A High-Precision Radar Heart Rate Detection Method Based on Fitted Matched Filtering and Double-CZT

Abstract

This invention discloses a high-precision radar heart rate detection method based on fitted matched filtering and Double-CZT. The method includes: Step 1: Acquiring an intermediate frequency (IF) signal data matrix containing respiratory and heartbeat signals of a target human using a millimeter-wave radar life detection system; Step 2: Preprocessing the IF signal data matrix to obtain an unwrapped signal and a preprocessed signal; Step 3: Fitting the unwrapped signal using smoothed splines to obtain a template signal; then convolving the template signal with the preprocessed signal, the convolution result being the output of the matched filter; Step 4: Variationally extracting the heartbeat signal from the matched filter output; Step 5: Estimating the heart rate using the obtained heartbeat signal using bilinear frequency-modulated Z-transform (Double-CZT). This invention offers high measurement accuracy, strong stability, and is easy for real-time heart rate detection and engineering implementation.

Description

Technical Field

[0001] This invention relates to the fields of radar signal processing and physiological signal detection, specifically to a heart rate detection method for millimeter-wave radar, and to non-contact heart rate detection under different conditions, specifically a high-precision radar heart rate detection method based on fitted matched filtering and Double-CZT. Background Technology

[0002] Currently, most routine vital sign monitoring instruments or devices used in hospitals and homes for heart rate, respiration, and other vital signs are measured using contact methods, such as electrocardiograms, electronic blood pressure monitors, smart bracelets, and finger-pressure pulse oximeters. While contact-based vital sign measurement methods offer high accuracy, they have some limitations in certain special applications: the sensors in contact-based monitoring devices need to be attached to the subject's skin in real time. For special subjects, such as patients with severe burns, significant changes in their skin surface pose challenges to contact measurements. Furthermore, for elderly individuals requiring long-term monitoring, patients with sleep apnea, and infants with sudden cardiac death syndrome, prolonged electrode contact and entanglement of electrode wires can significantly disrupt their normal sleep and daily life.

[0003] Non-contact vital sign detection effectively solves the problems caused by contact between traditional contact detection devices and the body surface. It also provides a convenient and efficient alternative for measuring the vital signs of ordinary patients in traditional medical and health care, possessing a huge application market. my country's aging population continues to deepen; it is estimated that by 2035, the number of people aged 60 and above in my country will exceed 400 million, making elderly care a significant social issue in the future. With the development of 5G and artificial intelligence in new infrastructure construction, medical care and elderly care will tend towards intelligence and interconnection. Non-contact vital sign detection combined with the Internet of Things will be one of the concrete manifestations of this intelligence. Its main application scenarios include:

[0004] (1) Centralized health monitoring of elderly people in nursing homes. The system can monitor the breathing and heartbeat signals of multiple elderly people in real time without the need for one-on-one measurements by guardians, saving manpower costs. At the same time, the system can work around the clock without interruption and can automatically alarm for abnormal data, effectively responding to emergencies and helping to realize smart nursing homes.

[0005] (2) Health monitoring of empty-nest elderly. For empty-nest elderly who live alone all year round, when abnormalities occur in their health conditions, they often cannot be detected in time. Using a non-contact radar vital sign monitoring system, the health condition of the elderly can be monitored remotely. Through a certain data transmission protocol, the health data of the elderly can be sent to their children's mobile phones in real time, which is convenient and fast.

[0006] (3) General health monitoring of family members. For patients with sleep apnea, patients with cardiovascular and cerebrovascular diseases, and family members with severe snoring, non-contact radar vital sign monitoring equipment should be installed indoors. When the monitored person's heartbeat or breathing stops, the equipment can sound an alarm in time, buying more time for rescue.

[0007] (4) Health monitoring of special patients in hospitals. For heart and respiratory monitoring of special patients such as severely burned patients, patients with infectious diseases, and premature infants, non-contact testing has great advantages over contact testing, bringing convenience to patients and medical staff.

[0008] Over the past two decades, significant efforts have been devoted to improving the accuracy of non-contact vital sign detection research. With advancements in semiconductor technology and electronic design methodologies, low-cost, small-size, low-power, and highly integrated radars have become possible. Beyond military applications, radar technology is increasingly being used in fields such as medicine. The most basic type of radar used for vital sign detection is continuous-wave radar, which uses a single frequency to detect phase shifts caused by chest displacement. However, continuous-wave radar suffers from low range resolution, making it difficult to distinguish multiple targets. Additionally, there is ultra-wideband pulse radar. While ultra-wideband pulse radar possesses range resolution and can detect multiple targets, its extremely short pulse transmission allows even relatively weak interference to be transmitted, reducing the signal-to-noise ratio, and its hardware implementation is relatively complex. To address this issue, FMCW radar has been considered and applied in the field of vital sign detection.

