Electronic sphygmomanometer human body pulse signal noise reduction method

Through IAO-CEEMDAN decomposition and wavelet threshold denoising technology, combined with threshold screening and residual reconstruction, the noise impact problem during electronic sphygmomanometer measurement is solved, efficient denoising of pulse signals is achieved, and measurement accuracy is significantly improved.

CN118981606APending Publication Date: 2024-11-19HANGZHOU DIANZI UNIV
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411082714.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-08
Publication Date
2024-11-19

AI Technical Summary

Technical Problem

Electronic blood pressure meters are susceptible to internal and external noise when measuring pulse signals of human bodies, resulting in measurement errors. The effect of existing noise reduction methods depends on the selection of wavelet basis and threshold values, and there is modal aliasing and endpoint effects that affect the noise reduction effect.

Method used

IAO-CEEMDAN (adaptive complete empirical modal decomposition based on Tianying Optimization Algorithm) is used to decompose the pulse signal, combine the thresholds of sample entropy, variance contribution rate and correlation coefficient to determine the effective IMF component, and use the wavelet threshold denoising method to remove it. Noise information and the noise-reduced pulse signal is reconstructed through residual components.

Benefits of technology

It effectively removes mixed noise in pulse signals, retains more effective information, and shows better noise reduction effect and stability than the previous denoising methods.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118981606A_ABST
    Figure CN118981606A_ABST
Patent Text Reader

Abstract

The invention discloses a human body pulse signal noise reduction method for an electronic sphygmomanometer, which comprises the following steps: firstly, carrying out preprocessing operation on a human body pulse signal to be subjected to noise reduction measured by the electronic sphygmomanometer, and carrying out IAO-CEEMDAN decomposition on the preprocessed signal to obtain a plurality of IMF components and a residual component; secondly, a sample entropy value, a variance contribution rate and a threshold value of a correlation coefficient are set, effective IMF components are determined, the effective IMF components are reserved, and invalid IMF components are rejected; and then denoising the effective IMF component by using a wavelet threshold, and carrying out secondary denoising. And finally, superposing the de-noised IMF components, and adding the residual component to complete reconstruction of the human body pulse signal so as to obtain a de-noised human body pulse signal. According to the method, mixed noise in the measured pulse signals is effectively removed, and more effective information in the pulse signals measured by the electronic sphygmomanometer is reserved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of pulse signal denoising, in particular to a method for denoising a human body pulse signal of an electronic sphygmomanometer. Background Art

[0002] In recent years, people in my country have paid more and more attention to their physical health, and blood pressure can roughly reflect a person's physical health. In order for everyone to measure their blood pressure at home, they usually use an electronic blood pressure monitor to measure their blood pressure. However, during the measurement process, the pressure sensor in the electronic blood pressure monitor will be affected by internal and external environmental noise, resulting in errors in the measured blood pressure value. Therefore, it is very necessary to effectively reduce the noise of the pulse signal obtained by the pressure sensor.

[0003] In response to such problems, the main denoising methods used in the early days were wavelet decomposition-based denoising methods, such as using an improved threshold function to perform wavelet denoising on the signal, which achieved certain results, but the denoising effect often depends on the selection of the wavelet basis and the threshold. The empirical mode decomposition (EMD) proposed later is suitable for nonlinear and non-stationary signal processing methods. This method decomposes the signal into several intrinsic mode components (IMFs) and a remainder (res) arranged in sequence from high frequency to low frequency. It has certain practical physical meaning and frequency stratification performance, and is a fully adaptive decomposition method. However, if there are abnormal mutations in the signal, modal aliasing and endpoint effects will occur, thus affecting the denoising effect. Summary of the invention

[0004] In view of the problems existing in the existing electronic sphygmomanometers, the present invention proposes a method for reducing noise of human pulse signals of an electronic sphygmomanometer, comprising the following steps:

[0005] S1. Preprocess the human pulse signal to be denoised measured by the electronic sphygmomanometer. According to the law of the human pulse signal, observe whether the measured signal is usable. Discard the signal that does not meet the measurement standard and retain the signal that meets the requirement to obtain the preprocessed signal. Decompose the preprocessed signal by IAO-CEEMDAN (adaptive complete empirical mode decomposition based on Sky Eagle optimization algorithm) to obtain multiple IMF components and one residual component.

