OFDM radar vital sign signal separation method based on FGO-VMD and multi-feature clustering

Abstract

The invention discloses an OFDM radar vital sign signal separation method based on FGO-VMD and multi-feature clustering, and belongs to the technical field of radar communication integration. According to the method, an OFDM signal radar is used for carrying out human body detection, multi-subcarrier phase difference linear regression is used for estimating displacement, thoracic cavity micro-motion signals are extracted, an FGO-VMD joint optimization model is provided, the modal number and penalty factor parameters of a VMD decomposition algorithm are adaptively determined through minimum envelope entropy, and a multi-feature clustering and weighted reconstruction strategy is designed. And in combination with frequency band screening and correlation coefficient weighting, the separation robustness of the respiratory signal and the heartbeat signal is enhanced, and high-precision separation of the vital sign signal is realized. According to the method, non-contact vital sign detection based on OFDM radar signals is achieved, the problems that in a traditional algorithm, parameters of modal numbers and penalty factors need to be manually set, and signal amplitude attenuation and modal aliasing are caused under complex noise are solved, the vital sign monitoring precision is remarkably improved, and high-precision separation of vital sign signals is achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of radar communication integration, and particularly relates to an OFDM radar vital sign signal separation method based on FGO-VMD and multi-feature clustering. Background Art

[0002] With the development of integrated radar and communication technology, non-contact vital sign monitoring based on OFDM signal radar has gradually become a research hotspot. Existing vital sign monitoring methods using contact sensors suffer from poor comfort and limited application scenarios. While non-contact radar-based detection technology can penetrate clothing and bedding for continuous monitoring, it still faces challenges in separating and extracting weak physiological signals (such as heartbeat and respiratory signals) in complex noisy environments.

[0003] In the prior art, variational mode decomposition (VMD) is used for micro-motion signal separation, but its performance is highly dependent on the preset number of modes and penalty factors. The number of modes and penalty factors need to be adjusted based on manual experience, but the decomposition efficiency is low and the adaptability is poor, which leads to problems such as modal aliasing and signal distortion in noise interference or signal aliasing scenarios, affecting the separation accuracy of respiratory and heartbeat signals. The existing phase unwrapping algorithm is susceptible to interference in a noisy environment, resulting in the accumulation of displacement estimation errors. Although existing optimization algorithms (such as genetic algorithms and particle swarm algorithms) can assist in parameter optimization, they have problems such as slow convergence speed and high computational complexity, making it difficult to meet real-time requirements. At the same time, the existing signal reconstruction strategy lacks robustness in the screening of multi-modal components, resulting in a significant decrease in signal separation performance in low signal-to-noise ratio scenarios. Summary of the Invention

[0004] The purpose of the present invention is to overcome the above-mentioned defects of the prior art and provide an OFDM radar vital sign signal separation method based on FGO-VMD and multi-feature clustering.

[0005] The technical problem proposed by the present invention is solved as follows:

[0006] An OFDM radar vital sign signal separation method based on FGO-VMD and multi-feature clustering includes the following steps:

[0007] Step 1: Mix the transmission signal carrying the communication information with the transmission carrier frequency to obtain the radar communication integrated transmission signal S after up-conversion processing. TX (t); After the radar communication integrated transmission signal detects the human chest cavity, the radar communication integrated echo signal S carrying vital signs information is obtained through down-conversion processing. RX (t);

[0008] Step 2: radar communication integrated echo signal S RX(t) After converting to N parallel data streams, remove the cyclic prefix operation, and then perform FFT transformation to obtain the frequency domain information Y[m,n] of the nth subcarrier echo signal on the mth symbol; the radar communication integrated transmission signal S TX (t) Perform an N-point FFT transform to obtain the frequency domain information X[m,n] of the n-th subcarrier transmitted signal on the m-th symbol, and calculate the phase information of the n-th subcarrier containing vital signs information on the m-th symbol Extract H m，n The phase of the nth subcarrier echo signal on the mth symbol is obtained by m，n ; Average all symbols on each subcarrier to obtain the average phase difference Δφ n , extract the phase difference sequence φ after unwrapping n ;

[0009] Step 3: Fit the phase difference sequence φ by weighted least squares method n With frequency f n The linear relationship between them is used to extract the estimated displacement value of the body; the estimated displacement value of the body is subtracted from the distance between the human body and the radar to obtain the vital sign related signal f(t);

[0010] Step 4: FGO-VMD parameter optimization

[0011] Perform FGO-VMD decomposition on the vital sign-related signal f(t) obtained in step 3, use the minimum envelope entropy as the fitness function, and use the fungal growth optimization FGO algorithm to obtain the optimal VMD decomposition parameters;

[0012] Step 5: Extract the global optimal solution obtained by the fungal growth optimization FGO algorithm in step 4 as the optimal VMD decomposition parameter, and perform VMD decomposition on the vital sign related signal f(t) to obtain An optimal narrowband mode function Calculate each narrowband mode function The spectrum peak Calculate each narrowband mode function and the Pearson correlation coefficient of the vital sign related signal f(t)

[0013] right and After normalization, the narrowband modal functions of the vital sign-related signals f(t) are used as clustering features. Based on the k-means clustering method, the narrowband modal functions of the vital sign-related signals f(t) are divided into modal combinations of respiratory signals and modal combinations of heartbeat signals. The clustering results are used to reconstruct the respiratory and heartbeat signals.

[0014] Step 6: Reconstruct the respiratory signal s obtained in step 5b (t) and heartbeat signal s h (t) Perform FFT transformation to obtain the spectrum of the extracted respiratory signal and heartbeat signal, extract the peak of the spectrum to obtain the extracted respiratory frequency and heartbeat frequency, and then complete the visualization output.