[0009] Healthy adults typically have heart rate frequencies of [0.12, 0.5] Hz and [0.8, 2] Hz at rest. Currently, many time-domain and frequency-domain methods have been proposed to extract heart rate signals. For example, filters with a cutoff frequency of [0.8, 2] Hz can be designed to filter out heart rate signals, but the heart rate signal obtained by this method is affected by respiratory harmonics, thus affecting the accuracy of heart rate estimation. In current research, mode decomposition is another method for extracting heart rate signals from complex cardiopulmonary signals. Common mode decomposition methods include empirical mode decomposition, integrated empirical mode decomposition, and fully integrated empirical mode decomposition with adaptive noise. These methods decompose the cardiopulmonary signal into multiple intrinsic mode components. However, they all suffer from varying degrees of mode aliasing and noise residue problems. Moreover, due to strong interference from respiratory signals, the intrinsic mode components of the obtained heart rate signal still contain residuals related to respiration and its harmonics, affecting heart rate detection.

[0010] Furthermore, non-contact vital sign detection is based on sensing minute physiological movements ranging from millimeters to centimeters. However, random body movements with displacements equivalent to or greater than the chest wall displacement caused by vital signs are substantial noise sources capable of corrupting the signal of interest and significantly attenuating the accuracy of vital sign estimation. The fundamental function of vital sign detection requires the detection device to extract low-frequency displacement information caused by these physiological movements from a high-frequency phase-modulated carrier signal, while large-scale random body movements relative to the minute displacements from vital signs corrupt the purity of the vital sign information. In 2016, Ferrars addressed the interference problem caused by random body movements in vital sign detection by installing one radar in front of the body and another behind it. This process reduces interference from random body movements by adding intermediate frequency signals from two radars, but the two radar systems need to measure synchronously, which is difficult to configure and implement in practice. Existing work has designed matched filtering techniques to recover damaged vital sign signals by effectively eliminating noise. However, the effectiveness of this method depends on the purity of the selected template signal.

[0011] Currently, key issues with non-contact vital sign detection based on FMCW radar include: the phase signal generated by the heartbeat is submerged due to random body movement and respiratory harmonics, significantly reducing heart rate detection accuracy. Furthermore, common mode decomposition methods suffer from varying degrees of mode aliasing and noise residue, leading to errors in the extracted heartbeat signal and affecting the final heart rate estimation. Moreover, current common frequency measurement methods either lack sufficient accuracy or involve enormous computational demands, making real-time heart rate detection difficult. Summary of the Invention

[0012] To address the shortcomings of existing technologies, the present invention aims to provide a high-precision radar heart rate detection method based on fitted matched filtering and Double-CZT. This invention reduces interference from noise such as random body movements, improves heart rate detection accuracy, and also exhibits high stability and real-time functionality.

[0013] To achieve the above-mentioned technical objectives, the technical solution adopted by the present invention is as follows:

[0014] A high-precision radar heart rate detection method based on fitted matched filtering and Double-CZT includes the following steps:

[0015] Step 1: Use a millimeter-wave radar life detection system to acquire an intermediate frequency signal data matrix containing the target human's breathing and heartbeat signals;

[0016] Step 2: Preprocess the intermediate frequency signal data matrix to obtain the unwrapped signal and the preprocessed signal;

[0017] Step 3: Apply smooth spline fitting to the unwrapped signal to obtain the template signal; then convolve the template signal with the preprocessed signal, and the convolution result is the output of the matched filter.

[0018] Step 4: Obtain the heartbeat signal from the matched filter output through variational extraction;

[0019] Step 5: Use the double-linear frequency modulated Z-transform (Double-CZT) to estimate the heart rate from the obtained heartbeat signal.

[0020] Preferably, the millimeter-wave radar life detection system includes a transmitter and a receiver;

[0021] The transmitting end includes a waveform transmitter, a voltage-controlled oscillator, and a transmitting antenna;

[0022] The receiving end includes a receiving antenna, a low-noise amplifier, a mixer, a bandpass filter, an analog-to-digital converter, and a digital signal processor connected in sequence.

[0023] The voltage-controlled oscillator is connected to the mixer.

[0024] Preferably, the method in step one specifically includes the following steps:

[0025] Step 1.1: The voltage-controlled oscillator receives the radio frequency signal emitted by the waveform transmitter and modulates it to obtain the transmitted signal. Part of the transmitted signal is transmitted by the transmitting antenna, and the other part of the transmitted signal enters the mixer as the local oscillator signal of the mixer.

[0026] Step 1.2: The transmitted signal emitted by the transmitting antenna is reflected when it encounters the target human body. The receiving antenna receives the reflected signal, i.e., the echo signal, which is amplified by a low-noise amplifier and then sent to the mixer.

