A blood flow signal denoising method based on PSO-EMD optimized wavelet threshold

Abstract

The application provides a blood flow signal denoising method based on PSO-EMD optimized wavelet threshold, which comprises the following steps: collecting photon signals scattered by tissues by using a photoelectric detector, processing the signals by a digital correlator, and obtaining blood flow signals by N-order linear algorithm calculation; decomposing the blood flow signals by EMD to obtain IMF signals; selecting a wavelet threshold denoising algorithm, and dynamically screening the threshold by using PSO, taking the signal-to-noise ratio as a target fitness function, and selecting optimal parameters by considering the target adaptation principle and the effective convergence principle of the PSO algorithm; removing noise from high-frequency IMF components in the IMF signals according to the wavelet threshold denoising algorithm of the optimal parameters, and superimposing and reconstructing the high-frequency IMF components, low-frequency IMF components and residual components to obtain denoised blood flow signals. The application can effectively remove noise and maximize the details of the effective information of the blood flow signals.

Description

Technical Field

[0001] This invention belongs to the field of signal processing technology, specifically relating to a method, apparatus, device, and medium for blood flow signal denoising based on PSO-EMD optimized wavelet threshold. Background Technology

[0002] Blood flow signals play an irreplaceable role in clinical diagnosis, intraoperative monitoring, and physiological research. However, in actual acquisition, blood flow signals are inevitably affected by factors such as high-frequency noise from the detector itself, electromagnetic interference from the acquisition equipment, and physiological activities. This prevents the complete acquisition of effective information from the blood flow signals and its use in analyzing their characteristics, severely impacting the accuracy and reliability of subsequent feature extraction, parameter estimation, and clinical diagnosis.

[0003] In the field of blood flow signal denoising, wavelet thresholding and empirical mode decomposition (EMD) denoising are the two most widely used methods. Wavelet thresholding, with its unique advantage of time-frequency dual-domain analysis, can accurately capture local signal details. Its algorithm is simple to implement and computationally efficient, demonstrating significant value in preserving details and rapidly processing non-stationary signals. EMD denoising, on the other hand, does not require preset basis functions; it decomposes complex signals into a series of intrinsic mode functions, better matching the natural fluctuation characteristics of blood flow signals and achieving preliminary separation of noise and useful components. However, both methods still have unavoidable limitations in blood flow signal processing: wavelet thresholding's threshold and wavelet basis function selection are highly dependent on experience, lacking a unified standard, making it difficult to adapt to the dynamic fluctuation characteristics of blood flow signals with changes in heart rate and respiration, easily leading to residual noise or distortion of instantaneous blood flow characteristics; EMD denoising generally suffers from mode aliasing, easily mixing key instantaneous features of the blood flow signal with high-frequency noise into the same IMF component, while also exhibiting significant endpoint effects and high computational complexity, making it difficult to meet the needs of real-time clinical monitoring.

[0004] In summary, addressing the core shortcoming of the above methods regarding threshold parameter selection, this invention proposes a blood flow signal denoising method based on PSO-EMD optimized wavelet threshold. This method utilizes the EMD algorithm to accurately separate IMFs components of different frequencies in the blood flow signal, and can adaptively and dynamically select the threshold based on the different characteristics of high-frequency IMFs components using the PSO algorithm to perform denoising processing, thereby achieving noise removal and effective information extraction from the blood flow signal. Summary of the Invention

[0005] The main objective of this invention is to provide a method, apparatus, device, and medium for denoising noisy blood flow signals acquired by the DCS method, which can solve the problem of obtaining effective information from blood flow signals containing noise interference.

[0006] To achieve the above objectives, the technical solution provided by the present invention is as follows:

[0007] In a first aspect, the present invention provides a blood flow signal denoising method based on PSO-EMD optimized wavelet threshold, comprising:

[0008] Photons scattered by tissue are collected using a photodetector, processed by a digital correlator, and the blood flow signal is calculated using an Nth-order linear algorithm.

