Heart rate and breath monitoring method based on millimeter wave radar

Abstract

The invention discloses a heart rate and breath monitoring method based on millimeter-wave radar, which comprises the following steps: (S1) system setting and signal acquisition: arranging a millimeter-wave radar sensor at a position facing a human chest area, and adjusting a radar beam center to be aligned with a target area; configuring a radar to transmit frequency modulation continuous wave signals, receiving echo signals reflected by human thoracic cavity micro-motion, and obtaining original ADC data after analog-to-digital conversion; (S2) distance dimension processing: carrying out distance dimension fast Fourier transform on the original ADC data to obtain a distance spectrum containing target distance information; (S3) distance gate locking: searching the maximum amplitude value of the distance spectrum in a preset human body target distance range, and determining a main distance gate where the chest of the human body is located; and (S4) phase extraction: continuously extracting a phase value of each frame period at the main range gate to form an original phase signal. The heart rate and breath monitoring can be better carried out.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of monitoring methods, in particular to a heart rate and respiration monitoring method based on millimeter wave radar. BACKGROUND

[0002] In the existing heart rate and respiration monitoring technology, non-contact monitoring is easily disturbed by environmental noise and human motion, resulting in a decrease in signal quality; there are problems of phase wrapping and abnormal phase jumps caused by non-uniform motion in the phase processing process, affecting the continuity of the phase signal; the fundamental frequency spectrum peak is easily covered by noise during spectrum estimation, and the frequency extraction accuracy is insufficient; the passband of the band-pass filter and the interference judgment threshold are mostly fixed settings, which cannot adapt to the physiological feature differences of the target object and the real-time frequency fluctuations and environmental noise changes; when multiple radars or multiple channels are monitored, the data are not effectively weighted and fused, making it difficult to take advantage of the complementary advantages of multi-source data; at the same time, the original phase signal is disturbed by noise and amplitude fluctuations, further affecting the subsequent processing accuracy, and the above problems result in insufficient anti-interference ability, environmental adaptability and result reliability of heart rate and respiration monitoring, which cannot meet the precise monitoring needs in complex scenarios, therefore, a heart rate and respiration monitoring method based on millimeter wave radar is proposed. SUMMARY

[0003] The present application solves the above technical problems by the following technical solutions, the present application comprises the following steps:

[0004] Step (S1) system setting and signal acquisition: set the millimeter wave radar sensor at a position facing the human chest cavity region, adjust the radar beam center to be aligned with the target region; configure the radar to transmit a frequency-modulated continuous wave signal, and receive the echo signal reflected by the micro-motion of the human chest cavity, and obtain the original ADC data after analog-to-digital conversion;

[0005] Step (S2) distance dimension processing: performing distance dimension fast Fourier transform on the original ADC data to obtain a distance spectrum containing target distance information;

[0006] Step (S3) distance gate locking: searching for the maximum value of the amplitude of the distance spectrum within a preset target human distance range to determine the main distance gate where the human chest cavity is located;

[0007] Step (S4) phase extraction: continuously extracting the phase value of each frame period at the main distance gate to form an original phase signal;

[0008] Step (S5) phase unwrapping: performing phase unwrapping processing on the phase signal obtained in step (S4) to obtain an unwrapped phase signal;

[0009] Step (S6) phase difference: performing first-order difference operation on the unwrapped phase signal to obtain a difference phase signal;

[0010] Step (S7) band separation: filtering the differential phase signal with a band-pass filter to separate the respiratory band signal and the heartbeat band signal;

[0011] Step (S8) spectrum estimation and frequency extraction: performing spectrum estimation on the respiratory band signal and the heartbeat band signal respectively to extract the preliminary respiratory frequency and the preliminary heartbeat frequency;

[0012] Step (S9) motion interference suppression: performing motion interference evaluation and signal quality judgment on the heartbeat band signal, screening stable segment signals based on an adaptive threshold, and checking or correcting the preliminary heartbeat frequency accordingly;

[0013] Step (S10) result fusion and output: if there are preliminary frequencies from multiple radars or multiple processing channels, performing data fusion to obtain a final fusion frequency; outputting the final respiratory frequency and the final heartbeat frequency.

[0014] Further, the motion interference evaluation and signal quality judgment in step (S9) specifically includes:

[0015] performing short-time frame processing on the heartbeat band signal, calculating the amplitude standard deviation and the spectrum entropy of each frame of signal; obtaining an interference index of each frame of signal by weighted fusion of the amplitude standard deviation and the spectrum entropy; setting a dynamic threshold, if the interference index of a frame exceeds the dynamic threshold, it is determined that the frame is seriously interfered by motion and is rejected; keeping consecutive frames with interference index below the threshold as the stable segment signal.

[0016] Further, the setting method of the dynamic threshold is:

[0017] calculating the mean and variance of the interference index in the historical monitoring data; the dynamic threshold is obtained by adding K times the variance to the mean, wherein K is a coefficient adaptively adjusted according to the environmental noise level, and the calculation process is:

[0018] ;

[0019] wherein, T is the dynamic threshold, is the mean of the historical interference index, is the standard deviation of the historical interference index, and K is a regulation coefficient negatively correlated with the environmental signal-to-noise ratio.

[0020] Further, the phase unwrapping processing in step (S5) adopts an improved algorithm combining local gradient detection and polynomial fitting, and the calculation process includes:

[0021] (S5a) Base unwrapping: calculating the phase difference value of adjacent sampling points in the phase signal to be unwrapped; when the absolute value of the phase difference value is greater than π, it is determined that phase wrapping occurs, and preliminary unwrapping is performed by adding or subtracting an integer multiple of 2π to all subsequent phase points to obtain an initial unwrapped signal;

[0022] (S5b) Jump detection and positioning: calculating the local gradient of the initial unwrapped signal; if the absolute value of the local gradient exceeds the preset physiological reasonable change rate threshold in a sliding time window, it is determined that there is an abnormal phase jump caused by non-uniform motion in the window;

[0023] (S5c) Jump correction: for the window with abnormal jumps, a quadratic polynomial is used to fit the normal phase points at both ends of the window; specifically, based on the position index and phase value of each of the M normal phase points at both ends of the window, the coefficients of a quadratic polynomial are solved by least squares method, which is used to describe the smooth trend of the phase change in the window; using the obtained coefficients, the fitting phase value corresponding to each sampling point index in the abnormal jump window is calculated to replace the original phase value in the window, so as to realize smooth correction; the fitting process is realized by the following process:

[0024] ;

[0025] wherein, is the fitted phase value, a, b, and c are the fitting coefficients determined by the least squares method, and n is the sampling point index; finally, the high-continuous unwrapped phase signal is obtained.