[0027] Step 1.3: The mixer performs frequency mixing on the echo signal and the local oscillator signal received by the receiving antenna, and sends the mixed signal to the bandpass filter for further processing to obtain the intermediate frequency signal.

[0028] Step 1.4: The intermediate frequency signal enters the analog-to-digital converter for A / D sampling, and then is processed by the digital signal processor to obtain the intermediate frequency signal data matrix.

[0029] Preferably, step two specifically includes the following steps:

[0030] Step 2.1: Perform a Q-point range-fourth ...

[0031] Then, after removing the static clutter from the range matrix R[m,n], the preprocessed range matrix R'[m,n] is obtained.

[0032]

[0033] Where Q is the number of points in the Range-FFT, and m and n represent the row and column numbers of the distance matrix, respectively;

[0034] Step 2.2: Extract the phase signal containing vital sign information from the preprocessed distance matrix R'[m,n].

[0035] Step 2.3, to make the phase signal The phase signal is always within [-π, π]. Phase unwrapping is performed to obtain the unwrapped signal φ(t).

[0036]

[0037] Where φ(t) represents the unwrapping signal at time t. This represents the phase signal at time t+1;

[0038] Step 2.4: Perform phase difference processing on the unwrapped signal φ(t) to obtain a phase difference signal, thereby reducing frequency offset and enhancing the heartbeat signal;

[0039] Step 2.5: After pulse noise removal processing of the phase difference signal, the preprocessed signal y(t) is obtained.

[0040] Preferably, step three specifically includes the following steps:

[0041] Step 3.1: Based on the phase changes caused by breathing and random body movements, smooth spline fitting is performed on the unwrapped signal φ(t) to remove noise interference and obtain the template signal h(t).

[0042]

[0043] in, It is a phase change caused by breathing and random body movements.

[0044] Step 3.2: The obtained template signal h(t) and preprocessed signal y(t) are used as the input signals of the matched filter, and then the two are matched to obtain the output s(t) of the matched filter.

[0045] s(t)=y(t)*h * (-t)

[0046] Where h * (-t) denotes the conjugate of antisymmetric h(t) in the time domain.

[0047] Preferably, step four specifically includes the following steps:

[0048] Step 4.1: Decompose the output s(t) of the matched filter into a heartbeat signal (desired pattern) u. r (t) and residual signal f k (t),

[0049] Then, based on the normal human heart rate range, and the heartbeat signal u r (t) at the center frequency w r Nearby, the bandwidth I1 of the heartbeat signal is obtained through minimization processing:

[0050]

[0051] Where δ(t) is the Dirac distribution, This represents the partial derivative with respect to variable t, * represents convolution, and i and j are complex units;

[0052] Step 4.2, based on the residual signal f k (t) and heartbeat signal u r Minimizing the spectral overlap between (t) defines a penalty function I2.

[0053]

[0054] Where γ(t) is the impulse response of the filter;

[0055] Step 4.3, set the heartbeat signal u r (t), residual signal f k (t) and center frequency w r The three satisfy each other The convergence condition is given by η, where η is the equilibrium parameter. Then, the heartbeat signal u is obtained through iterator optimization. r (t), the specific steps are as follows:

[0056] Step 4.3.1, combine the output of the matched filter s(t) and the heartbeat signal u r (t), residual signal f k The values ​​of λ(t) and the Lagrange multiplier λ(t) are transformed into the frequency domain representation of the matched filter output through variational extraction. Heartbeat signal residual signal and Lagrange multipliers

[0057] Step 4.3.2: Initialize the heartbeat signal in the frequency domain. residual signal and center frequency w rLet the loop variable be m, and let m = 0.

[0058] Step 4.3.3: Calculate using the alternating direction method of the multiplier algorithm. and Calculate the Lagrange multipliers in the frequency domain using the dual ascent method.

[0059]

[0060]

[0061]

[0062]

[0063] Where τ is the update parameter, and w is the frequency domain representation at time t;

[0064] Step 4.3.4: Set the discrimination precision ζ, where ζ > 0.

[0065] like and When the convergence between them satisfies the following equation,

[0066]

[0067] Then stop iterating, that is, obtain the heartbeat signal in the frequency domain.

[0068] Otherwise, let m = m + 1 and continue with steps 4.3.3-4.3.4;

[0069] Step 4.4, convert the heartbeat signal obtained in the frequency domain of step 4.3 into a frequency domain representation. Converted into heartbeat signal u r (t) is obtained.

[0070] Preferably, step five specifically includes the following steps:

[0071] Step 5.1, define the number of sampling points for the first and second sampling of the bilinear frequency modulated Z-transform as N and M, respectively.