[0006] S2. Set the thresholds of sample entropy, variance contribution rate and correlation coefficient to determine the valid IMF components, retain the valid IMF components and discard the invalid IMF components.

[0007] S3, the effective IMF component is subjected to secondary denoising by adopting the wavelet threshold denoising method, and the wavelet threshold denoising method of the sym8 wavelet basis function is selected to remove the noise information in the effective IMF component.

[0008] S4. Superimpose the denoised IMF components and add the residual components to complete the reconstruction of the human pulse signal and obtain the denoised human pulse signal.

[0009] In step S1, the pulse signal decomposition process based on IAO-CEEMDAN (adaptive complete empirical mode decomposition based on Skyhawk optimization algorithm) is:

[0010] S1.1. Initialize the parameters of CEEMDAN and IAO.

[0011] S1.2. Use the current CEEMDAN parameters to decompose the human pulse signal and obtain a set of IMF components.

[0012] S1.3. Select a fitness function, and calculate the fitness value of each IMF component based on the selected fitness function.

[0013] S1.4. Use IAO to optimize CEEMDAN parameters to minimize the fitness value.

[0014] S1.5. When the fitness value is minimized, the parameters of CEEMDAN are updated according to the results of IAO optimization.

[0015] S1.6. Based on the obtained parameters, the human pulse signal is decomposed by CEEMDAN to obtain multiple IMF components and residual components, where the decomposition formula of CEEMDAN is:

[0016]

[0017] Where: x(t) represents the original signal measured by the electronic sphygmomanometer; R(t) represents the monotone indecomposable residual component; IMF k Represents the k IMF components after decomposition.

[0018] Furthermore, in step S2, according to the result of decomposition in S1, the IMF components whose sample entropy is lower than the set threshold and whose variance contribution rate and correlation coefficient are higher than the set threshold are retained, otherwise they are discarded. Such preliminary processing and subsequent wavelet threshold denoising have the best effect.

[0019] Furthermore, the wavelet threshold denoising step in step S3 is as follows: temporarily name the remaining valid IMF components as signal x(t), select the wavelet basis sym8, decompose the number of layers into 5 layers, use the heursure rule as the threshold rule, and use the soft threshold method as the threshold function. First, perform discrete wavelet transform on the signal according to the selected rule to obtain a set of wavelet transform coefficients W(a, b) which can be expressed as:

[0020]

[0021] Where: W(a,b) is the wavelet transform coefficient, is the discrete wavelet basis function, and a and b are the scale factor and translation factor; by changing the value (a, b), a series of wavelet coefficients {W(a, b)} are obtained; after obtaining the wavelet coefficients {W(a, b)} at various scales, {W(a, b)} is screened by the threshold function to achieve the denoising effect. The soft threshold function is expressed as:

[0022]

[0023] Among them, W′(a,b) is the wavelet coefficient after threshold processing, W(a,b) is the wavelet coefficient after wavelet decomposition, and λ is the threshold; from the above formula, it can be seen that the soft threshold denoising method is to compare |W(a,b)| with λ. The coefficient less than λ is the high-frequency noise coefficient, and the value of W′(a,b) is directly set to zero; for the wavelet coefficient greater than the threshold, when W(a,b) is less than 0, the value of W′(a,b) is the sum of W(a,b) and λ, and when W(a,b) is greater than 0, the value of W′(a,b) is the difference between W(a,b) and λ.

[0024] The new wavelet coefficients {W′(a,b)} are used to reconstruct the noise signal x by using the inverse wavelet transform ′ (t) The inverse wavelet transform formula is shown as follows:

[0025]

[0026] Where x ′ (t) is the reconstructed IMF component signal, that is, the result after secondary noise removal. represents discrete wavelet basis function, a and b are scale factor and translation factor.

[0027] Further, in step S4, after performing wavelet threshold denoising on all retained IMF components, the denoised IMF components are obtained, and the pulse signal is reconstructed using the formula: The denoised pulse signal X(t) is reconstructed using the following formula:

[0028]

[0029] Where: IMF′ n is the IMF component after wavelet threshold denoising; N is the number of IMF components containing more valid signals screened out in step S2, X(t) is the pulse signal after denoising, and R(t) is the monotone indecomposable residual component after CEEMDAN decomposition.