[0026] Further, the spectrum estimation in step (S8) adopts a spectral peak enhancement algorithm based on harmonic weight superposition, and the processing procedures of the respiratory frequency band signal and the heartbeat frequency band signal are the same, and each includes the following sub-steps:

[0027] (S8a) Obtaining an initial spectrum: performing Fourier transform on the input frequency band signal to obtain its initial spectrum ;

[0028] (S8b) Fundamental frequency and harmonic identification: identifying the spectral peak corresponding to the vital sign fundamental frequency f0 in the initial spectrum, and finding the frequency position of the harmonic component , wherein k=2,3,...,N is the harmonic order;

[0029] (S8c) Calculating harmonic weight: for each identified harmonic frequency fk, calculate its local signal-to-noise ratio; the ratio of the local signal-to-noise ratio of each harmonic to the sum of the local signal-to-noise ratios of all harmonics is taken as the normalized weight wk of the harmonic;

[0030] (S8d) Spectrum Enhancement and Frequency Extraction: This involves extracting the spectral amplitude at each harmonic frequency. The harmonic amplitudes are weighted according to their corresponding normalized weights wk, and all weighted harmonic amplitude values ​​are summed. The resulting weighted sum is multiplied by an enhancement factor α and then superimposed onto the initial spectral amplitude at the fundamental frequency f0. The amplitude value of the enhanced fundamental spectrum is calculated accordingly. ;

[0031] The calculation process is as follows:

[0032] ;

[0033] Wherein, α is the enhancement factor that controls the intensity of harmonic energy feedback; finally, the frequency corresponding to the peak with the maximum spectral amplitude after enhancement is taken as the initial value of the extracted frequency.

[0034] Furthermore, the data fusion described in step (S10) specifically involves:

[0035] If multiple millimeter-wave radars are used or various parameter configurations are employed to monitor the same target from different angles, the preliminary values ​​of respiratory rate and heart rate acquired independently are fused separately. A weight wi is assigned to each frequency value to be fused, and this weight is positively correlated with the signal-to-noise ratio or spectral peak significance of the signal channel that yielded the frequency value in step (S8) spectral estimation. The fusion process is achieved through weighted summation, i.e., each preliminary frequency value fi is multiplied by its corresponding weight wi, and then all such weighted frequency values ​​are summed to obtain the final fused frequency f. final The calculation process is as follows:

[0036] ;

[0037] Where M is the number of frequency values ​​to be fused, fi is the initial value of the i-th frequency, and wi is its corresponding weight, and satisfies .

[0038] Furthermore, the passband range of the bandpass filter in step (S7) is adaptively adjusted according to the real-time spectral characteristics of the target object, and the calculation process includes:

[0039] (S7a) Initial estimation: During the startup phase or when frequency tracking fails, a preset broadband filter is used to filter and perform spectral analysis on the differential phase signal to obtain an initial estimate of the respiratory rate. and initial estimate of heart rate ;

[0040] (S7b) Dynamically setting the passband: Based on the initial estimated value, dynamically set the passband range of the bandpass filter used in the next processing cycle; specifically, the lower limit of the passband of the respiratory signal filter is set based on the initial estimated value. Subtract The passband upper limit is calculated by multiplying it by a scaling factor β, while the passband upper limit is calculated by adding the same product.

[0041] Similarly, the upper and lower limits of the passband of the heartbeat signal filter are respectively determined by... Subtraction and addition With another proportionality coefficient The product is used for calculation; this calculation process is achieved through the following steps:

[0042] ;

[0043] ;

[0044] ;

[0045] ;

[0046] Wherein, β and γ are proportionality coefficients preset based on individual differences and the measurement environment;

[0047] (S7c) Periodic update: Take the latest extracted frequency preliminary value in step (S8) as the new center frequency estimate, and repeat step (S7b) to realize the periodic tracking and update of the filter passband.

[0048] 7. Furthermore, after step (S4) and before step (S5), a phase signal enhancement step is also included:

[0049] (S4a) Wavelet denoising: for the original phase signal Wavelet transform and threshold denoising are performed to reconstruct the denoised phase signal. ;

[0050] (S4b) Adaptive Amplitude Normalization: Normalize the denoised phase signal to suppress amplitude fluctuations; calculate the root mean square value of all phase values ​​within a sliding window centered at each sampling point n, and record this value as... The instantaneous amplitude at that point is used to characterize the phase value; then the denoised phase value of each sampling point is... Divide by its corresponding instantaneous amplitude value The amplitude-normalized enhanced phase signal is obtained; this normalization calculation process is implemented through the following steps:

[0051] ;

[0052] in, To prevent small positive numbers from being divided by zero;

[0053] The enhanced phase signal This will be used as the input signal for the phase unwrapping process described in step (S5).

[0054] Compared with existing technologies, this invention has the following advantages: This heart rate and respiration monitoring method based on millimeter-wave radar is based on the technical principles and algorithm logic of millimeter-wave radar vital sign monitoring. It integrates a refined signal processing flow and anti-interference design. It accurately locates the target area through range-dimensional Fourier transform and range gate locking, cyclically extracts the phase and obtains continuous and stable phase information through an improved phase unwrapping algorithm, then enhances the vital sign signal and eliminates phase drift through phase difference, and achieves accurate signal separation based on the frequency difference between respiration and heartbeat using adaptive bandpass filtering. It is further enhanced by a spectral peak enhancement algorithm based on harmonic characteristics. The system ensures accurate frequency extraction and, combined with the micro-Doppler principle, filters stable data through spectrum segmentation, threshold judgment, and confidence verification. This effectively suppresses interference from human relative position movement and environmental factors. If multiple radar or multi-channel data are available, weighted fusion is used to further optimize the results. The entire technical solution is progressive, enhancing the continuity of phase signal processing and the significance of spectral peaks, while also improving the accuracy of signal separation and anti-interference capabilities. This significantly improves the reliability and stability of heart rate and respiratory rate monitoring, making it suitable for various application scenarios. Its wide applicability and strong practicality make this system more worthy of widespread adoption. Attached Figure Description

[0055] Figure 1 This is the overall flowchart of the present invention. Detailed Implementation

[0056] The embodiments of the present invention are described in detail below. These embodiments are implemented based on the technical solution of the present invention, and provide detailed implementation methods and specific operation processes. However, the scope of protection of the present invention is not limited to the following embodiments.