[0072] Step 5.2, first linear frequency modulation Z-transform, define the frequency measurement range as the normal human heart rate range of 0.8-2Hz, and perform frequency sampling spectrum analysis on N points along an arbitrary spiral on the Z plane to obtain the position N1 of the frequency peak;

[0073] Step 5.3, the second linear frequency modulation Z-transform, defines the frequency measurement range as [0.8+(N1-2)*Δf,0.8+(N1+2)*Δf], where Δf=(2-0.8) / N is the frequency resolution of the first linear frequency modulation Z-transform.

[0074] Then, the position N2 of the spectral peak is obtained by frequency sampling and spectral analysis of point M along an arbitrary spiral on the Z plane within this frequency range.

[0075] Step 5.4: The positions N1 and N2 of the spectral peaks obtained by measuring the heartbeat signal through bilinear frequency modulation Z-transform, along with the frequency resolutions Δf and Δf' of the two bilinear frequency modulation Z-transforms, determine the estimated heart rate f. h ,

[0076] f h =0.8+(N1-2)*Δf+N2*Δf'

[0077] Where Δf'=4*Δf / M is the frequency resolution of the second linear frequency modulated Z-transform.

[0078] Compared with the prior art, the beneficial technical effects of this invention are:

[0079] This invention achieves high-precision heart rate detection based on a heart rate detection technique that combines smooth spline fitting with matched filtering, variational extraction to obtain the heartbeat signal, and bilinear frequency modulated Z-transform (Double-CZT) frequency measurement. This invention offers high measurement accuracy, strong stability, and is easy to implement in real-time heart rate detection and engineering applications.

[0080] This invention creatively employs smooth spline fitting to fit the unwrapped signal, eliminating static clutter interference and removing noise to obtain a relatively pure template signal. Simultaneously, convolving the template signal with the preprocessed signal yields an enhanced phase signal after heartbeat information is obtained, improving the signal-to-noise ratio of the heartbeat signal and effectively reducing interference from random body movements and respiratory harmonics. Furthermore, compared to existing empirical mode decomposition methods, the heartbeat signal obtained through variational extraction in this invention is purer, with significantly reduced computational load. Finally, this invention creatively proposes a Double-CZT frequency measurement method for measuring the heartbeat signal, achieving high-precision heart rate detection within a shorter data window. Attached Figure Description

[0081] Figure 1 This is an overall flowchart of the method of the present invention.

[0082] Figure 2 This is the result of smoothing the spline fitting.

[0083] Figure 3 The spectrum diagrams are of the preprocessed signal and the template signal.

[0084] Figure 4 This is the spectrum of the matched filter result.

[0085] Figure 5 This is an accuracy graph for different frequency measurement methods.

[0086] Figure 6 Heart rate curves obtained using different frequency measurement methods.

[0087] Figure 7 The heart rate estimate and reference heart rate value of the present invention are in a resting state. Detailed Implementation

[0088] To further illustrate the present invention, specific implementation examples are given below. It should be noted that the following descriptions are merely embodiments of the present invention and are not intended to limit the present invention. All equivalent modifications, substitutions, and improvements made within the spirit and principles of the present invention should be included within the patent coverage of the present invention.

[0089] A high-precision radar heart rate detection method based on fitted matched filtering and Double-CZT includes the following steps:

[0090] Step 1: Use a millimeter-wave radar life detection system to acquire an intermediate frequency signal data matrix containing the target human's breathing and heartbeat signals;

[0091] Step 2: Preprocess the intermediate frequency signal data matrix to obtain the unwrapped signal and the preprocessed signal;

[0092] Step 3: Apply smooth spline fitting to the unwrapped signal to obtain the template signal; then convolve the template signal with the preprocessed signal, and the convolution result is the output of the matched filter.

[0093] Step 4: Obtain the heartbeat signal from the matched filter output through variational extraction;

[0094] Step 5: Use the double-linear frequency modulated Z-transform (Double-CZT) to estimate the heart rate from the obtained heartbeat signal.

[0095] The millimeter-wave radar life detection system includes a transmitter and a receiver.

[0096] The transmitting end includes a waveform transmitter, a voltage-controlled oscillator, and a transmitting antenna;

[0097] The receiving end includes a receiving antenna, a low-noise amplifier, a mixer, a bandpass filter, an analog-to-digital converter, and a digital signal processor connected in sequence.

[0098] The voltage-controlled oscillator is connected to the mixer.

[0099] The method described in step one specifically includes the following steps:

[0100] Step 1.1: The voltage-controlled oscillator receives the radio frequency signal emitted by the waveform transmitter and modulates it to obtain the transmitted signal. Part of the transmitted signal is transmitted by the transmitting antenna, and the other part of the transmitted signal enters the mixer as the local oscillator signal of the mixer.