[0015] Furthermore, in step 1, the radar communication integrated echo signal S RX (t) is expressed as:

[0016]

[0017] Where, 0≤n≤N-1, N is the number of subcarriers of the transmission carrier frequency, 0≤m≤M-1, M is the total number of OFDM signal symbols, A(n,m) represents the communication information modulated by the mth symbol on the nth subcarrier, j is the imaginary symbol, Δf is the subcarrier spacing, t is the time, τ is the delay, f c is the carrier frequency, noise is the noise generated by the radar communication channel;

[0018] The time delay τ is expressed as:

[0019]

[0020] Where c0 is the speed of light, R is the initial distance from the human chest to the radar, d(t) is the tiny displacement of the chest caused by breathing and heartbeat, and A b is the amplitude of chest movement caused by breathing, f b is the respiratory rate, A h is the amplitude of chest movement caused by heartbeat, f h is the heart rate, is the phase difference caused by heartbeat and breathing.

[0021] Furthermore, the specific process of step 3 is:

[0022] Phase difference sequence φ n With frequency f n There is a linear relationship between:

[0023]

[0024] in, is the rate of change of the body's tiny displacement, is the estimated displacement of the body, f n is the center frequency of the nth subcarrier, f n =f c +nΔf;

[0025] The linear relationship between phase difference and frequency is fitted by weighted least square method. and Solve:

[0026]

[0027] According to the phase difference formula Comparative linear relationship The estimated displacement of the body is for:

[0028]

[0029] Will Subtract R to obtain the vital sign related signal f(t).

[0030] Furthermore, in step 4, the specific steps of using the fungal growth optimization FGO algorithm to obtain the optimal VMD decomposition parameters are as follows:

[0031] Step 4-1. Initialize parameters

[0032] Set the range of the number of modal components K and the quadratic penalty factor α of the VMD decomposition parameters, and set the parameters of the FGO algorithm: the trade-off coefficient EM of exploration and development, the probability of environmental effect Ep, the convergence rate ER, the size of the bacterial colony P, and the maximum number of iterations T max , the upper and lower bounds of the VMD decomposition parameter penalty factor α [lb1, ub1] and the upper and lower bounds of the number of decomposition modal components K [lb2, ub2], initialize the bacterial population;

[0033] Initialize a bacterial colony containing P hyphae individuals, each hyphae individual represents a set of VMD parameters α and K; the i-th hyphae individual contains a quadratic penalty factor parameter S i，1 and decomposed modal component parameters S i，2 , specifically expressed as:

[0034] S i，j =lb j +rand(0,1)×(ub j -lb j )

[0035] Where, 1≤i≤P, j=1,2; rand(0,1) represents a random number in the interval [0,1];

[0036] Step 4-2, fitness evaluation

[0037] For each mycelium individual, the corresponding parameter α = S i，1 and K = S i，2 , substitute it into the VMD algorithm for signal decomposition, and decompose the vital sign related signal f(t) into K narrowband modal functions u with different frequencies k (t), calculate the minimum envelope entropy fitness function under each mode;

[0038] Step 4-3: Update the global optimal solution

[0039] The minimum envelope entropy fitness function F is used as the fitness value, and the mycelium individual with the minimum fitness value is found, and its corresponding decomposition parameter is As the best decomposition coefficient of the current VMD, as the latest global optimal solution;

[0040] Step 4-4, iterative update

[0041] In each iteration, nutrient allocation and mycelium individual update are performed; according to the current number of iterations T and the maximum number of iterations T max , calculate the nutritional value of each mycelium individual, which is used to control the growth behavior of the mycelium individual; after the mycelium individual is updated, the iteration counter is increased by 1, T = T + 1, and returns to step 4-2 until T = T max , output the global optimal solution S best ;

[0042] Furthermore, the specific process of step 4-2 is as follows:

[0043] Decompose the vital sign related signal f(t) into K signals with different frequencies ω k The narrowband mode function u k The process of (t) is expressed as:

[0044]

[0045] Where, 1≤k≤K, K is the number of narrowband modes, u k (t) is the kth narrowband mode function, ω k represents the kth mode center frequency, δ(t) is the Dirac function, represents the partial derivative with respect to time t, || ||2 represents the second norm, and * represents convolution;

[0046] By introducing the Lagrange multiplier λ(t) and the quadratic penalty term α, the above constrained problem is transformed into the following unconstrained variational problem:

[0047]

[0048] Among them, L({u k},{ω k}, λ(t)) represents the augmented Lagrangian function, and <> represents the inner product;

[0049] The above unconstrained variational problem is solved iteratively by the alternating direction multiplier method to obtain K with different center frequencies ω k The narrowband modal function u k (t);

[0050] The k-th narrowband mode function u obtained by VMD decomposition k The envelope entropy H of (t) k for:

[0051]

[0052] Among them, p k (t) is the normalized envelope signal, satisfying x k (t) is the kth narrowband modal function u k (t) Hilbert transform envelope signal;

[0053] The minimum envelope entropy fitness function F is defined as the minimum value of all modal envelope entropies, that is:

[0054]

[0055] Furthermore, the process of step 4-4 is:

[0056] First, distribute nutrients according to the stage: in the early stage of iteration Nutrient allocation is used to explore the search space; in the later stages of the iteration, Nutrient allocation is used to develop the search space;

[0057] Nutrients assigned to the i-th mycelium individual i for:

[0058]