[0009] EMD is used to decompose blood flow signals to obtain IMF signals;

[0010] The wavelet thresholding denoising algorithm is selected, and PSO is used to dynamically filter the threshold. The signal-to-noise ratio is used as the target fitness function, and the optimal parameters are selected by taking into account both the target fitness principle and the convergence and effectiveness principle of the PSO algorithm.

[0011] The high-frequency IMFs components in the IMF signal are removed by a wavelet threshold denoising algorithm with optimal parameters, and then superimposed with the low-frequency IMFs components and residual components to reconstruct the denoised blood flow signal.

[0012] Furthermore, the blood flow signal is decomposed using EMD to obtain the IMF signal, including:

[0013] Find all the maximum and minimum points of the blood flow signal x(t);

[0014] The upper envelope e of the original data is formed by fitting a cubic spline interpolation function. + (t) and lower envelope e - (t);

[0015] Calculate the mean m(t) of the upper and lower envelopes.

[0016] Subtract the mean m(t) from the blood flow signal x(t) to obtain h(t). Determine if h(t) satisfies the IMF condition. If not, treat the obtained h(t) as the original signal and decompose it according to the above steps. Repeat this process K times until there are no negative local maxima or positive local minima, resulting in an IMF component, denoted as n(t), where n(t) = imf(t) = h. K (t), where imf(t) represents the intrinsic mode function and K represents the number of iterations;

[0017] The residual signal r(t) is obtained by subtracting n(t) from the blood flow signal. After obtaining each IMF component, the residual signal is further decomposed using EMD until it no longer contains any IMFs, thus obtaining high-frequency IMFs and low-frequency IMFs. The original signal after EMD decomposition is expressed as follows:

[0018]

[0019] Where x(t) represents the original blood flow signal; n i (t) represents the i-th IMF component; l represents the number of IMF decomposition layers; r(t) is the residual signal.

[0020] Furthermore, a wavelet thresholding denoising algorithm is selected, and PSO is used to dynamically filter the threshold. The signal-to-noise ratio is used as the target fitness function, and the optimal parameters are selected by considering both the target fit principle and the convergence and effectiveness principle of the PSO algorithm. These parameters include:

[0021] The threshold is transformed into a "particle position vector" that PSO can handle. Population size N, maximum number of iterations k, learning factor, and inertia weight ω are set. The learning factor includes cognitive attraction c1 and social attraction c2.

[0022] Under the above conditions, each particle is given an initial velocity v. i and initial position x i It also limits the maximum velocity and position range of the particles, with the absolute value of each particle's position representing a candidate threshold and the velocity representing the amount of change of the threshold each time.

[0023] We choose signal-to-noise ratio as the evaluation metric and define the fitness function.

[0024] Update the local and global optima of the particles. Compare the best position pbest that the particle has experienced at the current fitness (i.e., the individual optimum of each particle) and update pbest. Select the best pbest in the population as the global optimum gbest, also denoted as P. g ;

[0025] The update velocity and position of the particles are calculated, and it is determined whether the particle independent variables exceed the parameter limits, which is the convergence and validity principle of the PSO algorithm.

[0026] Determine if the number of iterations has reached the maximum number of iterations. If the maximum number of iterations has been reached, terminate the check and use the threshold corresponding to the final gbest as the optimal threshold for subsequent noise reduction; otherwise, repeat the above steps.

[0027] Furthermore, the initial velocity of the particle is V. id =(v i1 v i2 , ..., v iD The initial position is X. id =(x i1 x i2 , ..., x iD ), where D is the dimension of the search space, V id Let X represent the set of velocity vectors of particle i from the 1st to the Dth dimension. idLet represent the set of position vectors of particle i from the 1st dimension to the Dth dimension.