[0101] Step 1.2: The transmitted signal emitted by the transmitting antenna is reflected when it encounters the target human body. The receiving antenna receives the reflected signal, i.e., the echo signal, which is amplified by a low-noise amplifier and then sent to the mixer.

[0102] Step 1.3: The mixer performs frequency mixing on the echo signal and the local oscillator signal received by the receiving antenna, and sends the mixed signal to the bandpass filter for further processing to obtain the intermediate frequency signal.

[0103] Step 1.4: The intermediate frequency signal enters the analog-to-digital converter for A / D sampling, and then is processed by the digital signal processor to obtain the intermediate frequency signal data matrix.

[0104] Step two specifically includes the following steps:

[0105] Step 2.1: Perform a Q-point range-fourth ...

[0106] Then, after removing the static clutter from the range matrix R[m,n], the preprocessed range matrix R'[m,n] is obtained.

[0107]

[0108] Where Q is the number of points in the Range-FFT, and m and n represent the row and column numbers of the distance matrix, respectively;

[0109] Step 2.2: Extract the phase signal containing vital sign information from the preprocessed distance matrix R'[m,n].

[0110] Step 2.3, to make the phase signal The phase signal is always within [-π, π]. Phase unwrapping is performed to obtain the unwrapped signal φ(t).

[0111]

[0112] Where φ(t) represents the unwrapping signal at time t. This represents the phase signal at time t+1;

[0113] Step 2.4: Perform phase difference processing on the unwrapped signal φ(t) to obtain a phase difference signal, thereby reducing frequency offset and enhancing the heartbeat signal;

[0114] Step 2.5: After pulse noise removal processing of the phase difference signal, the preprocessed signal y(t) is obtained.

[0115] Step three specifically includes the following steps:

[0116] Step 3.1: Based on the phase changes caused by breathing and random body movements, smooth spline fitting is performed on the unwrapped signal φ(t) to remove noise interference and obtain the template signal h(t).

[0117]

[0118] in, It is a phase change caused by breathing and random body movements.

[0119] Step 3.2: The obtained template signal h(t) and preprocessed signal y(t) are used as the input signals of the matched filter, and then the two are matched to obtain the output s(t) of the matched filter.

[0120] s(t)=y(t)*h * (-t)

[0121] Where h * (-t) denotes the conjugate of antisymmetric h(t) in the time domain.

[0122] Step four specifically includes the following steps:

[0123] Step 4.1: Decompose the output s(t) of the matched filter into a heartbeat signal (desired pattern) u. r (t) and residual signal f k (t),

[0124] Then, based on the normal human heart rate range, and the heartbeat signal u r (t) at the center frequency w r Nearby, the bandwidth I1 of the heartbeat signal is obtained through minimization processing:

[0125]

[0126] Where δ(t) is the Dirac distribution,

[0127] * indicates convolution.

[0128] i and j are complex units;

[0129] Step 4.2, based on the residual signal f k (t) and heartbeat signal u rMinimizing the spectral overlap between (t) defines a penalty function I2.

[0130]

[0131] Where γ(t) is the impulse response of the filter;

[0132] Step 4.3, set the heartbeat signal u r (t), residual signal f k (t) and center frequency w r The three satisfy each other The convergence condition is given by η, where η is the equilibrium parameter. Then, the heartbeat signal u is obtained through iterator optimization. r (t), the specific steps are as follows:

[0133] Step 4.3.1, combine the output of the matched filter s(t) and the heartbeat signal u r (t), residual signal f k The values ​​of λ(t) and the Lagrange multiplier λ(t) are transformed into the frequency domain representation of the matched filter output through variational extraction. Heartbeat signal residual signal and Lagrange multipliers

[0134] Step 4.3.2: Initialize the heartbeat signal in the frequency domain. residual signal and center frequency w r Let the loop variable be m, and let m = 0.

[0135] Step 4.3.3: Calculate using the alternating direction method of the multiplier algorithm. and Calculate the Lagrange multipliers in the frequency domain using the dual ascent method.

[0136]

[0137]

[0138]

[0139]

[0140] Where τ is the update parameter, and w is the frequency domain representation at time t;

[0141] Step 4.3.4: Set the discrimination precision ζ, where ζ > 0.

[0142] like and When the convergence between them satisfies the following equation,

[0143]

[0144] Then stop iterating, that is, obtain the heartbeat signal in the frequency domain.

[0145] Otherwise, let m = m + 1 and continue with steps 4.3.3-4.3.4;

[0146] Step 4.4, convert the heartbeat signal obtained in the frequency domain of step 4.3 into a frequency domain representation. Converted into heartbeat signal u r (t) is obtained.