[0030] Beneficial effects of the present invention:

[0031] The denoising method for the pulse signal of an electronic sphygmomanometer based on the IAO-CEEMDAN-WT threshold provided by the present invention is convenient for effectively removing the mixed noise in the measured pulse signal. Compared with the previous denoising method, it is more conducive to retaining more effective information in the pulse signal measured by the electronic sphygmomanometer. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The present invention can be better understood by referring to the description given below in conjunction with the accompanying drawings, which together with the following detailed description are included in this specification and form a part of this specification, and are used to further illustrate the preferred embodiments of the present invention and explain the principles and advantages of the present invention.

[0033] Figure 1 is a schematic diagram of a pulse simulation signal with noise provided by an embodiment of the present invention;

[0034] Figure 2 A convergence diagram of the fitness function of the Skyhawk optimization algorithm provided by an embodiment of the present invention;

[0035] Figure 3 A schematic diagram of the distribution of IMF components of a pulse signal after IAO-CEEMDAN decomposition is performed according to an embodiment of the present invention;

[0036] Figure 4 A schematic diagram of the variance contribution rate of each IMF component provided in an embodiment of the present invention;

[0037] Figure 5 Schematic diagram of CEEMDAN, WT threshold, and IAO-CEEMDAN-WT threshold after denoising provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0038] A method for reducing the noise of human pulse signals of an electronic sphygmomanometer is firstly to obtain a set of pulse signals measured by an electronic sphygmomanometer that meet the application standards, as shown in the schematic diagram. Figure 1 As shown in the figure, there are spectrum diagrams and frequency domain diagrams of pulse signals containing noise. It is known that the frequency of human pulse wave signals is mainly distributed in the range of 0.5Hz to 10Hz, so an algorithm is needed to remove the high-frequency noise in the pulse signal measured by the electronic blood pressure monitor.

[0039] First, the parameters of IAO and CEEMDAN are randomly filled in, and the human pulse signal is decomposed using the current CEEMDAN parameters to obtain a set of IMF components.

[0040] Select a fitness function, and then calculate the fitness value of each IMF based on the selected fitness function; the selection of the fitness function of the Eagle optimization algorithm has the following options based on the characteristics of CEEMDAN: 1. Minimum permutation entropy: the fitness function will minimize the permutation entropy of each IMF component after CEEMDAN decomposition; 2. Minimum envelope entropy: the fitness function will minimize the entropy value of the envelope of each IMF component after CEEMDAN decomposition; 3. Minimum sample entropy: the fitness function will minimize the sample entropy of each IMF component after CEEMDAN decomposition; 4. Minimum information entropy: the fitness function will minimize the information entropy of each IMF component after CEEMDAN decomposition.

[0041] The IAO algorithm is used to optimize the CEEMDAN parameters to minimize the fitness function. The steps of optimizing the CEEMDAN parameters using the IAO algorithm are as follows:

[0042] The Eagle optimization algorithm needs to randomly initialize the population position matrix X:

[0043]

[0044] X ij =rand×(UB j -LB j )+LB j ,i=1,2,…,N,j=1,2,…,Dim

[0045] Where N represents the population size and Dim represents the dimension of the search space.

[0046] The Eagle optimization algorithm is divided into the following stages:

[0047] (1) Expansion phase: In this strategy, the eagle flies from a high altitude and explores the search space extensively. Once the eagle identifies the area where the prey is located, it dives vertically. The mathematical model of this behavior is expressed as:

[0048]

[0049] In the formula, X(t) and X(t+1) represent the individual positions of the AO algorithm in the tth iteration and the t+1th iteration respectively, Xbest(t) represents the best individual obtained by the algorithm up to the tth iteration, and X M (t) represents the average position of the population at the tth iteration, and T represents the maximum number of iterations of the algorithm.

[0050] (2) Reduction phase: The eagle hovers over the target prey, preparing to land, and then descends in the selected area to attack the prey. The mathematical formula is:

[0051] X2(t+1)=X best (t)×Levy(D)+X R (t)+(yx)*rand

[0052] Where Levy(D) represents the Levy flight distribution strategy, which is the expression of the selected fitness function.