[0028] Furthermore, the particle velocity and position update formulas are as follows:

[0029]

[0030] Where r1 and r2 are random numbers in the interval [0, 1], increasing the randomness of the search; Let d be the velocity vector of particle i in the kth iteration; Let be the d-th dimension position vector of particle i in the k-th iteration; Let be the historical optimal position of particle i in the d-th dimension during the k-th iteration, that is, the optimal solution obtained by the i-th particle after the k-th iteration; ω represents the historical best position of the swarm in the d-th dimension during the k-th iteration, i.e., the best solution in the entire particle swarm after the k-th iteration; ω is the inertial weight, representing the influence of the velocity of the previous generation of particles on the velocity of the current generation of particles.

[0031] Furthermore, noise is removed from the high-frequency IMFs components in the IMF signal using a wavelet threshold denoising algorithm with optimal parameters. This removed signal is then superimposed with the low-frequency IMFs components and residual components to reconstruct the denoised blood flow signal, including:

[0032] Wavelet transform is performed on the high-frequency IMFs components to obtain the wavelet coefficients ω. j，k ;

[0033] For the obtained wavelet coefficients ω j，k Thresholding is performed by dynamically selecting the optimal threshold for high-frequency IMFs components, retaining wavelet coefficients above the optimal threshold, thus completing the thresholding process for the wavelet coefficients and obtaining the processed wavelet coefficients ω′. j，k ;

[0034] For wavelet coefficients ω′ j，k Perform inverse wavelet transform to obtain the denoised high-frequency IMFs components;

[0035] The processed high-frequency IMFs components, the directly retained low-frequency IMFs components, and the residual signal are superimposed to reconstruct the denoised signal under the current particle parameters.

[0036] Furthermore, the blood flow signal is obtained by processing with a digital correlator and calculating using an Nth-order linear algorithm, including:

[0037] The input signal of the digital correlator is the photon detected by the photodetector, and the output is the calculated optical field time autocorrelation function g2(τ). The electric field autocorrelation function g1(τ) is calculated according to the Sigert relation, and then the blood flow data is obtained using an Nth-order linear algorithm.

[0038] Secondly, the present invention also provides a blood flow signal denoising device based on PSO-EMD optimized wavelet threshold, comprising:

[0039] The first unit uses a photodetector to collect photon signals scattered by tissue, processes them through a digital correlator, and calculates the blood flow signal using an Nth-order linear algorithm.

[0040] The second unit uses EMD to decompose blood flow signals to obtain IMF signals;

[0041] The third unit is used to select the wavelet threshold denoising algorithm and dynamically filter the threshold using PSO. The signal-to-noise ratio is used as the target fitness function, and the optimal parameters are selected by taking into account both the target fitness principle and the convergence and effectiveness principle of the PSO algorithm.

[0042] The fourth unit is used to remove noise from the high-frequency IMFs components in the IMF signal using a wavelet threshold denoising algorithm with optimal parameters, and then superimposes and reconstructs the denoised blood flow signal by superimposing it with the low-frequency IMFs components and residual components.

[0043] Thirdly, the present invention also provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the processor; wherein the memory stores instructions executable by the processor, the instructions being executed by the processor to enable the processor to perform the method described thereon.

[0044] Fourthly, the present invention also provides a computer-readable storage medium for storing one or more programs, said one or more programs including computer instructions for causing a computer to perform the method.

[0045] Compared with the prior art, the beneficial effects of this invention are:

[0046] 1. The EMD algorithm in this invention effectively improves upon the limitation of wavelet thresholding, which relies on pre-selecting a suitable wavelet basis and appropriate decomposition level rather than the inherent characteristics of the signal. These artificially chosen factors can easily lead to incomplete extraction of effective information from the blood flow signal after decomposition and reconstruction. This invention creatively combines the adaptive decomposition capability of EMD with the fine filtering capability of wavelet thresholding. First, EMD decomposes the signal into a series of IMFs (Integrated Motion Filters). Then, optimal wavelet thresholding is applied differentially based on the frequency characteristics and physical meaning of different high-frequency IMFs. This maximizes the preservation of useful details in the original signal and effectively avoids signal distortion caused by the crude discarding of high-frequency components.

