A triple constant false alarm rate target detection and multi-frame correlation tracking vital sign monitoring method

Abstract

The application provides a kind of triple constant false alarm target detection and multi-frame association tracking vital sign monitoring method, which comprises the following steps: 1, the original chirp signal (Chirp signal) is preprocessed by difference and normalization;2, the signal after preprocessing is detected by distance gate triple CFAR, and multi-frame association is extracted;3, the target distance gate is extracted;4, the phase is filtered and Fourier transformed, and the phase contains the heartbeat signal spectrum energy, the respiratory signal spectrum energy;5, the heartbeat, respiratory signal is tracked by multi-frame association;6, the frequency data of the heartbeat, respiratory signal after tracking is output.In the use of CFAR for target detection, the application uses the synergistic effect of global, local and fixed triple threshold, which can effectively suppress noise and interference.At the same time, the multi-frame association tracking of vital signs adapts to the continuity of physiological signals, ensuring the stability and accuracy of monitoring.

Description

Technical Field

[0001] This invention belongs to the field of health monitoring, and in particular relates to a method for triple constant false alarm target detection and multi-frame correlation tracking for monitoring vital signs. Background Technology

[0002] Currently, non-contact vital sign monitoring technology has become a research hotspot in the medical field. Frequency modulated continuous wave (FMCW) radar vital sign monitoring is a new type of non-contact vital sign monitoring technology. By transmitting frequency modulated continuous wave signals and receiving reflected signals, it uses the phase change of radar echo signals to obtain displacement information of moving targets, thereby obtaining high-precision vital sign information of human respiration and heartbeat. It has the advantages of being non-contact, highly sensitive, and low-cost.

[0003] Radar systems in operation are inevitably affected by various types of clutter, such as ground clutter, sea clutter, weather clutter, and chaff clutter. Target detection in cluttered environments is a crucial step in radar signal processing. Traditional mean-based constant false alarm rate (CFAR) detectors often fail to achieve satisfactory detection results. To improve target detection performance, clutter map processing techniques are typically incorporated into clutter background detection, which necessitates the use of CFAR detection technology. CFAR technology has also been applied to multi-target range cell detection, combining phase extraction and signal processing algorithms to achieve efficient extraction of breathing and heartbeat signals.

[0004] For demodulation of orthogonal continuous wave radar, researchers have proposed various demodulation methods, such as arctangent demodulation, complex signal demodulation, parameterized demodulation, linear demodulation, and differential cross-multiplication. In parameter estimation, time and frequency domain analysis are fundamental methods in signal processing. Time domain methods include signal autocorrelation and time-domain statistical methods. Frequency domain methods include Fourier transform and Chirp-Z transform; the former is the most commonly used algorithm in signal processing, while the latter provides higher accuracy when the sample size is small. Summary of the Invention

[0005] The purpose of this invention is to solve the above-mentioned technical problems. This invention proposes a method for triple constant false alarm rate (CFAR) target detection and multi-frame correlation tracking for vital sign monitoring, comprising the following steps:

[0006] Step 1: Perform preprocessing on the original chirp signal by differential and normalization, and discretize it to obtain x(n,m), n=1,2,...,N,m=1,2,...,M, where N is the number of sampling points of the discrete signal and M is the number of distance gates;

[0007] Step 2: Perform triple CFAR detection with distance gate on the preprocessed signal, correlate multiple frames, and extract the target distance gate;

[0008] Step 3: Extract the target range gate phase Φ(w);

[0009] Step 4: Filter and perform Fourier transform on the phase to separate the spectral energy of the heartbeat signal contained in the phase. and respiratory signal spectral energy

[0010] Step 5: Perform multi-frame correlation tracking of heartbeat and respiratory signals;

[0011] Step 6: Output the number of heartbeats and respiratory signals after tracking.

[0012] Furthermore, the method of step 1 is as follows:

[0013] Step 1.1: Acquire raw chirp signal data from the FMCW radar, a single chirp signal x T The time-domain representation of (t) is as follows:

[0014]

[0015] Among them, f c S is the initial frequency, S is the slope, t is time, and T is the time interval. c The duration of the signal;

[0016] By filtering and eliminating the sum of terms in the signal after mixing the transmitted and received pulses, the intermediate frequency signal x is obtained. OUT (t):

[0017]

[0018] Among them, f IF φ0 is the intermediate frequency, τ is the time delay, d is the distance between the monitored target and the radar, c is the speed of light, and λ is the wavelength of the signal used by the radar.