[0053] The IAO algorithm carefully searches the area selected by the target prey (function) in preparation for the attack. The mathematical expression of this behavior is:

[0054]

[0055] y=r×cos(θ)

[0056] x=r×sin(θ)

[0057] r=r1+U×D1

[0058] θ=-ω×D1+θ1

[0059]

[0060] Where s and β are constants equal to 0.01 and 1.5 respectively, μ and v are random numbers between 0 and 1, y and x are spiral shape parameters in the search process; D is the problem dimension; r1 has a value between 1 and 20, indicating the number of searches; U = 0.00565; D1 is an integer from 1 to D; ω is 0.005;

[0061] (3) Expansion and development: In the third strategy, when the area where the prey is located is accurately locked, the eagle makes a preliminary attack by vertical descent to detect the prey's reaction. AO develops the designated area to approach and attack the target prey (function). This behavior expression:

[0062] X3(t+1)=(X best (t)-X M (t))×α-rand+

[0063] ((UB-LB)×rand+LB)×δ

[0064] Among them, α and δ are a certain smaller mining adjustment value, the value is 0.1; LB is the lower limit of the problem; UB is the upper limit of the problem

[0065] (4) Narrowing development: When the eagle approaches its prey, it will chase it and attack the prey on the ground. The formula is:

[0066] X4(t+1)=QF×X best(t) - (G1 × X(t) × r6) - G2 × Levy(D) + r7 × G1

[0067]

[0068] G1 = 2 × r8 - 1

[0069] Among them, x(t) is the current position; QF(t) represents the quality function value for balancing the search technology; G1 represents the motion parameter of the eagle during prey monitoring in flight, which is a random number between [-1, 1]; G2 represents the flight slope when following the prey during the journey, and its value decreases from 2 to 0. r6, r7, and r8 are random numbers between 0 and 1.

[0070] The overall process of the IAO (Eagle Optimization Algorithm) is as follows:

[0071] S1: Initialize the algorithm parameters, including the population size, the maximum number of iterations T, and the exploration and exploitation parameters α, δ.

[0072] S2: Initialize the population position X, the initial population fitness, and the best individual.

[0073] S3: When t < T, the AO starts to loop.

[0074] S4: Expand the exploration stage, calculate the average position of the population, and update the population position X1(t + 1).

[0075] S5: Narrow the exploration stage, and update the population position X2(t + 1).

[0076] S6: Expand the exploitation stage, and update the population position X3(t + 1).

[0077] S7: Narrow the exploitation stage, and update the population position X4(t + 1).

[0078] S8: Calculate the fitness of the updated population, and obtain the current best individual position and fitness.

[0079] S9: Compare the current best individual with the fitness of the best individual found in the t-th generation, and retain the better individual position.

[0080] S10: Determine whether the algorithm termination condition is met. If not, repeat steps 4 - 9.

[0081] S11: Jump out of the loop and output the optimal solution and the best fitness.

[0082] After the above process, the convergence curve of the Eagle Optimization Algorithm is as Figure 2As shown in the figure, the best fitness curve has converged, indicating that the best fitness function parameters have been obtained. The best fitness parameters mainly include the noise standard deviation and the number of CEEMDAN realizations required later. These two best parameters are input into CEEMDAN to perform CEEMDAN decomposition on the pulse signal with noise reduction.

[0083] CEEMDAN decomposition is an improvement on the basis of the EMD algorithm. The EMD algorithm has many advantages in non-stationary signal analysis and is suitable for denoising of ultrasonic signals. Its main disadvantage is the mode aliasing effect, which makes some signals obtain invalid IMF components after being decomposed by the EMD algorithm. In order to overcome the mode aliasing effect, Huang proposed the EEMD algorithm, which eliminates the mode aliasing effect by adding white noise with a certain signal-to-noise ratio to the original signal. However, in the denoising process, some of the added white noise still exists and cannot be eliminated, thereby affecting the denoising performance. CEEMDAN also eliminates the mode aliasing effect of the EMD algorithm by adding white noise, but unlike the EEMD algorithm, CEEMDAN adds an IMF component containing auxiliary noise. n(t) is decomposed by EMD instead of directly adding a Gaussian white noise signal, which effectively solves the problem of white noise residue. Therefore, this embodiment adopts the fully adaptive noise set empirical mode decomposition CEEMDAN algorithm.