[0147] Step five specifically includes the following steps:

[0148] Step 5.1, define the number of sampling points for the first and second sampling of the bilinear frequency modulated Z-transform as N and M, respectively.

[0149] Step 5.2, first linear frequency modulation Z-transform, define the frequency measurement range as the normal human heart rate range of 0.8-2Hz, and perform frequency sampling spectrum analysis on N points along an arbitrary spiral on the Z plane to obtain the position N1 of the frequency peak;

[0150] Step 5.3, the second linear frequency modulation Z-transform, defines the frequency measurement range as [0.8+(N1-2)*Δf,0.8+(N1+2)*Δf], where Δf=(2-0.8) / N is the frequency resolution of the first linear frequency modulation Z-transform.

[0151] Then, the position N2 of the spectral peak is obtained by frequency sampling and spectral analysis of point M along an arbitrary spiral on the Z plane within this frequency range.

[0152] Step 5.4: The positions N1 and N2 of the spectral peaks obtained by measuring the heartbeat signal through bilinear frequency modulation Z-transform, along with the frequency resolutions Δf and Δf' of the two bilinear frequency modulation Z-transforms, determine the estimated heart rate f. h ,

[0153] f h =0.8+(N1-2)*Δf+N2*Δf'

[0154] Where Δf'=4*Δf / M is the frequency resolution of the second linear frequency modulated Z-transform.

[0155] In this embodiment, a Texas Instruments millimeter-wave AWR 1642 radar operating at 77-81 GHz is used. The main parameters of the radar system are shown in Table 1.

[0156] Table 1. Main parameters of the radar system

[0157] parameter numerical values initial frequency 77GHz Frequency bandwidth 3.99GHz Frequency slope 70MHz Number of samples per frame 200 ADC sampling time 50μs ADC initial time 7μs

[0158] To verify the effectiveness of this invention, data was collected using a DCA 1000 acquisition board and transmitted to a computer terminal via a USB interface. Then, MATLAB was used for simulation and analysis.

[0159] Experimental verification

[0160] To verify the technical solution of the present invention, the following control experiment was conducted.

[0161] This experiment, conducted under normal heart rate conditions, set up two groups to estimate heart rate. Heart rate was measured using a POLAR H10 chest strap sensor as a reference heart rate, and mean absolute error (MAE) was used as the test metric.

[0162]

[0163] in:

[0164] W represents the total number of time windows within the observation period;

[0165] BPM true (l) represents the reference value within the l-th time window;

[0166] BPM est (l) represents the measurement value within the l-th time window.

[0167] Experiment 1: Estimating heart rate f by frequency measurement of heartbeat signal using Fourier Transform (FFT) h .

[0168] Experiment 2: The method of this invention, based on matched filtering, variational extraction, and Double-CZT, measures the frequency of the heartbeat signal to obtain an estimated heart rate f. h .

[0169] Furthermore, in this example, the number of points in the Range-FFT is set to Q = 256; the center frequency is w. r =1.2; discrimination accuracy ζ =1e-7; balance parameter η =20000; the number of sampling points N and M for the first and second sampling of the bilinear frequency modulated Z-transform are both 128.

[0170] Five individuals of different ages and genders were selected to participate in multiple experiments under Experiment 1 and Experiment 2. During the experiments, the subjects wore a Polar H10 chest heart rate sensor, sat 1 meter away from the radar, and had their chests at the same level as the radar. Before the test, the subjects kept their breathing and heart rate stable and remained as still as possible during the test, with no other human targets within the sensor's detection range.

[0171] The experimental results are shown in Table 3 and Figure 7 As shown, the average heart rate MAE of 5 volunteers at rest was 3.239 bpm and 0.685 bpm, respectively, measured using Experiment 1 and Experiment 2. The results indicate that the MAE of the heart rate measured by this invention is smaller, suggesting that the heart rate measured by this method is closer to the true value and has higher accuracy. Figure 7 The results show that the heart rate estimate obtained by this invention is very close to the reference heart rate value and exhibits the same fluctuation trend, accurately reflecting the change in the current heart rate; while the heart rate estimated in Experiment 1 deviates from the true heart rate value in the 9th-20th and 29th-35th sliding windows.

[0172] Table 3. MAE of heart rate measured in different experiments under resting conditions

[0173]

[0174] The advantages of this invention are that it is the first to remove interference and noise based on matched filtering of smooth spline fitting, which reduces the interference of noise such as random body movement and breathing, improves the signal-to-noise ratio of the heartbeat signal, and obtains a phase signal with enhanced heartbeat information; at the same time, the proposed Double-CZT frequency measurement method can accurately reflect heart rate changes, thereby realizing the frequency measurement of the heartbeat signal.