[0019] The frequency of the intermediate frequency (IF) signal after mixing is determined solely by the target distance. For different distances, the IF signal is considered as a superposition of multiple single-frequency signals, i.e.:

[0020]

[0021] in, The intermediate frequency corresponding to the distance to the m-th target. Let m be the phase corresponding to the m-th target distance, where m = 1, 2, ..., M, M is the number of distance gates, and m is the index of the distance gate;

[0022] Discretizing the above equation in the time domain, treating each single frequency point as a distance gate, yields the discretized mixing signal x(n,m):

[0023]

[0024] Where n is the number of discrete sampling points, N is the total number of sampling points, m is the index of the distance gate, and M is the number of distance gates;

[0025] Step 1.2: Perform differential processing on the mixed intermediate frequency signal to obtain the differential signal X. diff (n,m):

[0026]

[0027] Among them, X diff Let (n,m) be the differentiated signal, and let X... diff (1,m)=x(1,m);

[0028] Step 1.3: For the differentiated signal X diff Normalize (n,m) to obtain the normalized signal X. norm (n,m):

[0029]

[0030] Among them, X norm (n,m) represents the normalized signal, which is further amplified by the target range gate through normalization. This indicates that the maximum amplitude is taken from the m-th column of the differential signal.

[0031] Furthermore, in step 2, the preprocessed signal is subjected to triple CFAR detection with a range gate to extract the target range gate, specifically including the following steps:

[0032] Step 2.1: Let n = 1;

[0033] Step 2.2: Normalize the signal X norm A full scan is performed along the m-dimensional plane (n, m) to achieve peak detection; the peak points in the signal are the potential target candidate points m. peak Non-peak points are considered as background noise points m noise ;

[0034] Step 2.3: Calculate the global threshold detection threshold based on CA-CFAR, as follows:

[0035] Based on potential target candidate points m peak Set all target candidate points m peak The potential peak set Ω peak ;

[0036] Based on background noise point m noise Set by all background noise points m noise The set of noise points Ω noise ;

[0037] Based on the false alarm probability P fa To generate a global threshold detection threshold based on noise power, first calculate the global noise power μ. global :

[0038]

[0039] Among them, M K This refers to the number of non-peak distance gates.

[0040] Based on the false alarm probability P fa Number of non-peak distance gates M K Calculate the global threshold factor α global :

[0041]

[0042] Based on global noise power μ global and global threshold factor α global Generate global amplitude domain threshold T global :

[0043]

[0044] Step 2.4, Local limit dynamic calculation based on CA-CFAR, the process is as follows:

[0045] Step 2.4.1, with m peak Centered on the protection unit set Ω, the protection unit set is configured as follows: guard :

[0046] Ω guard =[m peak -G,m peak +G]

[0047] Where, m peak ∈Ω peak For each potential target peak point, G is the half-width of the protection unit, if m peak If -G < 0, then the protection unit interval is [0, m peak +G], if m peak If +G>M, then the protection unit interval is [m peak -G,M];

[0048] Reference unit intervals Ω are symmetrically set on both sides of the protection unit. ref :

[0049] Ωref =[m peak -GR,m peak -G-1]∪[m peak +G+1,m peak +G+R]

[0050] Where R is the number of single-sided reference elements, if m peak -G<0, then the reference cell interval is [m peak +G+1,m peak +G+R], if m peak If +G>M, then the protection unit interval is

[0051] [m peak -GR,m peak -G-1], if m peak If -GR < 0, then the reference cell interval is [0, m peak -G-1]∪[m peak +G+1,m peak +G+R], if m peak If +G+R>M, then the protection unit interval is [m peak -GR,m peak -G-1]∪[m peak +G+1,M];

[0052] Step 2.4.2: Extract the signal amplitude from the reference cell and calculate the average local noise power μ. local :

[0053]

[0054] Step 2.4.3: Based on the false alarm probability P fa And reference cell, calculate local limitation factor α locak :

[0055]

[0056] Based on the average power of local noise μ kocal And local departmental limiting factor α local Generate local limited T local :

[0057]

[0058] Step 2.5: Fixed threshold calculation, setting the minimum signal strength threshold T. fixed :

[0059]

[0060] Where β is an empirical coefficient, ω minDetermined by hardware sensitivity, it is the minimum detectable power of the system;