[0057] like Figure 1 As shown, this embodiment provides a technical solution: a heart rate and respiration monitoring method based on millimeter-wave radar, comprising the following steps:

[0058] Step (S1) System setup and signal acquisition: Set the millimeter-wave radar sensor to face the chest area of ​​the human body, and adjust the radar beam center to align with the target area; configure the radar to transmit frequency-modulated continuous wave signals and receive the echo signals reflected by the slight movement of the chest cavity of the human body, and obtain the raw ADC data after analog-to-digital conversion;

[0059] Step (S2) Range dimension processing: Perform a fast Fourier transform on the original ADC data in the range dimension to obtain a range spectrum containing target distance information;

[0060] Step (S3) Distance gate locking: Within the preset target distance range for the human body, search for the maximum amplitude of the distance spectrum to determine the main distance gate where the human chest cavity is located;

[0061] Step (S4) Phase Extraction: Continuously extract the phase value of each frame period at the main distance gate to form the original phase signal;

[0062] Step (S5) Phase unwrapping: The phase signal obtained in step (S4) is subjected to phase unwrapping processing to obtain the unfolded phase signal;

[0063] Step (S6) Phase Difference: Perform a first-order difference operation on the expanded phase signal to obtain a differential phase signal;

[0064] Step (S7) Band separation: The differential phase signal is filtered using a bandpass filter to separate the respiratory band signal and the heartbeat band signal;

[0065] Step (S8) Spectrum estimation and frequency extraction: Spectrum estimation is performed on the respiratory frequency band signal and the heartbeat frequency band signal respectively, and preliminary values ​​of respiratory frequency and heartbeat frequency are extracted;

[0066] Step (S9) Motion interference suppression: Motion interference assessment and signal quality judgment are performed on the heartbeat frequency band signal, stable segment signals are screened based on adaptive threshold, and the preliminary value of the heartbeat frequency is verified or corrected accordingly.

[0067] Step (S10) Result fusion and output: If there are preliminary frequency values ​​from multiple radars or multiple processing channels, data fusion is performed to obtain the final fused frequency; the final respiratory rate and heart rate are output.

[0068] The motion interference assessment and signal quality judgment in step (S9) specifically include:

[0069] The heartbeat frequency band signal is processed by short-time framing, and the amplitude standard deviation and spectral entropy of each frame signal are calculated. The interference index of each frame signal is obtained by weighted fusion of the amplitude standard deviation and the spectral entropy. A dynamic threshold is set. If the interference index of a frame exceeds the dynamic threshold, the frame is determined to be severely affected by motion interference and is removed. Continuous frames with interference indices below the threshold are retained as the stable segment signal.

[0070] The heart rate band signal is decomposed into independent analysis units through short - time frame processing. Combining the standard deviation of amplitude to reflect the amplitude fluctuation of the signal and the spectral entropy to characterize the discreteness of the spectral distribution, the weighted fusion of two - dimensional indicators can comprehensively depict the impact of motion interference on the signal, avoid the one - sidedness of single - indicator evaluation, and the dynamic threshold can flexibly adapt to different motion intensities and environmental noise scenarios, accurately eliminate severely interfered frames, and retain stable and effective signals, providing high - quality data support for the verification or correction of the initial heart rate value, significantly improving the anti - interference ability and result accuracy of heart rate monitoring;

[0071] For example, in the daily home scenario of a lonely elderly person, the elderly person is in three states in the bedroom: lying still and resting, slowly turning over to adjust the sleeping position, and getting up to get water. The motion interference is evaluated and the signal quality is judged based on the above method;

[0072] Set the sampling frequency of the heart rate band signal to 60Hz, the short - time frame length to 2s (each frame contains 120 sampling points), the weight of the standard deviation of amplitude wA = 0.45, the weight of the spectral entropy wH = 0.55, and the dynamic threshold is calculated based on the interference index of the historical 80 - frame signal , , take K = 1.7, from get the dynamic threshold T = 0.65+1.7×0.2 = 0.99;

[0073] State 1: Lying still and resting. Take 1 frame of signal in this state and calculate the standard deviation of amplitude. The formula is , where N = 120, and the signal mean , , then ; Perform Fourier transform on this frame of signal to obtain the spectrum P(k) (k = 1,2,...,60, frequency resolution 0.5Hz), and normalize it to get the probability distribution . The formula for spectral entropy is . Calculate to get H1 = 0.55; Through weighted fusion, the interference index I1 = 0.45×0.316 + 0.55×0.55 = 0.4472. Since I1 < T, it is determined that the signal in the stable segment is retained;

[0074] State 2: Slowly turning over to adjust the sleeping position. Take 1 frame of signal in this state, and the signal mean , then ;

[0075] The spectral dispersion increases slightly, and calculate to get H2 = 0.9;

[0076] The interference index I2 = 0.45×0.447 + 0.55×0.9 = 0.70615. Since I2 < T, it is determined that the signal in the stable segment is retained;

[0077] State 3: Get up and get out of bed to fetch water. Take one frame of signal in this state and calculate the signal average. , ,but The spectrum is extremely dispersed, and H3 is calculated to be 2.

[0078] Interference index I3 = 0.45 × 0.999 + 0.55 × 2.0 = 1.44955. Since I3 > T, it is determined to be a severely interfering frame and is therefore removed.

[0079] Based on the stable signals from the first two frames, the initial heart rate value of 68 beats / min was corrected to 66 beats / min, with an error of only 1 beat / min compared to the actual heart rate of 65 beats / min, thus verifying the effectiveness of the method.

[0080] The method for setting the dynamic threshold is as follows:

[0081] Based on the distribution of the interference index in historical monitoring data, its mean and variance are calculated; the dynamic threshold is obtained by adding K times the variance to the mean, where K is a coefficient that adaptively adjusts according to the environmental noise level. The calculation process is as follows:

[0082] ;

[0083] Where T is the dynamic threshold. This represents the average of historical interference indices. is the standard deviation of the historical interference index, and K is the adjustment coefficient that is negatively correlated with the environmental signal-to-noise ratio;

[0084] The dynamic threshold is constructed based on the mean and variance of the interference index of historical monitoring data. The adjustment coefficient K is negatively correlated with the environmental signal-to-noise ratio. It can adaptively match different environmental noise intensity changes, avoid misjudgment or omission in noise fluctuation scenarios with fixed thresholds, accurately adapt to the interference judgment needs in complex environments, further optimize the accuracy of stable segment signal screening, provide a judgment standard that fits the real-time environment for the verification or correction of the initial value of heart rate, and significantly improve the environmental adaptability and result stability of heart rate monitoring.