[0059] Among them, F i is the fitness value of the ith mycelium individual, rand(P) is a random number in the interval [0, 1];

[0060] Nutrients i Normalized, the normalized nutrients are the nutrients assigned to the i-th mycelium individual i Expressed as:

[0061]

[0062] Among them, r8 is a random number in the interval [0, 1];

[0063] Then iterate and update:

[0064] For each mycelium individual S i，j Calculate energy E r and probability p i :

[0065]

[0066]

[0067] Among them, minF and maxF are the minimum and maximum fitness values of each hyphae individual in the current iteration, and ε is a very small positive number used to avoid the denominator being zero;

[0068] If p i <E r , then enter the exploration phase and update the mycelium individuals according to the following formula:

[0069]

[0070] E=e fit

[0071]

[0072] in, and represents the i-th mycelium individual in the T+1th iteration and the Tth iteration respectively; U 1，j is a random binary vector, the elements of which are 0 or 1, controlling the update position; S a，j and S b，j are the values of the j-th dimension of two different mycelium individuals selected randomly;

[0073] If p i ≥E r , then proceed to the development phase and update the mycelium individuals according to the following formula:

[0074]

[0075] E c =(r1-0.5)×r2×(S a，j -S b，j )

[0076]

[0077] Among them, D i，j is the update direction of the i-th mycelium individual; E c is a random perturbation term that increases exploration diversity; Ep is the probability of environmental effect, 0<Ep<1, which determines whether to add perturbation; r1, r2, r3, r4, r5, r6, and r7 are all random numbers uniformly distributed in the range [0, 1];

[0078] If the updated The fitness is less than that before the update The fitness value of

[0079] Furthermore, in step 5, based on the k-means clustering method, the narrowband modal function of the vital sign related signal f(t) is divided into the modal combination of the respiratory signal and the modal combination of the heartbeat signal. The specific process of reconstructing the respiratory signal and the heartbeat signal using the clustering results is as follows:

[0080] Step 5-1, initialize cluster center

[0081] Randomly select 2 of the K cluster features as the initial cluster centers, which are respectively recorded as cluster centers C1: C1(μ f1 , μ ρ1 ) and cluster center C2: C2(μ f2 , μ ρ2 ); μ f1 and μ ρ1 Represent the spectrum peak and Pearson correlation coefficient of the first randomly selected cluster feature, μ f2 and μ ρ2 represent the spectrum peak and Pearson correlation coefficient of the second randomly selected cluster feature, respectively;

[0082] Step 5-2: Assign clustering features to cluster centers

[0083] For each narrowband mode function Calculate the Euclidean distance between its clustering features and the two cluster centers and assign it to the cluster with the closest distance, that is:

[0084]

[0085] in, is a narrowband mode function The distance from the cluster center C1, is a narrowband mode function The distance from the cluster center C2;

[0086] Step 5-3, Update cluster center

[0087] For two clusters, the mean of all cluster features in the cluster is used as the cluster center of the current cluster;

[0088] Step 5-4: Repeat steps 5-2 and 5-3 until the cluster center no longer changes or all cluster features are traversed, thereby dividing the narrowband modal function of the vital sign-related signal f(t) into a modal combination of the respiratory signal and a modal combination of the heartbeat signal;

[0089] Step 5-5: For the two obtained cluster modal combinations, the one with a lower cluster center frequency is selected as the modal combination of the respiratory signal, and the one with a higher cluster center frequency is selected as the modal combination of the heartbeat signal;

[0090] The narrowband modal functions of the clustering results of the heartbeat signal and respiratory signal are screened, and the respiratory signal modal combination within 0.1Hz to 0.5Hz is retained as the screened respiratory signal modal combination S b , retain the modal combination of heartbeat signals from 0.8Hz to 2.5Hz as the filtered heartbeat signal modal combination S h ;

[0091] The weight distribution is performed based on the Pearson correlation coefficient between each narrowband modal function and its vital sign related signal f(t):

[0092] Respiration signal weight Expressed as:

[0093]

[0094] Among them, k b is the filtered respiratory signal modal combination S b Each narrowband mode function of is the corresponding filtered respiratory signal modal combination S b Pearson correlation coefficient for each modality in ;

[0095] Heartbeat signal weight Expressed as:

[0096]

[0097] Among them, k h is the filtered heartbeat signal modal combination S b Each narrowband mode function of is the corresponding filtered heartbeat signal modal combination S b Pearson correlation coefficient for each modality in ;

[0098] Reconstruct the respiratory signal and heartbeat signal according to their weights and corresponding narrowband modal functions to obtain the reconstructed respiratory signal s b (t) and heartbeat signal s h (t):

[0099]

[0100] The beneficial effects of the present invention are:

[0101] The method described in the present invention adopts a multi-subcarrier phase difference linear regression model to estimate chest displacement, and combines frequency band energy screening with correlation coefficient weighted reconstruction strategy to effectively suppress the interference of complex environmental noise (such as body movement, electromagnetic interference) on micro-motion signals and reduce the accumulation of displacement estimation errors. The modal number and penalty factor of VMD are dynamically optimized by the FGO algorithm, which avoids the modal aliasing and signal distortion problems caused by the traditional VMD relying on manual experience to adjust parameters, significantly improves the separation accuracy of respiratory and heartbeat signals, and can still accurately extract the frequency characteristics of weak physiological signals in low signal-to-noise ratio scenarios. Compared with traditional genetic algorithms and particle swarm algorithms, the fungal growth optimization algorithm (FGO) has faster convergence speed and lower computational complexity, can quickly complete VMD parameter optimization and signal decomposition, meet the real-time requirements of vital signs monitoring systems, and is suitable for continuous dynamic monitoring scenarios. Through multi-feature clustering, the decomposed modal components are intelligently screened. Combined with frequency band screening and correlation coefficient weighting strategies, the breathing, heartbeat and noise components are adaptively distinguished to avoid invalid modal interference and ensure the authenticity and integrity of the reconstructed signal. This solves the problems of poor comfort and limited applicable scenarios of traditional contact sensors. It is suitable for complex scenarios such as postoperative monitoring, sleep monitoring, disaster relief, etc., and has broad practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0102] Figure 1 Schematic diagram of the process of the present invention;