[0061] Step 2.6: Perform threshold decision on peak points to suppress false alarms at non-peak locations, i.e., for m∈Ω peak A target is considered valid if it meets all three of the following threshold conditions, and this target is stored in set Ω. trg In (n):

[0062] Global threshold check:

[0063] |X norm (n,m)|≥T global

[0064] Limited testing by the department:

[0065] |X norm (n,m)|≥T local

[0066] Fixed threshold test:

[0067] |X norm (n,m)|≥T fixed

[0068] Conversely, if the signal strength at the peak point is lower than any threshold, it is determined that there is no target at that location;

[0069] Step 2.7: Let n = n + 1, and determine whether the following conditions are satisfied:

[0070] n≤N

[0071] If the condition is met, proceed to step 2.2; otherwise, proceed to step 2.8.

[0072] Step 2.8: Calculate the number of times the target distance gate appears, F(m):

[0073]

[0074] in, This is an indicator function; it is 1 if the corresponding distance gate in the set has a valid value, and 0 otherwise.

[0075] Step 2.9: Select the target distance gate from the set that is greater than the specified threshold.

[0076]

[0077] Where, m valid Let be the set of target distance gates, and thr be the specified threshold.

[0078] Furthermore, in step 3, the phase information of the target distance gate containing breathing and heartbeat information is extracted, specifically including the following steps:

[0079] Step 3.1, let m valid The number of effective distance gates obtained is K, and k is initialized to 1;

[0080] Step 3.2: Phase Φ of the range gate for the k-th target (k) Extracting the complex signal from the target distance gate using (n). For complex signals The real part, Complex signal The imaginary part, where i is the imaginary unit, is given by:

[0081]

[0082] Among them, tan -1 () represents the arctangent function;

[0083] Step 3.3, for phase Φ (k) (n) Unwrap to obtain the unwrapped phase.

[0084] ΔΦ (k) =Φ (k) (n)-Φ (k) (n-1)

[0085]

[0086] Where, when n=1, let Φ (k) (n-1)=0;

[0087] Step 3.4: Perform a difference operation on the unwound phase to obtain the differential phase.

[0088]

[0089] Where, when n=1, let

[0090] Step 3.5: Differentiate the phase Remove trending items:

[0091] Assuming trend component tr (k) (n) is a polynomial function of degree q with respect to the independent variable n:

[0092] tr (k) (n)=α n n q +α n-1 n q-1 +…+α1n+β

[0093] Least squares fitting of the trend using all phase data points:

[0094]

[0095] Where, α n ,α n-1 ...α1,β are the polynomial coefficients, This is an estimate of the trend term. These are the estimated polynomial coefficients;

[0096] Subtracting the trend term yields the phase Φ after removing the trend. dtr (n):

[0097]

[0098] Furthermore, in step 4, the phase from step 3 is filtered using the following method to separate the heartbeat signal spectral energy contained in the phase. and respiratory signal spectral energy Specifically, the following steps are included:

[0099] Step 4.1: Assuming the angular frequency range of respiration is [w1, w2] and the angular frequency range of heartbeat is [w3, w4], the respiratory rate of an adult is 0.1–0.5 Hz and the heart rate is 0.8–2.0 Hz, so there is no frequency aliasing. The ideal respiratory filter H... b (w) and the ideal heartbeat filter H h The design of (w) is as follows:

[0100]

[0101]

[0102] Where w is the frequency index;

[0103] Step 4.2: Divide the phase into a sliding window, and let the starting index of the sliding window be n. start =1, the sliding window length is set to W L Set the sliding window step size to S1 and initialize the sliding window index j = 0.

[0104] Step 4.3: Calculate the energy of the spectrum of a partial phase under the j-th sliding window, specifically the energy of the breathing spectrum. and heartbeat spectrum energy Defined as:

[0105]

[0106] n start =n start +S1

[0107] Step 4.4: Let j = j + 1, and determine whether the following conditions are satisfied:

[0108]

[0109] If the above conditions are met, proceed to step 4.3; otherwise, proceed to step 5.

[0110] Furthermore, in step 5, the following method is used to perform multi-frame correlation tracking of breathing and heartbeat of multiple or single individuals using a sliding window, specifically including the following steps:

[0111] Step 5.1: Initialize j = 1, and let J be the total number of sliding windows;

[0112] Step 5.2: Voting on dual thresholds for respiration and heartbeat signals:

[0113] Step 5.2.1, First-level voting: Perform preliminary screening on the respiration and heartbeat spectral energy distributions in the j-th analysis window to obtain the first-level candidate frequency set for respiration and heartbeat.