[0085] For example, in the scenario of home monitoring of elderly people living alone, the environment changes from quiet and undisturbed (high signal-to-noise ratio) to street traffic noise with the window open (medium signal-to-noise ratio), and then to family members walking around indoors (low signal-to-noise ratio). The interference is judged based on the previous dynamic threshold method.

[0086] The mean value was calculated from the interference index distribution of 100 frames of historical heartbeat frequency band signals. Standard deviation The adjustment coefficient K is negatively correlated with the environmental signal-to-noise ratio, specifically the relationship is as follows: signal-to-noise ratio In a quiet environment, K=1.5. (In a moderate noise environment) K=2, (In a low-noise environment) K=2.5; the dynamic threshold calculation formula is as follows: ;

[0087] Scenario 1: Quiet environment, signal-to-noise ratio (SNR) = 25dB, corresponding to K = 1.5. Substituting into the formula, we get... Combined with the interference indices of the three frames in the previous elderly scenario, I1=0.4472, I2=0.70615, and I3=1.44955, I1 and I2 are both less than T1, so they are judged as stable segments of signal retention, while I3 is greater than T1, so it is judged as a severely interfered frame to be removed, ensuring that the interference-free valid data is completely retained.

[0088] Scenario 2: Street traffic noise is present when the window is open; signal-to-noise ratio (SNR) = 15dB, corresponding to K = 2.0. Calculations show... ;

[0089] I1=0.4472 and I2=0.70615 are still less than T2, so they are retained as stable signals. I3=1.44955 is greater than T2 and is discarded. This is suitable for medium noise environments and avoids signals affected by slight noise being misjudged as interference.

[0090] Scenario 3: Family members are moving around indoors, signal-to-noise ratio (SNR) = 8dB, corresponding to K = 2.5, calculated as follows: ;

[0091] If I1=0.4472 and I2=0.70615 are less than T3, they are retained; if I3=1.44955 is greater than T3, they are discarded. At this time, due to the increase in environmental noise, the threshold is appropriately increased to retain the effective signal while avoiding strong interference data from affecting the heart rate calculation.

[0092] Based on the stable signals from the first two frames, the initial heart rate value of 68 beats / min was corrected to 66 beats / min, with an error of only 1 beat / min compared to the elderly person's actual heart rate of 65 beats / min. This verifies the method's adaptability to different noise environments and its accuracy in interference judgment.

[0093] The phase unwrapping process described in step (S5) employs an improved algorithm combining local gradient detection and polynomial fitting. Its calculation process includes:

[0094] (S5a) Basic unwrapping: Calculate the phase difference between adjacent sampling points in the phase signal to be unwrapped; when the absolute value of the phase difference is greater than π, it is determined that phase winding has occurred, and preliminary unwrapping is performed by accumulating or subtracting integer multiples of 2π to the current and subsequent phase points to obtain the initial unfolded signal;

[0095] (S5b) Jump detection and localization: Calculate the local gradient of the initial unfolding signal; within the sliding time window, if the absolute value of the local gradient exceeds the preset physiological reasonable rate of change threshold, it is determined that there is an abnormal phase jump caused by non-uniform motion within the window.

[0096] (S5c) Jump Correction: For windows with abnormal jumps, a quadratic polynomial is used to fit the normal phase points at both ends of the window. Specifically, based on the position indices and phase values ​​of M normal phase points at both ends of the window, the coefficients of a quadratic polynomial are solved using the least squares method. This polynomial describes the smooth trend of phase changes within the window. Using the obtained coefficients, the fitted phase value corresponding to each sampling point index within the abnormal jump window is calculated, and this value replaces the original phase value within the window to achieve smooth correction. The fitting process is implemented through the following steps:

[0097] ;

[0098] in, The fitted phase value is given by a, b, and c, which are the fitting coefficients determined by the least squares method, and n is the sampling point index; finally, the highly continuous expanded phase signal is obtained.

[0099] An improved phase unwrapping algorithm combining local gradient detection and polynomial fitting is adopted. First, the phase winding problem is effectively solved by basic unwrapping. Then, the abnormal phase jump caused by non-uniform motion is accurately located by local gradient detection. Finally, the jump window is smoothed by quadratic polynomial fitting. This avoids the phase discontinuity defects that traditional unwrapping algorithms cannot handle, and ensures the high continuity and stability of the unfolded phase signal. This provides high-quality signal support for subsequent processes such as phase difference and frequency band separation. It is effectively adapted to scenarios where the monitored object has slight non-uniform motion, and improves the anti-interference ability and reliability of heart rate and respiration monitoring results.

[0100] During home-based bed rest monitoring of elderly people living alone, the elderly person slowly turned over during the monitoring process, causing an abnormal jump in the original phase signal. The phase unwrapping method in this case was used to process the signal.

[0101] A specific example is as follows: Set the original phase signal sampling frequency to 60Hz, and the sampling period... The sliding time window length is 0.2s (including 12 sampling points), the physiologically reasonable rate of change threshold is 0.5rad / s, the number of normal phase points at both ends of the window is M=3, and the original phase signals to be unwrapped (n is the sampling point index, unit: rad) are: n=1:1.2, n=2:1.5, n=3:1.8, n=4:2.1, n=5:2.4, n=6:5.8 (jump start point), n=7:6.2, n=8:6.5, n=9:6.8, n=10:3.2 (jump end point), n=11:3.5, n=12:3.8, n=13:4.1, n=14:4.4, n=15:4.7;

[0102] Basic unwrapping: Calculate the phase difference between adjacent sampling points (n=1 to 14), when n=5 The absolute value is greater than If phase winding is detected, the phase points from n=6 to n=10 are cumulatively reduced. After correction: n=6: 5.8-6.28=-0.48rad, n=7: 6.2-6.28=-0.08rad, n=8: 6.5-6.28=0.22rad, n=9: 6.8-6.28=0.52rad, n=10: 3.2-6.28=-3.08rad. The phase values ​​at other points remain unchanged, thus obtaining the initial expanded signal. ;

[0103] Jump detection and localization: Calculating the local gradient of the initial unfolded signal The sliding window takes values ​​from n=4 to n=11, and the window contains... All of them exceeded the physiologically reasonable rate of change threshold, and n=6 to n=10 were determined to be abnormal jump windows;