[0103] Figure 2 Flowchart of optimizing VMD parameters using FGO algorithm in the method of the present invention;

[0104] Figure 3 This is a flow chart of classifying respiratory and heartbeat signals using a clustering algorithm in the method of the present invention;

[0105] Figure 4 This is a diagram of displacement estimation results in the method described in the embodiment;

[0106] Figure 5 A comparison diagram of the reconstructed respiratory signal and heartbeat signal in the method described in the embodiment and the actual situation;

[0107] Figure 6 Spectrum diagram of the respiratory signal and heartbeat signal extracted by the method described in the embodiment. DETAILED DESCRIPTION

[0108] The present invention will be further described below with reference to the accompanying drawings and examples.

[0109] This embodiment provides an OFDM radar vital sign signal separation method based on FGO-VMD and multi-feature clustering, the flow chart of which is as follows: Figure 1 As shown, the following steps are included:

[0110] Step 1: Mix the transmission signal carrying the communication information with the transmission carrier frequency to obtain the radar communication integrated transmission signal S after up-conversion processing. TX (t); After the radar communication integrated transmission signal detects the human chest cavity, the radar communication integrated echo signal S carrying vital signs information is obtained through down-conversion processing. RX (t);

[0111] Radar communication integrated echo signal S RX (t) is expressed as:

[0112]

[0113] Where, 0≤n≤N-1, N is the number of subcarriers of the transmission carrier frequency, 0≤m≤M-1, M is the total number of OFDM signal symbols, A(n,m) represents the communication information modulated by the mth symbol on the nth subcarrier, j is the imaginary symbol, Δf is the subcarrier spacing, t is the time, τ is the delay, f c is the carrier frequency, noise is the noise generated by the radar communication channel;

[0114] The time delay τ is expressed as:

[0115]

[0116] Where c0 is the speed of light, R is the initial distance from the human chest to the radar, d(t) is the tiny displacement of the chest caused by breathing and heartbeat, and A b is the amplitude of chest movement caused by breathing, f b is the respiratory rate, A h is the amplitude of chest movement caused by heartbeat, f h is the heart rate, is the phase difference caused by heartbeat and breathing.

[0117] Step 2: radar communication integrated echo signal S RX (t) After converting to N parallel data streams, remove the cyclic prefix operation, and then perform FFT transformation to obtain the frequency domain information Y[m,n] of the nth subcarrier echo signal on the mth symbol; the radar communication integrated transmission signal S TX (t) Perform an N-point FFT transform to obtain the frequency domain information X[m,n] of the n-th subcarrier transmitted signal on the m-th symbol, and calculate the phase difference between the echo signal and the transmitted signal:

[0118]

[0119] Among them, H m，n It is represented as the phase information of the nth subcarrier containing vital signs information on the mth symbol;

[0120] Extract H m，n The phase of the nth subcarrier echo signal on the mth symbol is obtained by m，n ; Average all symbols on each subcarrier to obtain the average phase difference Δφ n , extract the phase difference sequence φ after unwrapping n ;

[0121] Step 3: Assume that the phase difference sequence φ after unwrapping is extracted n With frequency f n There is a linear relationship between:

[0122]

[0123] in, is the rate of change of the body's tiny displacement, is the estimated displacement of the body, f n is the center frequency of the nth subcarrier, f n =f c +nΔf;

[0124] The linear relationship between phase difference and frequency is fitted by weighted least square method. and Solve:

[0125]

[0126] According to the phase difference formula Comparative linear relationship The estimated displacement of the target is for:

[0127]

[0128] You can see It fluctuates at a fixed distance R, which is the distance between the human body and the radar. The tiny vibration changes are due to the movement of the chest cavity. Subtracting the fixed distance R can obtain the vital sign related signal f(t).

[0129] Step 4: FGO-VMD parameter optimization

[0130] The vital sign related signal f(t) obtained in step 3 is decomposed by FGO-VMD, and the minimum envelope entropy is used as the fitness function. The fungal growth optimization FGO algorithm is used to obtain the optimal VMD decomposition parameters, such as Figure 2 As shown;

[0131] The specific steps for obtaining the optimal VMD decomposition parameters using the fungal growth optimization FGO algorithm are as follows:

[0132] Step 4-1. Initialize parameters

[0133] Set the range of the number of modal components K and the quadratic penalty factor α of the VMD decomposition parameters, and set the parameters of the FGO algorithm: the trade-off coefficient EM of exploration and development, the probability of environmental effect Ep, the convergence rate ER, the size of the bacterial colony P, and the maximum number of iterations T max , the upper and lower bounds of the VMD decomposition parameter penalty factor α are [lb1, ub1] and the upper and lower bounds of the number of decomposition modal components K are [lb2, ub2], and the bacterial colony is initialized.