[0114]

[0115] in, `max()` selects the value of `w` that satisfies the relationship within the parentheses; `max()` selects the maximum value within the parentheses.

[0116] Step 5.2.2, Second-level voting: From the first-level candidate set The set of significant frequency points of respiratory and heartbeat energy was selected.

[0117]

[0118] in, To indicate from The set of significant frequency points of respiratory signals selected from the data. To indicate from The set of significant frequency points of heartbeat signals selected from the data, where || represents the number of elements in the set;

[0119] Step 5.3: Let j = j + 1, and determine whether the following conditions are satisfied:

[0120] j≤J

[0121] If the above conditions are met, return to step 5.2; otherwise, proceed to step 5.4.

[0122] Step 5.4: Initialize j = 1;

[0123] Step 5.5: Determine whether the following conditions are true:

[0124] j≤J-num+1

[0125] Where num is the length of the sliding window. If true, proceed to step 5.6; otherwise, proceed to step 6.

[0126] Step 5.6: Accumulation of respiratory weight and heart rate weight:

[0127]

[0128] in, Weight of breathing As the weight of heartbeats, This is an indicator function; it is 1 when the frequency point is matched, and 0 otherwise.

[0129] Step 5.7: Select the center of the frequency band with the highest cumulative weight as the estimated value of the optimal respiratory and heart rate. and

[0130]

[0131] Here, arg max() means selecting the maximum value from the values ​​that satisfy the condition;

[0132] Step 5.8, Dynamic Frequency Fluctuation Constraints and Corrections:

[0133] Step 5.8.1, Fluctuation Range Limitation:

[0134] First frame exemption: If j=1, the initial selection value is directly accepted as the final value of the respiratory rate. Final heart rate Right now:

[0135]

[0136] Conversely, the following constraints are applied:

[0137] Respiratory rate constraint:

[0138] If j≥2, check the deviation between the initial selected respiratory rate and the final value of the previous frame, i.e.:

[0139]

[0140] Define an exception flag. b And initialize it to 0. If the above conditions are met, let Flag... b =1, otherwise let Proceed to step 5.9, where,

[0141] Heart rate constraints:

[0142] If j≥2, check the deviation between the initial selected heart rate and the final value of the previous frame, i.e.:

[0143]

[0144] Define an exception flag. h And initialize it to 0. If the above conditions are met, let Flag... h =1, jump to step 5.8.2, otherwise let Proceed to step 5.9, where,

[0145] Step 5.8.2, Respiratory Rate Correction: If the respiratory rate is marked as an outlier, i.e., Flag... b =1, then smoothing is performed based on historical data from the previous L frames:

[0146]

[0147] Where L is the moving average window length, and L = j-1 when there is insufficient data to construct the sliding window;

[0148] Heart rate correction: If the heart rate is marked as an outlier, i.e., Flag h =1, then smoothing is performed based on historical data from the previous L frames:

[0149]

[0150] Where L is the moving average window length, and L = j-1 when historical data is insufficient;

[0151] Step 5.9: Calculation and output of final results for respiration and heart rate:

[0152]

[0153] Among them, BR (k) (j) represents the number of breaths per minute, HR (k) (j) represents the heart rate per minute;

[0154] Step 5.10: Let j = j + 1, then proceed to step 5.5.

[0155] Furthermore, in step 6, the frequency data of the tracked heartbeat and respiratory signals are output using the following method, specifically including the following steps:

[0156] Step 6.1: Determine whether the following conditions are true:

[0157] k <K

[0158] If true, let k = k + 1 and return to step 3.2; otherwise, proceed to step 6.2.

[0159] Step 6.2: Output the detected respiratory rate BR of group K. (k) (j) Heart rate (HR) (k) (j):

[0160] BR (k) (j),j=1,2,…,J,k=1,2,…,K

[0161] HR (k) (j),j=1,2,…,J,k=1,2,…,K

[0162] Output discrete heart rate and respiratory rate, then end step 6.