[0104] Jump correction: Take normal points n=3 (1.8 rad), n=4 (2.1 rad), and n=5 (2.4 rad) at the left end of the window, and normal points n=11 (3.5 rad), n=12 (3.8 rad), and n=13 (4.1 rad) at the right end, and construct a quadratic polynomial. (n∈[6,10]), the coefficients are solved using the least squares method:

[0105] Let the objective function be... Where ni is the index of 6 normal points, For the corresponding phase value, matrix operations are performed. ( , Solving for a and b, we get a≈0.001, b≈0.28, c≈0.96;

[0106] Calculate the fitted phase value within the window:

[0107] n=6:

[0108] ;

[0109] n=7:

[0110] ;

[0111] n=8:

[0112] ;

[0113] n=9:

[0114] ;

[0115] n=10:

[0116] ;

[0117] By replacing the original abnormal phase values ​​with fitted values, a highly continuous expanded phase signal was obtained. The extracted heart rate after phase difference and frequency band separation was 65 beats / min, which was consistent with the actual heart rate of the elderly, thus verifying the effectiveness of the method.

[0118] The spectrum estimation in step (S8) employs a peak enhancement algorithm based on harmonic weighting. The processing procedures for the respiratory frequency band signal and the heartbeat frequency band signal are the same, both including the following sub-steps:

[0119] (S8a) Obtain the initial spectrum: Perform a Fourier transform on the input frequency band signal to obtain its initial spectrum. ;

[0120] (S8b) Fundamental frequency and harmonic identification: In the initial spectrum, identify the spectral peak corresponding to the fundamental frequency f0 of vital signs, and find the frequency positions of its harmonic components. Where k = 2, 3, ..., N are the harmonic orders;

[0121] (S8c) Calculate harmonic weights: For each identified harmonic frequency fk, calculate its local signal-to-noise ratio; take the ratio of the local signal-to-noise ratio of each harmonic to the sum of the local signal-to-noise ratios of all harmonics as the normalized weight wk of that harmonic.

[0122] (S8d) Spectrum Enhancement and Frequency Extraction: This involves extracting the spectral amplitude at each harmonic frequency. The harmonic amplitudes are weighted according to their corresponding normalized weights wk, and all weighted harmonic amplitude values ​​are summed. The resulting weighted sum is multiplied by an enhancement factor α and then superimposed onto the initial spectral amplitude at the fundamental frequency f0. The amplitude value of the enhanced fundamental spectrum is calculated accordingly. ;

[0123] The calculation process is as follows:

[0124] ;

[0125] Where α is the enhancement factor that controls the intensity of harmonic energy feedback; finally, the frequency corresponding to the peak with the maximum spectral amplitude after enhancement is taken as the initial value of the extracted frequency.

[0126] A spectrum peak enhancement algorithm based on harmonic weight superposition is adopted. By identifying the fundamental frequency and harmonic components of vital signs signals, and combining the local signal-to-noise ratio to calculate the harmonic normalization weight, the harmonic energy is fed back to the fundamental frequency in a weighted manner. This significantly enhances the significance of the fundamental spectrum peak, effectively suppresses the influence of environmental noise and weak interference on the spectrum, avoids the problem of the fundamental peak being masked by noise in traditional spectrum estimation, improves the accuracy and reliability of the preliminary value extraction of respiratory and heart rate, and is suitable for monitoring scenarios with weak vital signs signals or strong noise interference.

[0127] For home-based monitoring of elderly people living alone, their heartbeat signals are weak and slightly affected by ambient noise from the indoor air conditioning. The heartbeat frequency is extracted based on the previous peak enhancement algorithm.

[0128] The sampling frequency of the heartbeat frequency band signal is set to 100Hz, the signal duration is 10s, the frequency resolution after Fourier transform is 0.1Hz, the enhancement factor α=0.8, and the harmonic order N=3 (i.e. k=2, 3).

[0129] Obtain the initial spectrum: Perform a Fourier transform on the input heartbeat frequency band signal to obtain the initial spectrum P0(f), where the fundamental frequency f0 = 1.1 Hz (corresponding to a heartbeat frequency of 66 beats / min) and its initial spectrum amplitude P0(f0) = 0.8.

[0130] Fundamental frequency and harmonic identification: Identify the spectral peak of fundamental frequency f0 = 1.1 Hz in the initial spectrum, and then apply the harmonic frequency formula... The harmonic frequencies are calculated to be f2 = 2 × 1.1 = 2.2 Hz and f3 = 3 × 1.1 = 3.3 Hz.

[0131] Calculating harmonic weights: Local signal-to-noise ratio (SNR) is defined as the ratio of the signal amplitude at the harmonic frequency to the average noise amplitude of the five adjacent frequency points. The calculated local SNR of f2 is SNR2 = 15 dB, and the local SNR of f3 is SNR3 = 10 dB. The sum of the local SNRs of all harmonics... According to the normalized weight formula Calculated ;

[0132] Spectrum enhancement and frequency extraction: The amplitudes of the harmonics in the initial spectrum are respectively , Substitute into the peak enhancement formula Calculated ;

[0133] Before enhancement, due to environmental noise interference, the base peak at f0 in the initial spectrum was partially masked by noise, and the initial extracted heart rate was 63 beats / min, which had an error of 3 beats / min compared with the elderly person's actual heart rate of 66 beats / min. After enhancement, the amplitude of the base spectrum peak at f0 increased from 0.8 to 1.136, and the peak significance was significantly enhanced, successfully avoiding noise interference. The initial extracted heart rate was 66 beats / min, which was completely consistent with the actual heart rate, verifying the effectiveness of the algorithm.

[0134] The data fusion described in step (S10) specifically involves:

[0135] If multiple millimeter-wave radars are used or various parameter configurations are employed to monitor the same target from different angles, the preliminary values ​​of respiratory rate and heart rate acquired independently are fused separately. A weight wi is assigned to each frequency value to be fused, and this weight is positively correlated with the signal-to-noise ratio or spectral peak significance of the signal channel that yielded the frequency value in step (S8) spectral estimation. The fusion process is achieved through weighted summation, i.e., each preliminary frequency value fi is multiplied by its corresponding weight wi, and then all such weighted frequency values ​​are summed to obtain the final fused frequency f. final The calculation process is as follows:

[0136] ;

[0137] Where M is the number of frequency values ​​to be fused, fi is the initial value of the i-th frequency, and wi is its corresponding weight, and satisfies ;

[0138] By weighted fusion of preliminary frequency values ​​acquired from multiple radars or multi-processing channels, with the weights being positively correlated with the signal-to-noise ratio or spectral peak significance of the signal channels, the contribution of high-quality signals can be highlighted, the influence of low-quality signals can be weakened, the complementary advantages of multi-source data can be integrated, errors caused by interference or weak signals in a single channel can be avoided, the robustness and accuracy of frequency results can be improved, the complementary needs of multi-source data in complex monitoring environments can be adapted, and the reliability of heart rate and respiration monitoring results can be further guaranteed.