[0134] First, a bacterial colony consisting of P hyphae individuals is initialized, where each hyphae individual represents a set of VMD parameters (α and K); the i-th hyphae individual contains a quadratic penalty factor parameter S i，1 and decomposed modal component parameters S i，2 , specifically expressed as:

[0135] S i，j =lb j +rand(0,1)×(ub j -lb j )

[0136] Among them, 1≤i≤P, yield 1, 2; rand(0, 1) represents a random number in the interval [0, 1].

[0137] Step 4-2, fitness evaluation

[0138] For each mycelium individual, the corresponding parameter α = S i，1 and K = S i，2 , substitute it into the VMD algorithm for signal decomposition, and decompose the vital sign related signal f(t) into K narrowband modal functions u with different frequencies k (t), calculate the minimum envelope entropy fitness function F under each mode i .

[0139] The specific process is as follows:

[0140] Decompose the vital sign related signal f(t) into K signals with different frequencies ω k The narrowband modal function u k The process of (t) is expressed as:

[0141]

[0142] Where, 1≤k≤K, K is the number of narrowband modes, u k (t) is the kth narrowband mode function, ω krepresents the kth mode center frequency, δ(t) is the Dirac function, represents the partial derivative with respect to time t, || ||2 represents the second norm, and * represents convolution.

[0143] To solve this constrained problem, the Lagrange multiplier λ(t) and the quadratic penalty term α are introduced to transform the above constrained problem into the following unconstrained variational problem:

[0144]

[0145] Among them, L({u k},{ω k}, λ(t)) represents the augmented Lagrangian function, and <> represents the inner product;

[0146] The above unconstrained variational problem is solved iteratively by the alternating direction multiplier method to obtain K with different center frequencies ω k The narrowband modal function u k (t).

[0147] For each narrowband mode function u k (t) The specific process of calculating the minimum envelope entropy fitness function F is:

[0148] The k-th narrowband mode function u obtained by VMD decomposition k The envelope entropy H of (t) k :

[0149]

[0150] Among them, p k (t) is the normalized envelope signal, satisfying x k (t) is the kth narrowband modal function u k (t) is the envelope signal of the Hilbert transform, which represents the instantaneous amplitude;

[0151] The minimum envelope entropy fitness function F is defined as the minimum value of all modal envelope entropies, that is:

[0152]

[0153] Step 4-3: Update the global optimal solution

[0154] The minimum envelope entropy fitness function F is used as the fitness value, and the mycelium individual with the minimum fitness value is found, and its corresponding decomposition parameter is As the optimal decomposition coefficient of the current VMD, as the latest global optimal solution S best .

[0155] Step 4-4, iterative update

[0156] In each iteration, nutrient allocation and mycelium individual update are performed. According to the current number of iterations T and the maximum number of iterations T max , calculate the nutritional value of each mycelium individual, and the nutritional value is used to control the growth behavior of the mycelium individual.

[0157] First, distribute nutrients according to the stage: in the early stage of iteration Nutrient allocation is used to explore the search space; in the later stages of the iteration Nutrient allocation is used to develop the search space;

[0158] Nutrients assigned to the i-th mycelium individual i for:

[0159]

[0160] Among them, F i is the fitness value of the i-th mycelium individual, and rand(P) is a random number in the interval [0, 1].

[0161] However, the scale of fitness values may be high, causing the updated solution to be outside the search boundary. i Normalized between 0 and 1, as defined below:

[0162]

[0163] in, is the sum of the nutritional values of all mycelium individuals, and r8 is a random number in the interval [0, 1].

[0164] Then iterate and update:

[0165] For each mycelium individual S i，j Calculate energy E r and probability p i :

[0166]

[0167] Among them, minF and maxF are the minimum and maximum fitness values of each hyphae individual in the current iteration, and ε is a very small positive number used to avoid the denominator being zero;

[0168] If p i <E r , then enter the exploration phase and update the mycelium individuals according to the following formula:

[0169]

[0170] E=e fit

[0171]

[0172] in, and represents the i-th mycelium individual in the T+1th iteration and the Tth iteration respectively; U 1，j is a random binary vector, the elements of which are 0 or 1, controlling the update position; S a，j and S b，j are the values of the j-th dimension of two different randomly selected mycelium individuals.

[0173] If p i ≥E r , then proceed to the development phase and update the mycelium individuals according to the following formula:

[0174]

[0175] E c =(r1-0.5)×r2×(S a，j -S b，j )

[0176]

[0177] Among them, D i，j is the update direction of the i-th mycelium individual; E c is a random perturbation term that increases exploration diversity; Ep is the probability of environmental effect, 0<Ep<1, which determines whether to add perturbation; r1, r2, r3, r4, r5, r6, and r7 are all random numbers uniformly distributed in the value range [0, 1].

[0178] If the updated The fitness is less than that before the update The fitness value of The iteration counter is incremented by 1, T = T + 1, and the process returns to step 4-2 until T = T max , output the global optimal solution S best .

[0179] Step 5: Extract the global optimal solution obtained by the fungal growth optimization FGO algorithm in step 4 As the optimal VMD decomposition parameter, the vital sign related signal f(t) is decomposed by VMD to obtain An optimal narrowband mode function Calculate each narrowband mode function The spectrum peak Calculate each narrowband mode function and the Pearson correlation coefficient of the vital sign related signal f(t)

[0180] right and After standardization, it is used as clustering feature and divided into two categories using k-means clustering algorithm, such as Figure 3 As shown, the specific process is:

[0181] Step 5-1, initialize cluster center

[0182] Randomly select Q = 2 from the K cluster features as the initial cluster centers, denoted as cluster centers C1: C1(μ f1 , μ ρ1 ) and cluster center C2: C2(μ f2 , μ ρ2 ); μ f1 and μ ρ1 Represent the spectrum peak and Pearson correlation coefficient of the first randomly selected cluster feature, μ f2 and μ ρ2 Represent the spectrum peak and Pearson correlation coefficient of the second randomly selected cluster feature, respectively.