[0163] Beneficial effects: Compared with the prior art, the technical solution of the present invention has the following beneficial technical effects:

[0164] 1. This invention innovatively employs a distance-gated triple constant false alarm rate (CFAR) target detection method. As shown in step 2, the triple method mainly refers to setting three threshold values ​​when using CFAR detection. The first is the global threshold T. global The first type is a uniform detection threshold generated based on the average noise power of the entire signal. This threshold adapts to the noise level of the entire signal, providing a unified detection benchmark; the second type is a local threshold T. local The first type is a threshold that adapts to changes in local noise, generated based on local noise estimation around the target candidate point. This threshold can adapt to changes in local noise and suppress interference from nearby strong targets. The third type is a fixed threshold T. fixed This can avoid false detections caused by noise fluctuations or system errors. Through the synergistic effect of the triple threshold, noise and interference can be effectively suppressed, improving the robustness of detection.

[0165] 2. Based on the range-gate triple constant false alarm target detection, this invention uses full-sampling-point reference CFAR result analysis, as shown in step 2. The results of a single CFAR detection are easily affected by clutter interference, leading to instability or false alarms. Using full-sampling-point multi-frame tracking better adapts to this characteristic while ensuring the stability and accuracy of monitoring.

[0166] 3. This invention extracts and analyzes the phase of the target detected in step 2 using the distance gate. As shown in step 3, the phase unwinding operation in step 3.2 restores the phase to its true value, eliminating the discontinuity caused by periodicity, thus more accurately reflecting the target's motion state. Step 3.3, phase difference, subtracts the phase of the intermediate frequency signal from the phase of the previous frame from the phase of the intermediate frequency signal in the next frame, which helps eliminate the phase offset constant and enhances the heartbeat signal component in vital signs. Step 3.4, the detrending term, eliminates relatively uniform body tremor interference, further preventing its interference with respiratory signals.

[0167] 4. This invention employs a multi-frame correlation tracking method for heartbeat and respiration signals. As shown in step 5, a sliding window is used to perform multi-frame correlation tracking of respiration and heartbeat in multiple or single individuals. This is because single-frame detection is susceptible to random noise and limb micro-movement interference, causing frequency estimation jumps, while multi-frame tracking can suppress such instantaneous noise. Furthermore, physiological signals such as respiration and heartbeat are continuous, and their frequencies do not change abruptly. Single-frame detection is difficult to accurately capture their changing trends, while multi-frame tracking can better adapt to this characteristic, while ensuring the stability and accuracy of monitoring. In step 5.2, we introduce a dual threshold voting mechanism. First, by screening frequency points with energy exceeding half of the maximum value, we initially eliminate interference frequencies with lower energy, reducing the possibility of misjudgment. Then, we further screen frequency points with significant energy from the first-level candidate set to ensure that the finally selected frequency points have high reliability and significance. In step 5.3, we innovatively added dynamic frequency fluctuation constraints and corrections. By limiting the fluctuation range of the heartbeat frequency, we can effectively avoid frequency abrupt changes caused by noise or interference. At the same time, frequency points that exceed the fluctuation range are marked, and outliers are smoothed through a moving average correction mechanism, which avoids the impact of single-frame noise on the results and enhances the anti-interference capability of the tracking. Attached Figure Description

[0168] Figure 1 This is a schematic flowchart of the method of the present invention;

[0169] Figure 2 This is a comparison chart of the original Chirp signal data and the data processed in step 1 in Example 1;

[0170] Figure 3 This is the result image after the triple CFAR detection of the distance gate in step 2 in Example 1;

[0171] Figure 4 This is a filter diagram designed in Example 1 to separate respiratory and heartbeat signals;

[0172] Figure 5 This is a diagram showing the effect of multi-frame correlation tracking of heartbeat and respiration after step 5 in Example 1; Detailed Implementation

[0173] To better understand the purpose, method, and function of this invention, the following description, in conjunction with the accompanying drawings, provides a more detailed account of a vital sign monitoring method based on distance gate triple constant false alarm target detection and multi-frame correlation tracking.

[0174] Example 1:

[0175] The test scenario uses a serial-to-USB converter on a TI IWR6843AOPEVM connected to a PC for display and control. A person sits stationary in a chair approximately 1.5 meters from the radar. The slow sampling rate f... s =50Hz, bandwidth BW=0.27GHz, Chirp scan time T sweep =27*10 -6 s.

[0176] Following step 1, 250 seconds of Chirp data were collected, with a total of 8 range gates, i.e., N = 12500, M = 8. The target is located at the 4th range gate. The raw radar data x(n,m) is then subjected to a first-order difference operation to obtain X. diff The differenced signal is normalized to obtain X. norm (n,m). A comparison chart before and after implementing step 1 is shown below. Figure 2 As shown.