[0047] 2. The PSO algorithm in this invention uses random search technology, which can easily search for the optimal solution or near-optimal solution of the optimization problem, that is, the optimal threshold of the wavelet threshold denoising algorithm in this invention. It globally and adaptively optimizes the threshold parameters and automatically adjusts them to the optimal parameter combination according to the characteristics of different high-frequency IMFs components (such as non-stationarity and signal-to-noise ratio level). It exhibits strong adaptability and robustness, and overcomes the limitations of existing technologies where parameters depend on expert experience. Attached Figure Description

[0048] To more clearly illustrate the embodiments and technical solutions of the present invention, the accompanying drawings used in the description of the embodiments and the technical solutions adopted will be briefly introduced below.

[0049] Figure 1 This is a flowchart illustrating the overall denoising algorithm described in an embodiment of the present invention.

[0050] Figure 2 This is a flowchart illustrating the signal decomposition process described in an embodiment of the present invention.

[0051] Figure 3 This is a flowchart illustrating the wavelet threshold parameter optimization process described in an embodiment of the present invention.

[0052] Figure 4 This is a flowchart of wavelet threshold denoising as described in an embodiment of the present invention;

[0053] Figure 5 The images show a comparison of blood flow signals after using the traditional wavelet threshold denoising algorithm and the PSO-EMD optimized wavelet threshold denoising algorithm, respectively, in embodiments of the present invention. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. These descriptions are only for further illustrating the features and advantages of the present invention, and not for limiting the claims of the present invention. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0055] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0056] Example 1: This example uses measured blood flow signals as an example. Figure 1 The flowchart of a blood flow signal denoising method based on PSO-EMD wavelet threshold optimization in an embodiment of the present invention is described in detail below:

[0057] S1. Use a photodetector to collect photon signals scattered by tissue, process them with a digital correlator, and calculate the blood flow signal using an Nth-order linear algorithm.

[0058] The described Nth-order linear algorithm is a novel algorithm for calculating tissue blood flow values, proposed to address the shortcomings of semi-infinite analytical solutions for obtaining blood flow. It overcomes the inhomogeneity and irregularity of the measured tissue. It does not seek solutions to the diffusion correlation equation. Instead, the algorithm is derived from the integral of the time autocorrelation function g1(τ) on a single photon, as shown in the following formula:

[0059]

[0060] In the above formula, P(s) is the probability distribution of the path length; To reduce the scattering step size, τ is the delay time; k0 is the light wave vector intensity. s represents the total path length traversed by a photon along a given path; <Δr 2 (τ)> represents the mean square displacement of the scattering particles, the form of which depends on the specific flow model used; from DCS theory, <Δr 2 (τ)>=6αD B τ is widely applicable to various tissue models and can effectively simulate experimental data from various tissues or organs (α is the ratio of moving scatterers to the total number of scatterers, αD). B Blood flow index (BFI) in biological tissues.

[0061] Further expansion of g1(τ) using an Nth-order Taylor series yields:

[0062]

[0063] Further differentiation of g1(τ) by the Nth order yields:

[0064]

[0065] In the above formula This represents the photon propagation path information within the tissue. When τ is sufficiently small, the linear expressions for the first order (N=1) and Nth order (N>1) are:

[0066]

[0067] make but H·αD B (1) The slope Sl between τ and the autocorrelation function g1(τ)-1 can be written as H·αD B (1)=Sl.

[0068] Furthermore, this can be generalized to order N (N>1) to obtain H·αD B (N) =Sl (N) N≥2. Determining the slope through iterative linear regression is a key step in extracting blood flow values ​​from autocorrelation data.