[0183] Step 5-2: Assign clustering features to cluster centers

[0184] For each narrowband mode function Calculate the Euclidean distance between its clustering features and the two cluster centers and assign it to the cluster with the closest distance, that is:

[0185]

[0186] in, is a narrowband mode function The distance from the cluster center C1, is a narrowband mode function The distance from the cluster center C2.

[0187] Step 5-3, Update cluster center

[0188] For two clusters, the mean of all cluster features in the cluster is used as the cluster center of the current cluster.

[0189] Step 5-4: Repeat steps 5-2 and 5-3 until the cluster center no longer changes or all cluster features are traversed.

[0190] The above-mentioned k-means clustering algorithm can be used to divide the narrowband modal function of the vital sign related signal f(t) into the modal combination of the respiratory signal and the modal combination of the heartbeat signal. For the two obtained cluster modal combinations, the one with the lower cluster center frequency is selected as the modal combination of the respiratory signal, and the one with the higher cluster center frequency is selected as the modal combination of the heartbeat signal.

[0191] The narrowband modal functions of the clustering results of the heartbeat signal and respiratory signal are screened, and the respiratory signal modal combination within 0.1Hz to 0.5Hz is retained as the screened respiratory signal modal combination S b The modal combination of the heartbeat signal of 0.8Hz to 2.5Hz is the filtered heartbeat signal modal combination S h .

[0192] The weight distribution is performed based on the Pearson correlation coefficient between each narrowband modal function and its vital sign related signal f(t):

[0193] Respiration signal weight Expressed as:

[0194]

[0195] Among them, k b is the filtered respiratory signal modal combination S b Each narrowband mode function of is the corresponding filtered respiratory signal modal combination S b Pearson correlation coefficient for each mode in .

[0196] Heartbeat signal weight Expressed as:

[0197]

[0198] Among them, k h is the filtered heartbeat signal modal combination S b Each narrowband mode function of is the corresponding filtered heartbeat signal modal combination S b Pearson correlation coefficient for each mode in .

[0199] Reconstruct the respiratory signal and heartbeat signal according to their weights and corresponding narrowband modal functions to obtain the reconstructed respiratory signal s b (t) and heartbeat signal s h (t):

[0200]

[0201] Step 6: Reconstruct the respiratory signal s obtained in step 5 b (t) and heartbeat signal s h (t) Performing FFT transformation to obtain the spectrum of the extracted respiratory signal and heartbeat signal, and extracting the peak of the spectrum to obtain the extracted respiratory frequency and heartbeat frequency.

[0202] The displacement estimation in the method described in this embodiment is as follows Figure 4As shown in the figure, the comparison between the reconstructed respiratory signal and heartbeat signal and the real situation is shown in Figure 5 As shown, the spectrum of the extracted respiratory signal and heartbeat signal is as follows Figure 6 shown.

[0203] The method described in the present invention can adaptively determine the optimal number of modes and penalty factors of VMD through the fungal growth optimization algorithm (FGO), solving the problems of modal aliasing and signal distortion caused by the traditional VMD's reliance on manual experience to set parameters. It also uses a multi-subcarrier phase difference linear regression model to estimate chest displacement, overcomes the error accumulation problem of the traditional phase unwrapping algorithm in a noisy environment, enhances the detection capability of weak vital sign signals, designs a multi-feature clustering and dynamic weighted reconstruction strategy, combines frequency band energy screening with correlation coefficient weighting, effectively distinguishes respiratory, heartbeat and noise components, suppresses modal aliasing under complex noise interference, and improves the robustness of low signal-to-noise ratio scenarios. Compared with existing intelligent optimization algorithms, the efficient global search characteristics of the FGO algorithm reduce the computational complexity while ensuring parameter optimization accuracy, meeting the real-time requirements of the vital sign monitoring system. The present invention uses FGO-VMD to optimize VMD parameters and improve signal separation and reconstruction strategies, effectively solving the limitations of traditional methods in complex noise environments, achieving high-precision separation of respiratory and heartbeat signals in complex noise environments, and providing a new technical path for achieving efficient and accurate vital sign monitoring.