[0177] Based on step 2, for X norm Perform triple CFAR detection with distance gates on (n,m) to extract the target distance gate. The target detection result after step 2 is as follows: Figure 3 As shown, by Figure 3 It can be seen that the detected target is at the 4th distance gate, K=1.

[0178] Based on step 3, the phase of the target range gate is extracted to obtain Φ(n), and Φ(n) is unwound to obtain Φ. uwp (n), then for Φ uwp (n) Performing a first-order difference yields Φ diff (n), finally for Φ diff (n) Removing the trend term yields Φ dtr (n).

[0179] Based on step 4, filters are designed to separate the heartbeat and respiratory signals according to their different frequencies, and then the signals are processed by Φ. 2tr (n) Perform sliding window segmentation, with a window length of 1000 and a sliding length of 500, resulting in a window number J = 24. In this embodiment, the respiratory rate range is 0.1–0.5Hz, and the heart rate range is 0.8–2.0Hz. The filter designed in this embodiment is as follows: Figure 4 As shown. Figure 4 The real data is collected by radar equipment while the user is wearing a heart monitor and breathing belt.

[0180] Based on step 5, multi-frame correlation tracking is performed on respiration and heartbeat, and the tracking effect is as follows. Figure 5 As shown. Among them Corresponding to ±6 BPM, Corresponding to ±12 BPM.

[0181] Based on step 6, output the number of breaths and heartbeats (BR). (k) (j),HR (k) (j), the number of outputs is as follows Figure 5 As shown.

[0182] It is understood that the present invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of the present invention.

Claims

1. A method for triple constant false alarm rate target detection and multi-frame correlation tracking for vital sign monitoring, characterized in that, Includes the following steps: Step 1: Perform preprocessing on the original chirped signal, including differencing and normalization, and then discretize it to obtain... , The number of sampling points for the discrete signal. This refers to the number of doors at a distance; Step 2: Perform triple CFAR detection with distance gate on the preprocessed signal, correlate multiple frames, and extract the target distance gate; Step 3: Set the target range gate phase. Extract; Step 4: Filter and perform Fourier transform on the phase to separate the spectral energy of the heartbeat signal contained in the phase. and respiratory signal spectral energy ; Step 5: Perform multi-frame correlation tracking of heartbeat and respiratory signals; Step 6: Output the number of heartbeats and respiratory signals after tracking; In step 2, the preprocessed signal is subjected to triple CFAR detection with a range gate to extract the target range gate, specifically including the following steps: Step 2.1: Let n=1; Step 2.2: Normalize the signal exist A comprehensive scan of all dimensions is performed to achieve peak detection; the peak points in the signal are the potential target candidate points m. peak Non-peak points are considered as background noise points m noise ; Step 2.3: Calculate the global threshold detection threshold based on CA-CFAR, as follows: Based on potential target candidate points m peak Set all target candidate points m peak The potential peak set Ω peak ; Based on background noise point m noise Set by all background noise points m noise The set of noise points Ω noise ; Based on false alarm probability To generate a global threshold detection threshold based on noise power, first calculate the global noise power. : ； in, This refers to the number of non-peak distance gates. Based on false alarm probability Non-peak distance gate number Calculate the global threshold factor : ； Based on global noise power and global threshold factor Generate global amplitude domain threshold : ； Step 2.4, Local limit dynamic calculation based on CA-CFAR, the process is as follows: Step 2.4.1, with m peak Centered on the protection unit set Ω, the protection unit set is configured as follows: guard ; ； in, For each potential target peak point, G is the half-width of the protection unit. Then the protection unit interval is ,like Then the protection unit interval is ; Reference unit intervals Ω are symmetrically set on both sides of the protection unit. ref : ; Where R is the number of single-sided reference elements, if Then the reference cell interval is ,like Then the protection unit interval is ,like Then the reference cell interval is ,like Then the protection unit interval is ; Step 2.4.2: Extract the signal amplitude from the reference cell and calculate the average power of local noise. : ； Step 2.4.3: Based on the false alarm probability And reference unit, calculate local limitation factor : ； Based on the average power of local noise and departmental limitations Generate local departmental limitations : ； Step 2.5: Fixed threshold calculation, setting the minimum signal strength threshold. : ; Where β is an empirical coefficient. Determined by hardware sensitivity, it is the minimum detectable power of the system; Step 2.6: Perform threshold judgment on peak points to suppress false alarms at non-peak locations, i.e., for... A target is considered valid and is stored in a set if it meets all three of the following threshold conditions. middle: Global threshold check: ； Limited testing by the department: ； Fixed threshold test: ； Conversely, if the signal strength at the peak point is lower than any threshold, it is determined that there is no target at that location; Step 2.7: Let n = n + 1, and determine whether the following conditions are satisfied: ； If the condition is met, proceed to step 2.2; otherwise, proceed to step 2.