[0069] S2. Use EMD to decompose the blood flow signal to obtain the IMF signal;

[0070] Specific steps are as follows: Figure 2 As shown:

[0071] S2.1 Find all the maximum and minimum points of the blood flow signal x(t);

[0072] S2.2. Use cubic spline interpolation function to fit and form the upper envelope e of the original data. + (t) and lower envelope e - (t);

[0073] S2.3 Calculate the mean value m(t) of the upper and lower envelopes.

[0074] S2.4. Subtract the mean m(t) from the blood flow signal x(t) to obtain h(t). Determine whether h(t) satisfies the IMF condition. If not, treat the obtained h(t) as the original signal and decompose it according to the above steps. Repeat this process K times until there are no negative local maxima or positive local minima. This yields an IMF component, denoted as n(t), where n(t) = imf(t) = h. K (t), where imf(t) represents the intrinsic mode function and K represents the number of iterations;

[0075] S2.5. Subtract n(t) from the blood flow signal to obtain the residual signal r(t). After obtaining each IMF component, continue to perform EMD decomposition on the residual signal until the residual signal no longer contains any IMFs, to obtain the high-frequency IMFs components and low-frequency [MFs components]. The original signal after EMD decomposition is expressed as follows:

[0076]

[0077] Where x(t) represents the original blood flow signal; n i (t) represents the i-th IMF component; l represents the number of IMF decomposition layers; r(t) is the residual signal.

[0078] S3. Select the wavelet threshold denoising algorithm and use PSO to dynamically filter the threshold. Use the signal-to-noise ratio as the target fitness function and select the optimal parameters while taking into account the target fitness principle and the convergence and effectiveness principle.

[0079] Specific steps are as follows: Figure 3 As shown:

[0080] S3.1. Convert the threshold into a "particle position vector" that PSO can process, and set the population size N=30, the maximum number of iterations k=50, the inertia weight ω=0.9, the cognitive attraction c1=1.5, and the social attraction c2=1.5;

[0081] S3.2. Under the condition of S3.1, each particle is given an initial velocity V. id =(v i1 v i2 , ..., v iD ) and initial position X id =(x i1 x i2 , ..., x iD D represents the dimension of the search space, which limits the maximum velocity and position range of the particles. The absolute value of each particle's position represents a candidate threshold, and the velocity represents the amount of change in the threshold each time. V id Let X represent the set of velocity vectors of particle i from the 1st to the Dth dimension. id This represents the set of position vectors of particle i from the 1st dimension to the Dth dimension;

[0082] S3.3 Select signal-to-noise ratio as the evaluation index and define the fitness function;

[0083] S3.4 Update the local and global optima of the particles. Compare the current fitness with the best position the particle has ever reached, pbest (i.e., the individual optimum of each particle), and update pbest. Select the best pbest in the population as the global optimum gbest, also denoted as P. g ;

[0084] S3.5 Calculate the particle update velocity and position and determine whether the particle independent variables exceed the parameter limits, i.e., the convergence and validity principle of the PSO algorithm. The particle velocity and position update formulas are as follows:

[0085]

[0086] Where r1 and r2 are random numbers in the interval [0, 1], increasing the randomness of the search; Let d be the velocity vector of particle i in the kth iteration; Let be the d-th dimension position vector of particle i in the k-th iteration; Let be the historical optimal position of particle i in the d-th dimension during the k-th iteration, that is, the optimal solution obtained by the i-th particle after the k-th iteration; ω represents the historical best position of the swarm in the d-th dimension during the k-th iteration, i.e., the best solution in the entire particle swarm after the k-th iteration; ω is the inertial weight, representing the influence of the velocity of the previous generation of particles on the velocity of the current generation of particles.

[0087] S3.6 Determine if the number of iterations has reached the maximum number of iterations. If the maximum number of iterations has been reached, terminate the check and use the threshold corresponding to the final gbest as the optimal threshold for subsequent noise reduction; otherwise, repeat S3.2-S3.5.