[0139] For home-based monitoring of elderly people living alone, two millimeter-wave radars were deployed (radar 1: directly opposite the head of the bed, radar 2: the corner of the bedroom facing the chest cavity) to simultaneously collect the elderly's heartbeat signals. Due to slight environmental noise and the installation angle of the radars, the initial values ​​of the heartbeat frequency obtained by the two radars differed. The final result was obtained based on the data fusion method.

[0140] Specific examples are as follows: Set the initial number of frequencies to be fused to M=2, the initial heartbeat frequency of radar 1 is f1=64 beats / min, its signal-to-noise ratio (SNR) in the spectrum estimation in step (S8) is SNR1=18dB, and the peak significance is S1=0.92 (peak significance is the ratio of the peak amplitude to the ambient noise amplitude); the initial heartbeat frequency of radar 2 is f2=67 beats / min, its spectrum estimation has an SNR_2=22dB, and the peak significance is S2=0.98;

[0141] The weighting is based on a combined score of signal-to-noise ratio and spectral peak significance, and the combined score formula is as follows: (The coefficients are used to balance the magnitudes of the two indicators), calculate the comprehensive score for the two channels:

[0142] ;

[0143] ;

[0144] According to the weighting formula (satisfy ), calculated to be:

[0145] ;

[0146] ;

[0147] Substitute into the data fusion formula Calculate the final fusion frequency:

[0148] times / minute;

[0149] The elderly person's actual heart rate was 65 beats / min. The error of a single channel of radar 1 was |64-65|=1 beat / min, and the error of a single channel of radar 2 was |67-65|=2 beats / min. After fusion, the error was reduced to |66-65|=1 beat / min, and the result was closer to the actual value. At the same time, the slight signal distortion caused by the slightly off-center installation angle of radar 1 was weakened, and the contribution of the high-quality signal of radar 2 was highlighted. This verifies that the data fusion method can improve the accuracy and robustness of the results.

[0150] The passband range of the bandpass filter in step (S7) is adaptively adjusted according to the real-time spectral characteristics of the target object. The calculation process includes:

[0151] (S7a) Initial estimation: During the startup phase or when frequency tracking fails, a preset broadband filter is used to filter and perform spectral analysis on the differential phase signal to obtain an initial estimate of the respiratory rate. and initial estimate of heart rate ;

[0152] (S7b) Dynamically setting the passband: Based on the initial estimated value, dynamically set the passband range of the bandpass filter used in the next processing cycle; specifically, the lower limit of the passband of the respiratory signal filter is set based on the initial estimated value. Subtract The passband upper limit is calculated by multiplying it by a scaling factor β, while the passband upper limit is calculated by adding the same product.

[0153] Similarly, the upper and lower limits of the passband of the heartbeat signal filter are respectively determined by... Subtraction and addition With another proportionality coefficient The product is used for calculation; this calculation process is achieved through the following steps:

[0154] ;

[0155] ;

[0156] ;

[0157] ;

[0158] Wherein, β and γ are proportionality coefficients preset based on individual differences and the measurement environment;

[0159] (S7c) Periodic update: Take the latest extracted frequency preliminary value in step (S8) as the new center frequency estimate, and repeat step (S7b) to realize the periodic tracking and update of the filter passband;

[0160] The passband range of the bandpass filter is adaptively adjusted based on the real-time spectral characteristics of the target object. The basic passband is determined through initial estimation and updated periodically with the latest preliminary frequency value as the center. The passbands of the respiratory and heartbeat filters are dynamically set according to corresponding proportional coefficients, which can accurately match the real-time frequency changes of respiratory and heartbeat signals, effectively isolate the frequency band interference between the two, avoid the problem of incomplete filtering or loss of effective signals when the signal frequency fluctuates in a fixed passband, adapt to the physiological characteristics of different individuals and frequency drift during the monitoring process, significantly improve the accuracy and stability of frequency band separation, and provide high-purity signal support for subsequent spectrum estimation and frequency extraction.

[0161] Monitoring of elderly people living alone while resting at home showed that their respiratory rate fluctuated slightly with the depth of sleep, and their heart rate changed slightly due to mild emotional relaxation. The respiratory and heart rate frequency bands were separated based on an adaptive bandpass filtering method.

[0162] Set the scaling factor β=0.3 (respiratory signal passband adjustment factor), γ=0.2 (heartbeat signal passband adjustment factor), and the sampling frequency to 100Hz. Use a preset broadband filter to obtain the initial estimate during the startup phase.

[0163] Initial estimation: Through broadband filtering and spectral analysis, the initial estimate of the respiratory rate is obtained. (Corresponding to 15 beats / min), initial estimate of heart rate (Corresponding to 66 times / minute);

[0164] Dynamically set the passband: Calculate the initial passband range according to the formula;

[0165] Respiratory signal bandgap: ;

[0166] Heartbeat signal passband: ;

[0167] At this point, the filter passband accurately covers the initially estimated signal frequency band, filtering out external low-frequency noise (such as floor vibration <0.1Hz) and high-frequency interference (such as electrical noise >1.5Hz).

[0168] Periodic Update: After one processing cycle (5s), the preliminary respiratory rate value extracted in step (S8) is updated to... (Corresponding to 16.8 beats / min), the initial heart rate value has been updated to... (Corresponding to 63 times / minute);

[0169] Update the passband with the new frequency value as the center:

[0170] Passband after respiratory signal update: ;

[0171] Passband after heartbeat signal update: ;

[0172] Continue monitoring for one cycle, and the initial frequency value is updated again. (16.2 times / minute) (64.8 times / min), passband synchronization update is as follows:

[0173] Respiratory band: 0.27×(1-0.3)=0.189Hz to 0.27×(1+0.3)=0.351Hz;

[0174] Heart rate band: 1.08×(1-0.2)=0.864Hz to 1.08×(1+0.2)=1.296Hz;

[0175] Throughout the process, the passband dynamically matched the actual frequency changes of the elderly's breathing and heartbeat. The purity of the separated respiratory frequency band signal reached 92%, and the purity of the heartbeat frequency band signal reached 94%. The respiratory rate of 16 breaths / min and the heart rate of 65 beats / min extracted by subsequent spectrum estimation were completely consistent with the actual physiological state of the elderly, verifying the effectiveness of the method.