[0175] like Figure 2-4 As shown, the unwrapped signal, after being fitted with a smooth spline, basically matches the unwrapped signal. Figure 2 The obtained template signal was weakened in the lower frequency range of respiration, and the maximum peak was more pronounced and basically the same as the reference heart rate. Figure 3 This allows the template signal to accurately characterize the heartbeat signal; the results after matched filtering show that the main peak of the heartbeat signal is more prominent, and spurious peaks caused by respiratory harmonics, random body movements, and other interferences can be effectively removed. Figure 4 This technology has obtained accurate heartbeat signals compared to existing technologies.

[0176] Furthermore, simulation experiments were conducted to compare the frequency measurement accuracy of the newly proposed Double-CZT and FFT frequency measurement methods. The simulation experiment settings were as follows: a sinusoidal signal s(t) = 0.2*sin(2π*1.2*t), a signal length of 64, and a sampling frequency of 10Hz. The root mean square error (RMSE) was used as the measurement metric, and 1000 Monte Carlo simulations were performed on the sinusoidal signal at each noise level. RMSE was defined as:

[0177]

[0178] Where N is the number of Monte Carlo experiments, and est(k) and true(k) represent the estimated and true frequencies of the sine wave, respectively. Simulation results are as follows: Figure 5 As shown, a smaller vertical axis value indicates a smaller error, meaning higher frequency measurement accuracy and better performance. This demonstrates that the Double-CZT algorithm proposed in this invention consistently outperforms the FFT frequency measurement method, regardless of whether the signal-to-noise ratio is low or high.

[0179] Figure 6 The results show that the heart rate estimate obtained through Double-CZT frequency measurement exhibits the same fluctuation trend as the reference value, accurately reflecting the current heart rate fluctuation. In contrast, the existing FFT technique shows a large deviation from the reference heart rate in the 25th-35th sliding window.

[0180] The above explains that the present invention removes interference and noise through matched filtering of smooth spline fitting. The Double-CZT frequency measurement method can accurately reflect heart rate changes, thereby realizing the frequency measurement of heartbeat signals.

[0181] Furthermore, we conducted 1000 simulation tests on a computer with an Intel(R) Core(TM) i7-1065G7 CPU @ 1.30GHz and 1.50GHz. The average runtime of the present invention was 0.158 milliseconds, indicating that the present invention can be implemented in real time.

[0182] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above features with, but not limited to, technical features disclosed in this application that have similar functions.