[0088] S4. The high-frequency IMFs components in the IMF signal are removed by the wavelet threshold denoising algorithm with the optimal parameters, and then superimposed with the low-frequency IMFs components and the residual components to reconstruct the denoised blood flow signal.

[0089] Specific steps are as follows: Figure 4 As shown:

[0090] S4.1 Perform wavelet transform on the high-frequency IMFs components to obtain the wavelet coefficients ω. j，k The wavelet transform formula is:

[0091]

[0092] Where a is the scale parameter, b is the displacement parameter, This is an energy normalization factor to ensure consistent wavelet energy across different scales, guaranteeing fair comparison of local signal features and ψ at each scale. a，b (t) is the mother wavelet function;

[0093] S4.2, The wavelet coefficients ω obtained in S4.1 j，k Thresholding is performed by dynamically selecting the optimal threshold for high-frequency IMFs components, retaining wavelet coefficients above the optimal threshold, thus completing the thresholding process for the wavelet coefficients and obtaining the processed wavelet coefficients ω′. j，k ;

[0094] S4.3, Regarding the wavelet coefficients ω′ j，k Performing an inverse wavelet transform yields the denoised high-frequency IMFs components. The inverse wavelet transform formula is:

[0095]

[0096] in, The constant of the wavelet function is...

[0097] S4.4. The processed high-frequency IMFs components, the directly retained low-frequency IMFs components, and the residual signal are superimposed to reconstruct the denoised signal under the current particle parameters.

[0098] The denoising results of the blood flow signal are referenced. Figure 5 As shown. From Figure 5 As can be seen, noise interference causes many spikes in the acquired blood flow signal, resulting in an uneven waveform. While the waveform obtained using traditional wavelet threshold denoising methods generally matches the original signal in outline, significant deviations occur at signal abrupt changes, and high-frequency information is not effectively preserved, leading to the loss of some effective information in the blood flow signal. The waveform obtained using the method proposed in this invention is smoother, with an outline highly consistent with the original signal, especially showing superior performance at abrupt changes and local variations. Using the signal-to-noise ratio (SNR) as an evaluation metric, a higher SNR generally indicates better denoising performance. In this embodiment, the SNR of the blood flow signal after wavelet denoising is 7.56 dB, and the SNR after PSO-EMD optimized wavelet denoising is 9.16 dB, indicating that the denoising effect of the method proposed in this invention is superior to that of wavelet denoising. Further verification shows that the method of this invention not only effectively suppresses most noise but also reconstructs weak high-frequency component signals relatively well.

[0099] Example 2: While Example 1 provided a blood flow signal denoising method based on PSO-EMD optimized wavelet thresholding, this example provides a blood flow signal denoising device based on PSO-EMD optimized wavelet thresholding, comprising:

[0100] The first unit uses a photodetector to collect photon signals scattered by tissue, processes them through a digital correlator, and calculates the blood flow signal using an Nth-order linear algorithm.

[0101] The second unit uses EMD to decompose blood flow signals to obtain IMF signals;

[0102] The third unit is used to select the wavelet threshold denoising algorithm and dynamically filter the threshold using PSO. The signal-to-noise ratio is used as the target fitness function, and the optimal parameters are selected by taking into account both the target fitness principle and the convergence and effectiveness principle of the PSO algorithm.

[0103] The fourth unit is used to remove noise from the high-frequency IMFs components in the IMF signal using a wavelet threshold denoising algorithm with optimal parameters, and then superimposes and reconstructs the denoised blood flow signal by superimposing it with the low-frequency IMFs components and residual components.

[0104] Example 3: This example provides an electronic device corresponding to the blood flow signal denoising method based on PSO-EMD optimized wavelet threshold provided in Example 1. The electronic device can be an electronic device for the client, such as a mobile phone, laptop, tablet computer, desktop computer, etc., to execute the method of Example 1.