[0176] 8. After step (S4) and before step (S5), a phase signal enhancement step is also included:

[0177] (S4a) Wavelet denoising: for the original phase signal Wavelet transform and threshold denoising are performed to reconstruct the denoised phase signal. ;

[0178] (S4b) Adaptive Amplitude Normalization: Normalize the denoised phase signal to suppress amplitude fluctuations; calculate the root mean square value of all phase values ​​within a sliding window centered at each sampling point n, and record this value as... The instantaneous amplitude at that point is used to characterize the phase value; then the denoised phase value of each sampling point is... Divide by its corresponding instantaneous amplitude value The amplitude-normalized enhanced phase signal is obtained; this normalization calculation process is implemented through the following steps:

[0179] ;

[0180] in, To prevent small positive numbers from being divided by zero;

[0181] The enhanced phase signal This will be used as the input signal for the phase unwrapping process described in step (S5);

[0182] The wavelet denoising effectively filters out environmental noise and random interference in the original phase signal, and then adaptive amplitude normalization suppresses the instantaneous amplitude fluctuation of the signal, avoiding the interference of amplitude changes on subsequent phase unwrapping, significantly improving the purity and stability of the phase signal, providing a high-quality input signal for phase unwrapping in step (S5), reducing phase jump misjudgment during unwrapping, adapting to scenarios with weak noise or unstable signal amplitude in the monitoring environment, and further ensuring the overall accuracy and anti-interference capability of heart rate and respiration monitoring;

[0183] For home-based monitoring of elderly people living alone, the original phase signal is affected by high-frequency noise generated by indoor air conditioning and slight vibration of radar sensor, resulting in noise interference and amplitude fluctuations. This signal is processed based on a phase signal enhancement method.

[0184] The original phase signal sampling frequency is set to 60Hz, and the sliding window length is 0.1s (including 5 sampling points), to prevent the removal of tiny zero normal values. The wavelet basis is selected as db4, and the noise reduction threshold is set to 0.1 rad.

[0185] Original phase signals (n is the sampling point index, unit: rad): n=1:1.23, n=2:1.51, n=3:1.78, n=4:2.05, n=5:2.32, n=6:2.59, n=7:2.87 (noise interference point), n=8:3.12, n=9:3.38, n=10:3.65, n=11:3.91, n=12:4.18, n=13:4.45, n=14:4.72, n=15:4.99;

[0186] Wavelet denoising: for the original phase signal A 3-level db4 wavelet decomposition was performed, and the detail coefficients were soft-thresholded (threshold 0.1 rad) to reconstruct the denoised phase signal. The signal values ​​after noise reduction are: n=1:1.22, n=2:1.50, n=3:1.79, n=4:2.04, n=5:2.31, n=6:2.58, n=7:2.85 (noise was filtered out), n=8:3.13, n=9:3.37, n=10:3.64, n=11:3.92, n=12:4.17, n=13:4.46, n=14:4.71, n=15:4.98;

[0187] Adaptive amplitude normalization: The root mean square formula for instantaneous amplitude is as follows ( (Number of sampling points in the window), calculate each sampling point :

[0188] When n=3, within the window , , ;

[0189] When n=7, the window... , , ;

[0190] Other sampling points The calculation is similar; substitute the values ​​into the normalization formula. This yields an enhanced phase signal:

[0191] , The remaining point values ​​are concentrated between 0.685 rad and 1.079 rad;

[0192] Before processing, the original signal fluctuation amplitude reached 0.63 rad. Direct input of phase unwrapping would misjudge n=7 as a transition point. After unwrapping, the extracted heart rate was 61 beats / min, with an error of 4 beats / min compared to the actual heart rate of 65 beats / min. After processing, the signal fluctuation amplitude decreased to 0.394 rad, noise was effectively filtered out, and no misjudgment occurred when the enhanced signal was input for phase unwrapping. The heart rate extracted in subsequent processes was 65 beats / min, which was completely consistent with the actual heart rate, verifying the effectiveness of the method.

[0193] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0194] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0195] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for monitoring heart rate and respiration based on millimeter-wave radar, characterized in that, Includes the following steps: Step (S1) System setup and signal acquisition: Set the millimeter-wave radar sensor to face the chest area of ​​the human body, and adjust the radar beam center to align with the target area; configure the radar to transmit frequency-modulated continuous wave signals and receive the echo signals reflected by the slight movement of the chest cavity of the human body, and obtain the raw ADC data after analog-to-digital conversion; Step (S2) Range dimension processing: Perform a fast Fourier transform on the original ADC data in the range dimension to obtain a range spectrum containing target distance information; Step (S3) Distance gate locking: Within the preset target distance range for the human body, search for the maximum amplitude of the distance spectrum to determine the main distance gate where the human chest cavity is located; Step (S4) Phase Extraction: Continuously extract the phase value of each frame period at the main distance gate to form the original phase signal; Step (S5) Phase unwrapping: The phase signal obtained in step (S4) is subjected to phase unwrapping processing to obtain the unfolded phase signal; Step (S6) Phase Difference: Perform a first-order difference operation on the expanded phase signal to obtain a differential phase signal; Step (S7) Band separation: The differential phase signal is filtered using a bandpass filter to separate the respiratory band signal and the heartbeat band signal; Step (S8) Spectrum estimation and frequency extraction: Spectrum estimation is performed on the respiratory frequency band signal and the heartbeat frequency band signal respectively, and preliminary values ​​of respiratory frequency and heartbeat frequency are extracted; Step (S9) Motion interference suppression: Motion interference assessment and signal quality judgment are performed on the heartbeat frequency band signal, stable segment signals are screened based on adaptive threshold, and the preliminary value of the heartbeat frequency is verified or corrected accordingly. Step (S10) Result Fusion and Output: If there are preliminary frequency values ​​from multiple radars or multiple processing channels, data fusion is performed to obtain the final fused frequency; Output the final respiratory rate and heart rate.