Claims

1. A high-precision radar heart rate detection method based on fitted matched filtering and Double-CZT, characterized in that, The method specifically includes the following steps: Step 1: Use a millimeter-wave radar life detection system to acquire an intermediate frequency signal data matrix containing the target human's breathing and heartbeat signals; Step 2: Preprocess the intermediate frequency signal data matrix to obtain the unwrapped signal and the preprocessed signal; Step 3: Apply smooth spline fitting to the unwrapped signal to obtain the template signal; then convolve the template signal with the preprocessed signal, and the convolution result is the output of the matched filter. Step 4: Variational extraction of the matched filter output to obtain the heartbeat signal; Step 5: Use double-linear frequency modulated Z-transform (Double-CZT) to estimate the heart rate from the obtained heartbeat signal; Step three specifically includes the following steps: Step 3.1: Based on the phase changes caused by breathing and random body movements, smooth spline fitting is performed on the unwrapped signal φ(t) to remove noise interference and obtain the template signal h(t). in, It is a phase change caused by breathing and random body movements. Step 3.2: The obtained template signal h(t) and preprocessed signal y(t) are used as the input signals of the matched filter, and then the two are matched to obtain the output s(t) of the matched filter. s(t)=y(t)*h * (-t) Where h * (-t) denotes the conjugate of antisymmetric h(t) in the time domain; Step four specifically includes the following steps: Step 4.1: Decompose the output s(t) of the matched filter into a heartbeat signal u. r (t) and residual signal f k (t), Then, based on the normal human heart rate range, and the heartbeat signal u r (t) at the center frequency w r Nearby, the bandwidth I1 of the heartbeat signal is obtained through minimization processing: Where δ(t) is the Dirac distribution, * is the convolution, and i and j are complex units; Step 4.2, based on the residual signal f k (t) and heartbeat signal u r Minimizing the spectral overlap between (t) defines a penalty function I2. Where γ(t) is the impulse response of the filter; Step 4.3, set the heartbeat signal u r (t), residual signal f k (t) and center frequency w r The three satisfy each other The convergence condition is given by η, where η is the equilibrium parameter. Then, the heartbeat signal u is obtained through iterator optimization. r (t), the specific steps are as follows: Step 4.3.1, combine the output of the matched filter s(t) and the heartbeat signal u r (t), residual signal f k The values ​​of λ(t) and the Lagrange multiplier λ(t) are transformed into the frequency domain representation of the matched filter output through variational extraction. Heartbeat signal residual signal and Lagrange multipliers Step 4.3.2: Initialize the heartbeat signal in the frequency domain. residual signal and center frequency w r Let the loop variable be m, and let m = 0. Step 4.3.3: Calculate using the alternating direction method of the multiplier algorithm. and Calculate the Lagrange multipliers in the frequency domain using the duality ascent method. Where τ is the update parameter, and w is the frequency domain representation at time t; Step 4.3.4: Set the discrimination precision ζ, where ζ > 0. like and When the convergence between them satisfies the following equation, Then stop iterating, that is, obtain the heartbeat signal in the frequency domain. Otherwise, let m = m + 1 and continue with steps 4.3.3-4. Step 4.4, convert the heartbeat signal obtained in the frequency domain of step 4.3 into a frequency domain representation. Converted into heartbeat signal u r (t) is obtained; Step five specifically includes the following steps: Step 5.1, define the number of sampling points for the first and second sampling of the bilinear frequency modulated Z-transform as N and M, respectively. Step 5.2, first linear frequency modulation Z-transform, define the frequency measurement range as the normal human heart rate range of 0.8-2Hz, and perform frequency sampling spectrum analysis on N points along an arbitrary spiral on the Z plane to obtain the position N1 of the spectrum peak; Step 5.3, the second linear frequency modulation Z-transform, defines the frequency measurement range as [0.8+(N1-2)*Δf,0.8+(N1+2)*Δf], where Δf=(2-0.8) / N is the frequency resolution of the first linear frequency modulation Z-transform. Then, the position N2 of the spectral peak is obtained by frequency sampling and spectral analysis of point M along an arbitrary spiral on the Z plane within this frequency range. Step 5.4: The positions N1 and N2 of the spectral peaks obtained by measuring the heartbeat signal through bilinear frequency modulation Z-transform, along with the frequency resolutions Δf and Δf' of the two bilinear frequency modulation Z-transforms, determine the estimated heart rate f. h , f h =0.8+(N1-2)*Δf+N2*Δf' Where Δf'=4*Δf / M is the frequency resolution of the second linear frequency modulated Z-transform.

2. The high-precision radar heart rate detection method based on fitted matched filtering and Double-CZT as described in claim 1, characterized in that, The millimeter-wave radar life detection system includes a transmitter and a receiver; The transmitting end includes a waveform transmitter, a voltage-controlled oscillator, and a transmitting antenna; The receiving end includes a receiving antenna, a low-noise amplifier, a mixer, a bandpass filter, an analog-to-digital converter, and a digital signal processor connected in sequence. The voltage-controlled oscillator is connected to the mixer; The method described in step one specifically includes the following steps: Step 1.1: The voltage-controlled oscillator receives the radio frequency signal emitted by the waveform transmitter and modulates it to obtain the transmitted signal. Part of the transmitted signal is transmitted by the transmitting antenna, and the other part of the transmitted signal enters the mixer as the local oscillator signal of the mixer. Step 1.2: The transmitted signal emitted by the transmitting antenna is reflected when it encounters the target human body. The receiving antenna receives the reflected signal, i.e., the echo signal, which is amplified by a low-noise amplifier and then sent to the mixer. Step 1.3: The mixer performs frequency mixing on the echo signal and the local oscillator signal received by the receiving antenna, and sends the mixed signal to the bandpass filter for further processing to obtain the intermediate frequency signal. Step 1.4: The intermediate frequency signal enters the analog-to-digital converter for A / D sampling, and then is processed by the digital signal processor to obtain the intermediate frequency signal data matrix.

3. The high-precision radar heart rate detection method based on fitted matched filtering and Double-CZT as described in claim 1, characterized in that, Step two specifically includes the following steps: Step 2.1: Perform a Q-point range-fourth ... Then, after removing the static clutter from the range matrix R[m,n], the preprocessed range matrix R'[m,n] is obtained. Where Q is the number of points in the Range-FFT, and m and n represent the row and column numbers of the distance matrix, respectively; Step 2.2: Extract the phase signal containing vital sign information from the preprocessed distance matrix R'[m,n]. Step 2.3, to make the phase signal The phase signal is always within [-π, π]. Phase unwrapping is performed to obtain the unwrapped signal φ(t). in, This represents the phase signal at time t+1; Step 2.4: Perform phase difference processing on the unwrapped signal φ(t) to obtain a phase difference signal, thereby reducing frequency offset and enhancing the heartbeat signal; Step 2.5: After pulse noise removal processing of the phase difference signal, the preprocessed signal y(t) is obtained.

Images