[0105] Example 4: This example provides a computer-readable storage medium for storing one or more programs, the one or more programs including computer instructions, which, when executed by a computer, cause the computer to perform the method provided in Example 1 above.

[0106] The above description only details the embodiments of the present invention. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention, and all such changes should be included within the protection scope of the present invention.

Claims

1. A blood flow signal denoising method based on PSO-EMD optimized wavelet threshold, characterized in that, include: Photons scattered by tissue are collected using a photodetector, processed by a digital correlator, and the blood flow signal is calculated using an Nth-order linear algorithm. EMD is used to decompose blood flow signals to obtain IMF signals; The wavelet thresholding denoising algorithm is selected, and PSO is used to dynamically filter the threshold. The signal-to-noise ratio is used as the target fitness function, and the optimal parameters are selected by taking into account both the target fitness principle and the convergence and effectiveness principle of the PSO algorithm. The high-frequency IMFs components in the IMF signal are removed by a wavelet threshold denoising algorithm with optimal parameters, and then superimposed with the low-frequency IMFs components and residual components to reconstruct the denoised blood flow signal.

2. The blood flow signal denoising method based on PSO-EMD optimized wavelet threshold as described in claim 1, characterized in that, EMD is used to decompose blood flow signals to obtain IMF signals, including: Find all the maximum and minimum points of the blood flow signal x(t); The upper envelope e of the original data is formed by fitting a cubic spline interpolation function. + (t) and lower envelope e - (t); Calculate the mean m(t) of the upper and lower envelopes. Subtract the mean m(t) from the blood flow signal x(t) to obtain h(t). Determine if h(t) satisfies the IMF condition. If not, treat the obtained h(t) as the original signal and decompose it according to the above steps. Repeat this process K times until there are no negative local maxima or positive local minima, resulting in an IMF component, denoted as n(t), where n(t) = imf(t) = h. K (t), where imf(t) represents the intrinsic mode function and K represents the number of iterations; The residual signal r(t) is obtained by subtracting n(t) from the blood flow signal. After obtaining each IMF component, the residual signal is further decomposed using EMD until it no longer contains any IMFs, thus obtaining high-frequency IMFs and low-frequency IMFs. The original signal after EMD decomposition is expressed as follows: Where x(t) represents the original blood flow signal; n i (t) represents the i-th IMF component; l represents the number of IMF decomposition layers; r(t) is the residual signal.

3. The blood flow signal denoising method based on PSO-EMD optimized wavelet threshold as described in claim 1, characterized in that, Wavelet thresholding denoising algorithm is selected, and PSO is used to dynamically filter the threshold. The signal-to-noise ratio is used as the target fitness function, and the optimal parameters are selected by considering both the target fit principle and the convergence and effectiveness principle of the PSO algorithm. These parameters include: The threshold is transformed into a "particle position vector" that can be handled by PSO. The population size N, the maximum number of iterations k, the learning factor, and the inertia weight ω are set. The learning factor includes cognitive attraction c1 and social attraction c2. Under the above conditions, each particle is given an initial velocity v. i and initial position x i It also limits the maximum velocity and position range of the particles, with the absolute value of each particle's position representing a candidate threshold and the velocity representing the amount of change of the threshold each time. We choose signal-to-noise ratio as the evaluation metric and define the fitness function. Update the local and global optima of the particles. Compare the best position pbest that the particle has experienced at the current fitness (i.e., the individual optimum of each particle) and update pbest. Select the best pbest in the population as the global optimum gbest, also denoted as P. g ; The update velocity and position of the particles are calculated, and it is determined whether the particle independent variables exceed the parameter limits, which is the convergence and validity principle of the PSO algorithm. Determine if the number of iterations has reached the maximum number of iterations. If the maximum number of iterations has been reached, terminate the check and use the threshold corresponding to the final gbest as the optimal threshold for subsequent noise reduction; otherwise, repeat the above steps.