2. The heart rate and respiration monitoring method based on millimeter-wave radar according to claim 1, characterized in that: The motion interference assessment and signal quality judgment in step (S9) specifically include: The heartbeat frequency band signal is subjected to short-time framing processing to calculate the amplitude standard deviation and spectral entropy of each frame signal; the amplitude standard deviation and the spectral entropy are weighted and fused to obtain the interference index of each frame signal; a dynamic threshold is set, and if the interference index of a frame exceeds the dynamic threshold, the frame is determined to be severely affected by motion interference and is discarded; consecutive frames with interference indices below the threshold are retained as the stable segment signal.

3. The heart rate and respiration monitoring method based on millimeter-wave radar according to claim 2, characterized in that: The method for setting the dynamic threshold is as follows: Based on the distribution of the interference index in historical monitoring data, its mean and variance are calculated; the dynamic threshold is obtained by adding K times the variance to the mean, where K is a coefficient that is adaptively adjusted according to the environmental noise level.

4. The heart rate and respiration monitoring method based on millimeter-wave radar according to claim 3, characterized in that: The phase unwrapping process described in step (S5) employs an improved algorithm combining local gradient detection and polynomial fitting. Its calculation process includes: (S5a) Basic unwrapping: Calculate the phase difference between adjacent sampling points in the phase signal to be unwrapped; when the absolute value of the phase difference is greater than π, it is determined that phase winding has occurred, and preliminary unwrapping is performed by accumulating or subtracting integer multiples of 2π to the current and subsequent phase points to obtain the initial unfolded signal; (S5b) Jump detection and localization: Calculate the local gradient of the initial unfolding signal; within the sliding time window, if the absolute value of the local gradient exceeds the preset physiological reasonable rate of change threshold, it is determined that there is an abnormal phase jump caused by non-uniform motion within the window. (S5c) Jump Correction: For windows with abnormal jumps, a quadratic polynomial is used to fit the normal phase points at both ends of the window. Specifically, based on the position index and phase value of each of the M normal phase points at both ends of the window, the coefficients of a quadratic polynomial are solved by the least squares method. This polynomial is used to describe the smooth trend of phase change within the window. Using the obtained coefficients, the fitted phase value corresponding to each sampling point index within the abnormal jump window is calculated, and this value replaces the original phase value within the window to achieve smooth correction, ultimately obtaining the highly continuous expanded phase signal.

5. A heart rate and respiration monitoring method based on millimeter-wave radar according to claim 3, characterized in that... The spectrum estimation in step (S8) employs a peak enhancement algorithm based on harmonic weighting. The processing procedures for the respiratory frequency band signal and the heartbeat frequency band signal are the same, both including the following sub-steps: (S8a) Obtain the initial spectrum: Perform a Fourier transform on the input frequency band signal to obtain its initial spectrum. ; (S8b) Fundamental frequency and harmonic identification: In the initial spectrum, identify the spectral peak corresponding to the fundamental frequency f0 of vital signs, and find the frequency positions of its harmonic components. Where k = 2, 3, ..., N are the harmonic orders; (S8c) Calculate harmonic weights: For each identified harmonic frequency fk, calculate its local signal-to-noise ratio; take the ratio of the local signal-to-noise ratio of each harmonic to the sum of the local signal-to-noise ratios of all harmonics as the normalized weight wk of that harmonic. (S8d) Spectrum Enhancement and Frequency Extraction: This involves extracting the spectral amplitude at each harmonic frequency. The harmonic amplitudes are weighted according to their corresponding normalized weights wk, and all weighted harmonic amplitude values ​​are summed. The resulting weighted sum is multiplied by an enhancement factor α and then superimposed onto the initial spectral amplitude at the fundamental frequency f0. The amplitude value of the enhanced fundamental spectrum is calculated accordingly. ; Finally, the frequency corresponding to the peak with the maximum spectral amplitude after enhancement is used as the initial value of the extracted frequency.

6. The heart rate and respiration monitoring method based on millimeter-wave radar according to claim 5, characterized in that: The data fusion described in step (S10) specifically involves: If multiple millimeter-wave radars are used or various parameter configurations are employed to monitor the same target from different angles, the preliminary values ​​of respiratory rate and heart rate acquired independently are fused separately. A weight wi is assigned to each frequency value to be fused, and this weight is positively correlated with the signal-to-noise ratio or spectral peak significance of the signal channel that yielded the frequency value in step (S8) spectral estimation. The fusion process is achieved through weighted summation, i.e., each preliminary frequency value fi is multiplied by its corresponding weight wi, and then all such weighted frequency values ​​are summed to obtain the final fused frequency f. final .

7. A heart rate and respiration monitoring method based on millimeter-wave radar according to claim 6, characterized in that: The passband range of the bandpass filter in step (S7) is adaptively adjusted according to the real-time spectral characteristics of the target object. The calculation process includes: (S7a) Initial estimation: During the startup phase or when frequency tracking fails, a preset broadband filter is used to filter and perform spectral analysis on the differential phase signal to obtain an initial estimate of the respiratory rate. and initial estimate of heart rate ; (S7b) Dynamically setting the passband: Based on the initial estimated value, dynamically set the passband range of the bandpass filter used in the next processing cycle; specifically, the lower limit of the passband of the respiratory signal filter is set based on the initial estimated value. Subtract The passband upper limit is calculated by multiplying it by a scaling factor β, while the passband upper limit is calculated by adding the same product. Similarly, the upper and lower limits of the passband of the heartbeat signal filter are respectively determined by... Subtraction and addition With another proportionality coefficient Calculated by the product; (S7c) Periodic update: Take the latest extracted frequency preliminary value in step (S8) as the new center frequency estimate, and repeat step (S7b) to realize the periodic tracking and update of the filter passband.

8. A heart rate and respiration monitoring method based on millimeter-wave radar according to claim 7, characterized in that: Previously, it also included a phase signal enhancement step: (S4a) Wavelet denoising: for the original phase signal Wavelet transform and threshold denoising are performed to reconstruct the denoised phase signal. ; (S4b) Adaptive Amplitude Normalization: Normalize the denoised phase signal to suppress amplitude fluctuations; calculate the root mean square value of all phase values ​​within a sliding window centered at each sampling point n, and record this value as... The instantaneous amplitude at that point is used to characterize the phase value; then the denoised phase value of each sampling point is... Divide by its corresponding instantaneous amplitude value This yields an enhanced phase signal after amplitude normalization; The enhanced phase signal This will be used as the input signal for the phase unwrapping process described in step (S5).

Images