8. Step 2.8: Calculate the number of target distance gates that appear. : ； in, ( ) is an indicator function; it is 1 if the set has a valid value for the distance gate, and 0 otherwise. Step 2.9: Select the target distance gate from the set that is greater than the specified threshold. ； in, For the set of target distance gates, The specified threshold.

2. The method for triple constant false alarm target detection and multi-frame correlation tracking for vital sign monitoring according to claim 1, characterized in that, The specific method for step 1 is as follows: Step 1.1: Acquire raw chirp signal data from the FMCW radar, single chirp signal. The time-domain representation is as follows: ； in, The starting frequency, The slope For time, The duration of the signal; The intermediate frequency signal is obtained by filtering and eliminating the sum of terms in the signal after mixing the transmitted and received pulses. : ； ； ； in, This is the intermediate frequency. τ is the phase, d is the time delay, c is the distance between the monitored target and the radar, λ is the speed of light, and λ is the wavelength of the signal used by the radar. The frequency of the intermediate frequency (IF) signal after mixing is determined solely by the target distance. For different distances, the IF signal is considered as a superposition of multiple single-frequency signals, i.e.: ； in, The intermediate frequency corresponding to the distance to the m-th target. The phase corresponding to the distance to the m-th target. , is the number of distance gates, and m is the index of the distance gate; Discretize the above equation in the time domain, treating each single frequency point as a distance gate, to obtain the discretized mixing signal. : ； Where n is the number of discrete sampling points, N is the total number of sampling points, and m is the index of the distance gate. This refers to the number of doors at a distance; Step 1.2: Perform differential processing on the mixed intermediate frequency signal to obtain the differential signal X. diff (n,m): ； Among them, X diff Let (n,m) be the differentiated signal, and let X... diff (1,m)=x(1,m); Step 1.3: For the differentiated signal X diff Normalize (n,m) to obtain the normalized signal X. norm (n,m): ； Among them, X norm (n,m) represents the normalized signal, which is further amplified by the target range gate through normalization. Indicates the signal after differential processing. Take the maximum value of the amplitude from the column.

3. The method for triple constant false alarm target detection and multi-frame correlation tracking for vital sign monitoring according to claim 1, characterized in that, Step 3 involves extracting the phase information of the target distance gate, which contains information about breathing and heartbeat. This includes the following steps: Step 3.1, set The number of effective distance gates obtained is K, and k is initialized to 1; Step 3.2: Phase of the range gate for the k-th target. Extract the complex signal of the target range gate. , Complex signal The real part, Complex signal The imaginary part, where i is the imaginary unit, is given by: ； Among them, tan -1 () represents the arctangent function; Step 3.3, Phase Unwinding yields the unwound phase. : ； ； Among them, when season ; Step 3.4: Perform a difference operation on the unwound phase to obtain the differential phase. : ； Among them, when season ; Step 3.5: Differentiate the phase Remove trending items: Assuming trend components It's about the independent variable. q-degree polynomial function: ； Least squares fitting of the trend using all phase data points: ； in, For polynomial coefficients, This is an estimate of the trend term. These are the estimated polynomial coefficients; Subtracting the trend term yields the phase after removing the trend. : 。 4. The method for triple constant false alarm target detection and multi-frame correlation tracking for vital sign monitoring according to claim 1, characterized in that, In step 4, the phase from step 3 is filtered using the following method to separate the heartbeat signal spectral energy contained in the phase. and respiratory signal spectral energy Specifically, it includes the following steps: Step 4.1, Assume the angular frequency range of breathing is... The range of heartbeat angular frequency is An adult's respiratory rate is 0.1–0.5 Hz, and their heart rate is 0.8–2.0 Hz, so there is no frequency aliasing, making it an ideal respiratory filter. and ideal heartbeat filter The design is as follows: ； ； Where w is the frequency index; Step 4.2: Divide the phase into a sliding window, and let the starting index of the sliding window be n. start =1, the sliding window length is set to W. L Set the sliding window step size to S1 and initialize the sliding window index j = 0. Step 4.3: Calculate the energy of the spectrum of a partial phase under the j-th sliding window, specifically the energy of the breathing spectrum. and heartbeat spectrum energy Defined as: ； ； ； Step 4.4: Let j = j + 1, and determine whether the following conditions are met: ； If the above conditions are met, proceed to step 4.3; otherwise, proceed to step 5.