Claims

1. A method for separating vital signs signals from OFDM radar based on FGO-VMD and multi-feature clustering, characterized in that: The following steps are involved: Step 1: Mix the transmission signal carrying the communication information with the transmission carrier frequency to obtain the radar communication integrated transmission signal S after up-conversion processing. TX (t); After the radar communication integrated transmission signal detects the human chest cavity, it is processed by down-conversion to obtain the radar communication integrated echo signal S carrying vital signs information. RX (t); Step 2: radar communication integrated echo signal S RX (t) After converting to N parallel data streams, remove the cyclic prefix operation, and then perform FFT transformation to obtain the frequency domain information Y[m,n] of the nth subcarrier echo signal on the mth symbol; the radar communication integrated transmission signal S TX (t) Perform an N-point FFT transform to obtain the frequency domain information X[m,n] of the n-th subcarrier transmitted signal on the m-th symbol, and calculate the phase information of the n-th subcarrier containing vital signs information on the m-th symbol Extract H m，n The phase of the nth subcarrier echo signal on the mth symbol is obtained by m，n ; Average all symbols on each subcarrier to obtain the average phase difference Δφ n , extract the phase difference sequence φ after unwrapping n ; Step 3: Fit the phase difference sequence φ by weighted least squares method n With frequency f n The linear relationship between them is used to extract the estimated displacement value of the body; the estimated displacement value of the body is subtracted from the distance between the human body and the radar to obtain the vital sign related signal f(t); Step 4: FGO-VMD parameter optimization Perform FGO-VMD decomposition on the vital sign-related signal f(t) obtained in step 3, use the minimum envelope entropy as the fitness function, and use the fungal growth optimization FGO algorithm to obtain the optimal VMD decomposition parameters; Step 5: Extract the global optimal solution obtained by the fungal growth optimization FGO algorithm in step 4 as the optimal VMD decomposition parameter, and perform VMD decomposition on the vital sign related signal f(t) to obtain An optimal narrowband mode function Calculate each narrowband mode function The spectrum peak Calculate each narrowband mode function and the Pearson correlation coefficient of the vital sign related signal f(t) right and After normalization, the narrowband modal functions of the vital sign-related signals f(t) are used as clustering features. Based on the k-means clustering method, the narrowband modal functions of the vital sign-related signals f(t) are divided into modal combinations of respiratory signals and modal combinations of heartbeat signals. The clustering results are used to reconstruct the respiratory and heartbeat signals. Step 6: Reconstruct the respiratory signal s obtained in step 5 b (t) and heartbeat signal s h (t) Perform FFT transformation to obtain the spectrum of the extracted respiratory signal and heartbeat signal, extract the peak of the spectrum to obtain the extracted respiratory frequency and heartbeat frequency, and then complete the visualization output.

2. The OFDM radar vital sign signal separation method based on FGO-VMD and multi-feature clustering according to claim 1 is characterized in that: In step 1, the radar communication integrated echo signal S RX (t) is expressed as: Where, 0≤n≤N-1, N is the number of subcarriers of the transmission carrier frequency, 0≤m≤M-1, M is the total number of OFDM signal symbols, A(n,m) represents the communication information modulated by the mth symbol on the nth subcarrier, j is the imaginary symbol, Δf is the subcarrier spacing, t is the time, τ is the delay, f c is the carrier frequency, noise is the noise generated by the radar communication channel; The time delay τ is expressed as: Where c0 is the speed of light, R is the initial distance from the human chest to the radar, d(t) is the tiny displacement of the chest caused by breathing and heartbeat, and A b is the amplitude of chest movement caused by breathing, f b is the respiratory rate, A h is the amplitude of chest movement caused by heartbeat, f h is the heart rate, is the phase difference caused by heartbeat and breathing.

3. The OFDM radar vital sign signal separation method based on FGO-VMD and multi-feature clustering according to claim 2 is characterized in that: The specific process of step 3 is: Phase difference sequence φ n With frequency f n There is a linear relationship between: in, is the rate of change of the body's tiny displacement, is the estimated displacement of the body, Qiang is the center frequency of the nth subcarrier, f n =f c +nΔf; The linear relationship between phase difference and frequency is fitted by weighted least square method. and Solve: According to the phase difference formula Comparative linear relationship The estimated displacement of the body is for: Will Subtract R to obtain the vital sign related signal f(t).

4. The OFDM radar vital sign signal separation method based on FGO-VMD and multi-feature clustering according to claim 3 is characterized in that: In step 4, the specific steps for obtaining the optimal VMD decomposition parameters using the fungal growth optimization FGO algorithm are as follows: Step 4-1. Initialize parameters Set the range of the number of modal components K and the quadratic penalty factor α of the VMD decomposition parameters, and set the parameters of the FGO algorithm: the trade-off coefficient EM of exploration and development, the probability of environmental effect Ep, the convergence rate ER, the size of the bacterial colony P, and the maximum number of iterations T max , the upper and lower bounds of the VMD decomposition parameter penalty factor α [lb1, ub1] and the upper and lower bounds of the number of decomposition modal components K [lb2, ub2], initialize the bacterial population; Initialize a bacterial colony containing P hyphae individuals, each hyphae individual represents a set of VMD parameters α and K; the i-th hyphae individual contains a quadratic penalty factor parameter S i，1 and decomposed modal component parameters S i，2 , specifically expressed as: S i，j =lb j +rand(0,1)×(ub j -lb j ) Where, 1≤i≤P, j=1,2; rand(0,1) represents a random number in the interval [0,1]; Step 4-2, fitness evaluation For each mycelium individual, the corresponding parameter α = S i，1 and K = S i，2 , substitute it into the VMD algorithm for signal decomposition, and decompose the vital sign related signal f(t) into K narrowband modal functions u with different frequencies k (t), calculate the minimum envelope entropy fitness function under each mode; Step 4-3: Update the global optimal solution The minimum envelope entropy fitness function F is used as the fitness value, and the mycelium individual with the minimum fitness value is found, and its corresponding decomposition parameter is As the best decomposition coefficient of the current VMD, as the latest global optimal solution; Step 4-4, iterative update In each iteration, nutrient allocation and mycelium individual update are performed; according to the current number of iterations T and the maximum number of iterations T max , calculate the nutritional value of each mycelium individual, which is used to control the growth behavior of the mycelium individual; after the mycelium individual is updated, the iteration counter is increased by 1, T = T + 1, and returns to step 4-2 until T = T max , output the global optimal solution S best .