[0084] The steps of CEEMDAN are as follows:

[0085] S12: Add a standard normally distributed white noise sequence n(t) to the measured human pulse signal x(t) to obtain a corresponding noise signal X(t); then perform EMD decomposition on the noise signal X(t) to obtain a group of component signals, and average the group of component signals to obtain the corresponding first modal component IMF1(n);

[0086]

[0087] Where: IMF1(n) represents the first modal component obtained by CEEMDAN decomposition; N represents the number of modal components obtained by EMD decomposition of the measured pulse signal X(t); Represents the components decomposed by EMD.

[0088] The first residual component is calculated by the following formula:

[0089] r1(n)=x(t)-IMF1(n);

[0090] S13: Add the standard normally distributed white noise sequence n(t) to the first residual component r1(n) to obtain the corresponding noise signal R1(t); then perform EMD decomposition on the noise signal R1(t) to obtain a group of component signals, and average the group of component signals to obtain the second modal component IMF2;

[0091]

[0092] IMF2(n) represents the second modal component obtained by CEEMDAN decomposition; N represents the number of modal components obtained by EMD decomposition of the residual component r1(n); Represents the components decomposed by EMD.

[0093] The second residual component is calculated by the following formula:

[0094] r2(n)=r1(n)-IMF2

[0095] S14: Execute steps S12 to S13 in a loop until the residual component finally obtained does not meet the conditions of EMD decomposition, and then stop decomposing to obtain j IMFs (modal components) and the final residual component R(t), which is described by the formula as follows:

[0096]

[0097] Where j represents the number of modal components decomposed by CEEMDAN.

[0098] According to the above CEEMDAN, 9 IMFs (intrinsic mode components) and a residual signal R(t) are obtained, as shown in Figure 3 As shown in the figure, the time domain information diagram of each IMF is included. For better explanation, the frequency distribution diagram corresponding to each IMF component is added. It can be shown that IMF1 to IMF4 contain a large amount of high-frequency information, and the frequency of human pulse signals is generally between 0.5Hz and 10Hz, so further processing and noise reduction are required.

[0099] After obtaining each IMF component by IAO-CEEMDAN decomposition, the sample entropy value, variance contribution rate and correlation coefficient of each IMF component are calculated. The variance contribution rate describes the contribution rate of each IMF component to the fluctuation of the original sequence, so as to characterize the influence of different period components on the formation of the original data. The calculation formula of the variance contribution rate is:

[0100]

[0101] Among them: SS res Represents the residual sum of squares, SS tot represents the total sum of squares.

[0102] The result range of the correlation coefficient is between [-1,1]. A correlation coefficient of 1 means that the two signals are most similar, and a correlation coefficient of -1 means that the two signals are completely opposite in similarity. As can be seen from the similarity, the larger the correlation coefficient, the better. The calculation formula of the correlation coefficient is:

[0103]

[0104] Where: Cov(X(t),IMF) represents the covariance between X(t) and IMF, σ X(t) is the standard deviation of X(t), σ IMF is the standard deviation of the IMF.

[0105] The distribution diagram of the calculation results of sample entropy value, variance contribution rate and correlation coefficient is as follows: Figure 4 As shown in the figure, the thresholds of sample entropy, variance contribution rate and correlation coefficient are set. The IMF components whose sample entropy is lower than the set threshold and whose variance contribution rate and correlation coefficient are higher than the set threshold are retained, otherwise they are discarded. After experimental verification, the initial denoising effect is best when the sample entropy threshold is 2.50, the variance contribution threshold is 0.05, and the correlation threshold is 0.10. For the IMF components after IAO-CEEMDAN decomposition, the retained modal components are IMF5, IMF6, IMF7, IMF8, and IMF9.