5. The method for triple constant false alarm target detection and multi-frame correlation tracking for vital sign monitoring according to claim 1, characterized in that, In step 5, the following method is used to perform multi-frame correlation tracking of breathing and heartbeat of multiple or single individuals using a sliding window, specifically including the following steps: Step 5.1: Initialize j=1, and let J be the total number of sliding windows; Step 5.2: Voting on dual thresholds for respiration and heartbeat signals: Step 5.2.1, First-level voting: Perform preliminary screening on the respiration and heartbeat spectral energy distributions in the j-th analysis window to obtain the first-level candidate frequency set for respiration and heartbeat. (j) (j): ； ； in, `max()` selects the value of `w` that satisfies the relationship within the parentheses; `max()` selects the maximum value within the parentheses. Step 5.2.2, Second-level voting: From the first-level candidate set , The set of significant frequency points of respiratory and heartbeat energy was selected. , : ； ； in, To indicate from The set of significant frequency points of respiratory signals selected from the data. To indicate from The set of significant frequency points of heartbeat signals selected from the data, where | represents the number of elements in the set; Step 5.3: Let j = j + 1, and determine whether the following conditions are met: ； If the above conditions are met, return to step 5.2; otherwise, proceed to step 5.

4. Step 5.4: Initialize j=1; Step 5.5: Determine whether the following conditions are true: ； Where num is the length of the sliding window. If true, proceed to step 5.6; otherwise, proceed to step 6. Step 5.6: Accumulation of respiratory weight and heart rate weight: ； ； in, Weight of breathing Weighting of heartbeats ( ) is an indicator function, which is 1 when the frequency point is matched, and 0 otherwise; Step 5.7: Select the center of the frequency band with the highest cumulative weight as the estimated value of the optimal respiratory and heart rate. and : ； ； Here, arg max() means selecting the maximum value from the values ​​that satisfy the condition; Step 5.8, Dynamic Frequency Fluctuation Constraints and Corrections: Step 5.8.1, Fluctuation Range Limitation: First frame exemption: If j=1, the initial value is directly accepted as the final value of the respiratory rate. Final heart rate ,Right now: ； ； Conversely, the following constraints are applied: Respiratory rate constraint: If j≥2, check the deviation between the initial selected respiratory rate and the final value of the previous frame, i.e.: ； Define an exception flag. b And initialize it to 0. If the above conditions are met, let Otherwise Proceed to step 5.9, where, ; Heart rate constraints: If j≥2, check the deviation between the initial selected heart rate and the final value of the previous frame, i.e.: ； Define an exception flag. h And initialize it to 0. If the above conditions are met, let If so, proceed to step 5.8.2; otherwise, let... Proceed to step 5.9, where, ; Step 5.8.2, Respiratory Rate Correction: If the respiratory rate is marked as an outlier, i.e. Then smoothing is performed based on historical data from the previous L frames: ； Where L is the moving average window length, and L=j-1 when there is insufficient data to construct the sliding window; Heart rate correction: If the heart rate is marked as an outlier, i.e. Then smoothing is performed based on historical data from the previous L frames: ； Where L is the moving average window length, and L=j-1 when historical data is insufficient; Step 5.9: Calculation and output of final results for respiration and heart rate: ； ； in, The number of breaths per minute. Heart rate per minute; Step 5.10: Let j = j + 1, then proceed to step 5.

5.

6. The method for triple constant false alarm target detection and multi-frame correlation tracking for vital sign monitoring according to claim 1, characterized in that, In step 6, the number of heartbeats and respiratory signals after tracking is output using the following method, specifically including the following steps: Step 6.1: Determine whether the following conditions are true: ； If true, let k = k + 1 and return to step 3.2; otherwise, proceed to step 6.

2. Step 6.2: Output the detected number of breaths in group K. Heart rate : ； ； Output discrete heart rate and respiratory rate, then end step 6.

Images