distances, and accurately in a resting state. The signal acquisition module consists of electrode pads and an electrode vest made primarily of flexible dry electrodes, which are worn and fitted to the user's upper left chest cavity to achieve real-time, long-distance, and accurate acquisition of ECG signals in a resting state. The acquired data is then directly sent to the signal processing module.

[0092] The signal processing module is used to perform ensemble empirical mode decomposition on the signal transmitted by the signal acquisition module and the Gaussian white noise signal to obtain the IMF components① and ② of the Gaussian white noise signal and the acquired raw electrocardiogram signal, respectively.

[0093] The composite noise reduction module is used to subtract each IMF component ① and IMF component ② from their corresponding center frequency bands to eliminate the white noise residue in the original ECG set empirical mode decomposition and obtain a new IMF component ③.

[0094] The respiratory reconstruction module is used to calculate the correlation between the new IMF component ③, the IMF component ② obtained from the original electrocardiogram signal and the measured respiratory signal, respectively. It compares the magnitude of the increment of the correlation coefficient between the two, determines the optimal amplitude noise figure α of the algorithm based on the principle of maximizing the increment of the correlation coefficient, selects each IMF component in the respiratory frequency band to sum, and calculates the reconstructed respiratory signal.

[0095] The above are merely preferred embodiments of the present invention and are not intended to limit the implementation methods and protection scope of the present invention. Those skilled in the art should recognize that any equivalent substitutions and obvious changes made based on the content of this specification should be included within the protection scope of the present invention.

Claims

1. A method for respiratory extraction based on composite empirical mode decomposition of electrocardiogram signals, characterized in that, It includes the following steps: Step S1: Acquire Gaussian white noise signal and collect raw ECG signal. Perform ensemble empirical mode decomposition on the collected Gaussian white noise signal and the collected raw ECG signal respectively to obtain each IMF component ① and IMF component ② of the Gaussian white noise signal and the collected raw ECG signal respectively. Step S2: Calculate and screen the respiratory frequency range in IMF component ① and IMF component ② using Fourier transform technology. Then, subtract each IMF component ① and IMF component ② from the corresponding center frequency range to eliminate the white noise residue in the original ECG set empirical mode decomposition and obtain the new IMF component ③. Step S3: Calculate the correlation between the new IMF component ③, the IMF component ② obtained from the original ECG signal, and the measured respiratory signal, respectively. Compare the increments of their correlation coefficients and determine the optimal amplitude noise figure of the algorithm based on the principle of maximizing the increment of the correlation coefficient. The reconstructed respiratory signal is obtained by summing the IMF components within the selected respiratory frequency band.

2. The method for respiratory extraction based on composite empirical mode decomposition of electrocardiogram signals according to claim 1, characterized in that, The specific implementation of the set empirical mode decomposition in step S1 includes the following sub-steps: Step S11: Set the original ECG signal as Gaussian white noise signal The preset overall average frequency is m The noise amplitude coefficient is ; Step S12: The next added white noise sequence with a standard normal distribution is: Calculate the first The noisy signal in this experiment: , ; Step S13: For , Proceed to the first After the second EMD treatment: ， ， in, , The EMD decomposition of the original electrocardiogram signal and Gaussian white noise respectively yielded the first... Average value of each IMF component; , It is the average value of its corresponding residual terms.

3. The method for respiratory extraction based on composite empirical mode decomposition of electrocardiogram signals according to claim 2, characterized in that, The specific implementation of the composite noise reduction process in step S2 includes the following sub-steps: Step S21: Subtract the IMF components obtained in steps S12 and S13 within their respective frequency bands to obtain new IMF components. and the average of the new residuals ; Step S22: Repeat steps S13 and S21 until... Please explain the process. The first EMD decomposition yielded the... Average of IMF components and the average of the residuals ; Step S23: The final result after denoising using Gaussian white noise is expressed as follows: ， in, n For the process The total number of IMF components obtained from each EMD decomposition.

4. The method for respiratory extraction based on composite empirical mode decomposition of electrocardiogram signals according to claim 3, characterized in that, In step S21, the respiratory frequency ranges in IMF component ① and IMF component ② are first screened, and the corresponding frequency bands are extracted from the screened IMF component ① and IMF component ② and subtracted. The remaining frequency bands are not subtracted.

5. The method for respiratory extraction based on composite empirical mode decomposition of electrocardiogram signals according to claim 1, characterized in that, Step S3 specifically includes: selecting a respiratory frequency range, and using the IMF component in the frequency band as the component for reconstructing the respiration. Using Pearson correlation coefficient The index is used to measure the correlation between IMF components and measured respiration, among which, for and covariance, for variance for The variance.

6. The method for respiratory extraction based on composite empirical mode decomposition of electrocardiogram signals according to claim 5, characterized in that, Compare the EEMD decomposition components of the electrocardiogram signal respectively Composite empirical mode decomposition components The optimal amplitude noise figure for the algorithm is determined by maximizing the correlation coefficient increment between the measured respiratory signal and the actual respiratory signal. It is used to reconstruct respiratory signals.

7. A composite empirical mode decomposition respiratory extraction system based on electrocardiogram signals, characterized in that, include: The signal acquisition module is used for real-time, long-term, and accurate acquisition of electrocardiogram signals from a person in a resting state. The signal processing module is used to perform ensemble empirical mode decomposition on the signal transmitted by the signal acquisition module and the Gaussian white noise signal to obtain the IMF components① and ② of the Gaussian white noise signal and the acquired raw electrocardiogram signal, respectively. The composite noise reduction module is used to subtract each IMF component ① and IMF component ② from their corresponding center frequency bands to eliminate the white noise residue in the original ECG set empirical mode decomposition and obtain a new IMF component ③. The respiratory reconstruction module is used to calculate the correlation between the new IMF component ③, the IMF component ② obtained from the original ECG signal, and the measured respiratory signal, respectively. It compares the increments of the correlation coefficients and determines the optimal amplitude noise figure of the algorithm based on the principle of maximizing the increment of the correlation coefficient. The reconstructed respiratory signal is obtained by summing the IMF components within the selected respiratory frequency band.

8. The composite empirical mode decomposition respiratory extraction system based on electrocardiogram signals according to claim 7, characterized in that: In the signal acquisition module, electrode pads and electrode vests made primarily of flexible dry electrodes are worn and fitted to the user's upper left chest cavity to achieve real-time, long-range, and accurate acquisition of electrocardiogram signals in the human body at rest, and the acquired data is directly sent to the signal processing module.

Images