[0106] The retained IMF5, IMF6, IMF7, IMF8, and IMF9 components are subjected to wavelet threshold denoising. The denoising process is as follows: the remaining IMF components are temporarily named as signal x(t). After experimental verification, the wavelet basis sym8 is selected, the number of decomposition layers is 5, the threshold rule adopts the heursure rule, and the threshold function adopts the soft threshold method. First, the signal is subjected to discrete wavelet transform according to the selected rule to obtain a set of wavelet transform coefficients W(a, b) which can be expressed as:

[0107]

[0108] Where: W(a,b) is the wavelet transform coefficient, is the discrete wavelet basis function, and a and b are the scale factor and translation factor; by changing the value (a, b), a series of wavelet coefficients {W(a, b)} are obtained; after obtaining the wavelet coefficients {W(a, b)} at various scales, {W(a, b)} is screened by the threshold function to achieve the denoising effect. The soft threshold function is expressed as:

[0109]

[0110] Among them, the function SIGN refers to the sign function, which is defined as follows: for any real number x, if x>0, then SIGN(x)=1; if x=0, then SIGN(x)=0; if x<0, then SIGN(x)=-1, W′(a,b) is the wavelet coefficient after threshold processing, W(a,b) is the wavelet coefficient after wavelet decomposition, and λ is the threshold; from the above formula, it can be seen that the soft threshold denoising method is to compare |W(a,b)| with λ, the coefficient less than λ is the high-frequency noise coefficient, and the value of W′(a,b) is directly set to zero; for the wavelet coefficient greater than the threshold, when W(a,b) is less than 0, the value of W′(a,b) is the sum of W(a,b) and λ, and when W(a,b) is greater than 0, the value of W′(a,b) is the difference between W(a,b) and λ.

[0111] For the obtained new wavelet coefficients {W′(a,b)}, the noise signal x is reconstructed by using the inverse wavelet transform ′ (t) The inverse wavelet transform formula is shown as follows:

[0112]

[0113] Where x ′ (t) is the reconstructed IMF component signal, represents discrete wavelet basis function, a and b are scale factor and translation factor.

[0114] After performing the above wavelet threshold denoising on IMF5, IMF6, IMF7, IMF8, and IMF9, we can obtain the denoised IMF5′, IMF6′, IMF7′, IMF8′, and IMF9′. Adding these signals to R(t) can obtain the denoised pulse data, which can be expressed as:

[0115] X(t)=IMF5′+IMF6′+IMF7′+IMF8′+IMF9′+R(t)

[0116] Where X(t) represents the denoised signal and R(t) is the monotone indecomposable residual.

[0117] In order to verify the performance of the method, the commonly used denoising methods are compared, and the comparison chart after denoising using the previous CEEMDAN and WT thresholds and the method of the present invention is shown as follows: Figure 5 As shown, it can be seen that the noise reduction effect of the present invention is smoother.

[0118] In order to more intuitively illustrate the superior performance of the method in this paper, more intuitive parameters are used to represent the performance of the algorithm. The parameters of the digital standard are the root mean square error. The smaller the root mean square error, the closer the denoised signal is to the real signal. The calculation formula is:

[0119]

[0120] Where RMSE is the root mean square error, N is the total signal length, t is the time of a single sampling point, f(t) is the original noisy signal value at a certain time t, and f′(t) is the denoised signal value at a certain time t.

[0121]

[0122] Signal-to-noise ratio: In modulated signal transmission, the signal-to-noise ratio generally refers to the ratio of the average power of the carrier signal at the channel output end, that is, the receiver input end, to the average power of the noise in the channel, which can be called the carrier-to-noise ratio. In analog communication systems, the signal-to-noise ratio generally refers to the ratio of the average power of the signal at the output end of the demodulator of the communication terminal to the average power of the noise, indicating that the larger this parameter is, the stronger the algorithm performance is. Its calculation formula is:

[0123]

[0124] Among them, SNR is the signal-to-noise ratio, and t is the time of a single sampling point.

[0125] Root mean square error and signal-to-noise ratio are important parameter standards. The detailed results are shown in Table 1 below:

[0126] Table 1

[0127] Denoising method / parameter index CEEMDAN WT threshold The present invention Root mean square error 0.062496 0.056679 0.037786 Signal-to-Noise Ratio 24.0786 24.9273 28.4491

[0128] This table shows that compared with previous denoising methods, the performance of the algorithm proposed in this paper is more outstanding.