5. The OFDM radar vital sign signal separation method based on FGO-VMD and multi-feature clustering according to claim 4 is characterized in that: The specific process of step 4-2 is as follows: Decompose the vital sign related signal f(t) into K signals with different frequencies ω k The narrowband mode function u k The process of (t) is expressed as: Where, 1≤k≤K, K is the number of narrowband modes, u k (t) is the kth narrowband mode function, ω k represents the kth mode center frequency, δ(t) is the Dirac function, Represents the partial derivative with respect to time t, ||||2 represents the second norm, and * represents convolution; By introducing the Lagrange multiplier λ(t) and the quadratic penalty term α, the above constrained problem is transformed into the following unconstrained variational problem: Among them, L({u k },{ω k }, λ(t)) represents the augmented Lagrangian function, and <> represents the inner product; The above unconstrained variational problem is solved iteratively by the alternating direction multiplier method to obtain K with different center frequencies ω k The narrowband mode function u k (t); The k-th narrowband mode function u obtained by VMD decomposition k The envelope entropy H of (t) k for: Among them, p k (t) is the normalized envelope signal, satisfying x k (t) is the kth narrowband modal function u k (t) Hilbert transform envelope signal; The minimum envelope entropy fitness function F is defined as the minimum value of all modal envelope entropies, that is:

6. The OFDM radar vital sign signal separation method based on FGO-VMD and multi-feature clustering according to claim 5 is characterized in that: The process of step 4-4 is: First, distribute nutrients according to the stage: in the early stage of iteration Nutrient allocation is used to explore the search space; in the later stages of the iteration, Nutrient allocation is used to develop the search space; Nutrients assigned to the i-th mycelium individual i for: Among them, F i is the fitness value of the ith mycelium individual, rand(P) is a random number in the interval [0, 1]; Nutrients i Normalized, the normalized nutrients are the nutrients assigned to the i-th mycelium individual i Expressed as: Among them, r8 is a random number in the interval [0, 1]; Then iterate and update: For each mycelium individual S i，j Calculate energy E r and probability p i : Among them, minF and maxF are the minimum and maximum fitness values of each hyphae individual in the current iteration, and ε is a very small positive number used to avoid the denominator being zero; If p i <E r , then enter the exploration phase and update the mycelium individuals according to the following formula: And=and fit in, and represents the i-th mycelium individual in the T+1th iteration and the Tth iteration respectively; U 1，j is a random binary vector, the elements of which are 0 or 1, controlling the update position; S a，j and S b，j are the values of the j-th dimension of two different mycelium individuals selected randomly; If p i ≥E r , then proceed to the development phase and update the mycelium individuals according to the following formula: E c =(r1-0.5)×r2×(S a，j -S b，j ) Among them, D i，j is the update direction of the i-th mycelium individual; E c is a random perturbation term that increases exploration diversity; Ep is the probability of environmental effect, 0<Ep<1, which determines whether to add perturbation; r1, r2, r3, r4, r5, r6, and r7 are all random numbers uniformly distributed in the range [0, 1]; If the updated The fitness is less than that before the update The fitness value of 7. The OFDM radar vital sign signal separation method based on FGO-VMD and multi-feature clustering according to claim 6, characterized in that: In step 5, based on the k-means clustering method, the narrowband modal function of the vital sign related signal f(t) is divided into the modal combination of the respiratory signal and the modal combination of the heartbeat signal. The specific process of reconstructing the respiratory signal and the heartbeat signal using the clustering results is as follows: Step 5-1, initialize cluster center Randomly select 2 of the K cluster features as the initial cluster centers, which are respectively recorded as cluster centers C1: C1(μ f1 , μ ρ1 ) and cluster center C2: C2(μ f2 , μ ρ2 ); μ f1 and μ ρ1 Represent the spectrum peak and Pearson correlation coefficient of the first randomly selected cluster feature, μ f2 and μ ρ2 represent the spectrum peak and Pearson correlation coefficient of the second randomly selected cluster feature, respectively; Step 5-2: Assign cluster features to cluster centers For each narrowband mode function Calculate the Euclidean distance between its clustering features and the two cluster centers and assign it to the cluster with the closest distance, that is: in, is a narrowband mode function The distance from the cluster center q, is a narrowband mode function The distance from the cluster center q; Step 5-3, Update cluster center For two clusters, the mean of all cluster features in the cluster is used as the cluster center of the current cluster; Step 5-4: Repeat steps 5-2 and 5-3 until the cluster center no longer changes or all cluster features are traversed, thereby dividing the narrowband modal function of the vital sign-related signal f(t) into a modal combination of the respiratory signal and a modal combination of the heartbeat signal; Step 5-5: For the two obtained cluster modal combinations, the one with a lower cluster center frequency is selected as the modal combination of the respiratory signal, and the one with a higher cluster center frequency is selected as the modal combination of the heartbeat signal; The narrowband modal functions of the clustering results of the heartbeat signal and respiratory signal are screened, and the respiratory signal modal combination within 0.1Hz to 0.5Hz is retained as the screened respiratory signal modal combination S b , retain the modal combination of heartbeat signals from 0.8Hz to 2.5Hz as the filtered heartbeat signal modal combination S h ; The weight distribution is performed based on the Pearson correlation coefficient between each narrowband modal function and its vital sign related signal f(t): Respiration signal weight Expressed as: Among them, k b is the filtered respiratory signal modal combination S b Each narrowband mode function of is the corresponding filtered respiratory signal modal combination S b Pearson correlation coefficient for each modality in ; Heartbeat signal weight Expressed as: Among them, k h is the filtered heartbeat signal modal combination S b Each narrowband mode function of is the corresponding filtered heartbeat signal modal combination S b Pearson correlation coefficient for each modality in ; Reconstruct the respiratory signal and heartbeat signal according to their weights and corresponding narrowband modal functions to obtain the reconstructed respiratory signal s b (t) and heartbeat signal s h (t):

Images