4. The blood flow signal denoising method based on PSO-EMD optimized wavelet threshold as described in claim 3, characterized in that, The initial velocity of the particle is V id =(v i1 v i2 , ..., v iD The initial position is X. id =(x i1 x i2 , ..., x iD ), where D is the dimension of the search space, V id Let X represent the set of velocity vectors of particle i from the 1st to the Dth dimension. id Let represent the set of position vectors of particle i from the 1st dimension to the Dth dimension.

5. A blood flow signal denoising method based on PSO-EMD optimized wavelet thresholding according to claim 4, characterized in that, The particle velocity and position update formulas are as follows: Where r1 and r2 are random numbers in the interval [0, 1], increasing the randomness of the search; Let d be the velocity vector of particle i in the kth iteration; Let be the d-th dimension position vector of particle i in the k-th iteration; Let be the historical optimal position of particle i in the d-th dimension during the k-th iteration, that is, the optimal solution obtained by the i-th particle after the k-th iteration; ω represents the historical best position of the swarm in the d-th dimension during the k-th iteration, i.e., the best solution in the entire particle swarm after the k-th iteration; ω is the inertial weight, representing the influence of the velocity of the previous generation of particles on the velocity of the current generation of particles.

6. The blood flow signal denoising method based on PSO-EMD optimized wavelet threshold as described in claim 1, characterized in that: The high-frequency IMFs components in the IMF signal are removed using a wavelet threshold denoising algorithm with optimal parameters. These removed components are then superimposed with the low-frequency IMFs components and residual components to reconstruct the denoised blood flow signal, which includes: Wavelet transform is performed on the high-frequency IMFs components to obtain the wavelet coefficients ω. j，k ; For the obtained wavelet coefficients ω j，k Thresholding is performed by dynamically selecting the optimal threshold for high-frequency IMFs components, retaining wavelet coefficients above the optimal threshold, thus completing the thresholding process for the wavelet coefficients and obtaining the processed wavelet coefficients ω′. j，k ; For wavelet coefficients ω′ j，k Perform inverse wavelet transform to obtain the denoised high-frequency IMFs components; The processed high-frequency IMFs components, the directly retained low-frequency IMFs components, and the residual signal are superimposed to reconstruct the denoised signal under the current particle parameters.

7. A blood flow signal denoising method based on PSO-EMD optimized wavelet thresholding according to claim 1, characterized in that: The blood flow signal is obtained by processing with a digital correlator and calculating using an Nth-order linear algorithm, including: The input signal of the digital correlator is the photon detected by the photodetector, and the output is the calculated optical field time autocorrelation function g2(τ). The electric field autocorrelation function g1(τ) is calculated according to the Sigert relation, and then the blood flow data is obtained using an Nth-order linear algorithm.

8. A blood flow signal denoising device based on PSO-EMD optimized wavelet threshold, characterized in that, The first unit uses a photodetector to collect photon signals scattered by tissue, processes them through a digital correlator, and calculates the blood flow signal using an Nth-order linear algorithm. The second unit uses EMD to decompose blood flow signals to obtain IMF signals; The third unit is used to select the wavelet threshold denoising algorithm and dynamically filter the threshold using PSO. The signal-to-noise ratio is used as the target fitness function, and the optimal parameters are selected by taking into account both the target fitness principle and the convergence and effectiveness principle of the PSO algorithm. The fourth unit is used to remove noise from the high-frequency IMFs components in the IMF signal using a wavelet threshold denoising algorithm with optimal parameters, and then superimposes and reconstructs the denoised blood flow signal by superimposing it with the low-frequency IMFs components and residual components.

9. An electronic device, characterized in that, include: At least one processor; And a memory communicatively connected to the processor; wherein the memory stores instructions executable by the processor, the instructions being executed by the processor to enable the processor to perform the method according to any one of claims 1-7.

10. A computer-readable storage medium for storing one or more programs, characterized in that, The one or more programs include computer instructions for causing a computer to perform the method according to any one of claims 1-7.

Images