Claims

1. A method for reducing noise of human pulse signal of electronic sphygmomanometer, characterized in that: The following steps are involved: S1. Preprocessing the human pulse signal to be denoised, measured by an electronic sphygmomanometer, and performing IAO-CEEMDAN decomposition on the preprocessed signal to obtain multiple IMF components and a residual component; S2. Set the thresholds of sample entropy, variance contribution rate and correlation coefficient to determine the valid IMF components, retain the valid IMF components and discard the invalid IMF components; S3, using wavelet threshold to denoise the effective IMF components and perform secondary denoising; S4. Superimpose the denoised IMF components and add the residual components to complete the reconstruction of the human pulse signal to obtain the denoised human pulse signal.

2. The method for reducing noise of human pulse signal of electronic sphygmomanometer according to claim 1, characterized in that: The specific process of IAO-CEEMDAN decomposition is as follows: S1.

1. Initialize the parameters of CEEMDAN and IAO; S1.

2. Decompose the human pulse signal using the current CEEMDAN parameters to obtain a set of IMF components; S1.3, select a fitness function, and calculate the fitness value of each IMF component according to the fitness function; S1.

4. Use IAO to optimize CEEMDAN parameters to minimize the fitness value. S1.5, when the fitness value is minimized, update the parameters of CEEMDAN according to the results of IAO optimization; S1.

6. Based on the obtained CEEMDAN parameters, the human pulse signal is decomposed by CEEMDAN to obtain multiple IMF components and residual components. The decomposition formula of CEEMDAN is: Where x(t) represents the original signal measured by the electronic sphygmomanometer; R(t) represents the monotone indecomposable residual component; IMF k represents the k IMF components after decomposition.

3. The method for reducing noise of a human pulse signal of an electronic sphygmomanometer according to claim 2, characterized in that: In step S2, according to the decomposition result of S1, the IMF components whose sample entropy is lower than the set threshold and whose variance contribution rate and correlation coefficient are higher than the set threshold are retained, otherwise they are discarded.

4. The method for reducing noise of a human pulse signal of an electronic sphygmomanometer according to claim 3, characterized in that: The specific process of wavelet threshold denoising is as follows: The effective IMF component is named as signal x(t), the wavelet basis sym8 is selected, the decomposition layer is 5, the threshold rule adopts the heursure rule, and the threshold function adopts the soft threshold method. First, the signal is discrete wavelet transformed according to the selected rule to obtain a set of wavelet transform coefficients W(a,b) expressed as: Where W(a,b) is the wavelet transform coefficient, is the discrete wavelet basis function, and a and b are the scale factor and translation factor; by changing the value (a, b), a series of wavelet coefficients {W(a, b)} are obtained; after obtaining the wavelet coefficients {W(a, b)} at various scales, {W(a, b)} is filtered by the threshold function to achieve the denoising effect. The soft threshold function is expressed as: Among them, W′(a,b) is the wavelet coefficient after threshold processing, W(a,b) is the wavelet coefficient after wavelet decomposition, and λ is the threshold; from the above formula, we can see that the soft threshold denoising method is to compare |W(a,b)| with λ. The coefficient less than λ is the high-frequency noise coefficient, and the W′(a,b) value is directly set to zero; for the wavelet coefficient greater than the threshold, when W(a,b) is less than 0, the value of W′(a,b) is the sum of W(a,b) and λ, and when W(a,b) is greater than 0, the value of W′(a,b) is the difference between W(a,b) and λ; The new wavelet coefficients {W′(a,b)} are used to reconstruct the noise signal x by using the inverse wavelet transform ′ (t) The inverse wavelet transform formula is shown as follows: Where x ′ (t) is the reconstructed IMF component signal, that is, the result after removing the secondary noise. represents the discrete wavelet basis function, a and b are the scale factor and translation factor.

5. The method for reducing noise of human pulse signal of electronic sphygmomanometer according to claim 4, characterized in that: The specific implementation process of step S4 is as follows: According to the denoised IMF components, the pulse signal is reconstructed using the formula: Where, IMF′ n is the IMF component after wavelet threshold denoising; N is the number of valid IMF components, and X(t) is the output pulse signal after denoising.

Citation Information

Cited By

  • Blood pressure measuring method

    CN119791625A