A rapid and accurate sign detection method based on millimeter wave radar

Abstract

The application discloses a kind of based on millimeter wave radar's sign rapid and accurate detection method.Using the present application can quickly detect vital signs signal, and detection precision is high, and complexity is low.The application first mixes the echo signal and the transmission signal, and the complex signal obtained by cross sampling;Then the variance based on the energy of slow time dimension of complex signal is used, the best distance unit is selected without destroying target phase signal, to ensure that the best distance unit contains human breathing and heartbeat signal;At the same time, before heart rate estimation, the high-order harmonic of breathing signal is first eliminated, and then the final result of heart rate is output according to the amplitude size, so that the heart rate estimation has high accuracy.The application reduces the requirement of signal processing and estimation signal time length under the condition of ensuring detection accuracy, so that the present application meets the accuracy requirement of high-precision rapid heart rate estimation, and can adapt to rapid detection in various scenes.

Description

Technical Field

[0001] This invention relates to the field of radar equipment technology, and specifically to a method for rapid and accurate detection of vital signs based on millimeter-wave radar. Background Technology

[0002] As the aging population trend deepens and the number of people with chronic diseases gradually increases, the demand for routine monitoring of vital signs such as respiration and heartbeat is growing to alleviate the strain on medical resources and to enable early detection and treatment of diseases. Contact-based vital sign monitoring devices, such as ECGs and electrodes, require contact with the human body to perform the monitoring, which restricts the user's range of motion, causes discomfort, and is unsuitable for special populations such as burn patients and patients with infectious diseases. Non-contact vital sign monitoring devices, such as cameras, may infringe on user privacy and are easily affected by environmental conditions. While WiFi devices or continuous wave radar devices lack the ability to detect multiple targets at a distance, FMCW radar, with its small size, low cost, high detection accuracy, and ability to achieve non-intrusive detection of multiple human targets while protecting privacy, has enormous application potential in home-based routine monitoring and medical monitoring scenarios.

[0003] FMCW radar can acquire target range information through frequency changes and obtain subtle motion information of the target through phase changes, such as chest rise and fall caused by breathing and heartbeat. Current methods for detecting vital signs such as breathing and heart rate based on FMCW radar mainly face the following challenges:

[0004] Firstly, the current common method for selecting range units is based on energy amplitude, such as the method mentioned in Chinese patent application CN110346790A "A non-contact vital sign detection method, device and system based on millimeter wave radar" which searches for the maximum peak value of a one-dimensional range image. This method often cannot guarantee the selection of the optimal range unit, that is, a range unit that simultaneously contains human breathing and heartbeat signals.

[0005] Secondly, pulse interference noise caused by local oscillator leakage or circuit interference can greatly affect the phase signal, thereby affecting the accurate estimation of respiratory and heart rate signals. Traditional methods use pulse noise interference detection and cubic spline interpolation to remove pulse noise, but the computational load is large and pulse interference noise cannot be completely removed.

[0006] Third, in order to improve the accuracy of vital sign detection, existing algorithms often use signals with long time windows and employ mode decomposition or wavelet decomposition to extract respiratory and heartbeat signals. However, these methods are often computationally complex and time-consuming. Moreover, frequency estimation for signals with long time windows can only obtain the average heart rate within that time window, which is not conducive to the rapid acquisition and observation of vital sign signals and has practical application defects. Summary of the Invention

[0007] In view of this, the present invention provides a rapid and accurate method for detecting vital signs based on millimeter-wave radar, which can quickly detect vital signs signals with high accuracy and low complexity.

[0008] The present invention provides a rapid and accurate method for detecting vital signs based on millimeter-wave radar, comprising:

[0009] Step 1: The radar transmits a frequency-modulated continuous wave signal toward the target; the received echo signal and the transmitted signal are mixed; the mixed signal is low-pass filtered and then orthogonally sampled to obtain a sampled complex signal;

[0010] Step 2: Perform a fast Fourier transform on the sampled complex signal in the fast time dimension to obtain the distance-time graph;

[0011] Step 3: Extract the energy variance in the slow time dimension of the sampled complex signal of each distance cell; extract the phase of the complex signal of the distance cell with the largest variance and perform phase unwrapping;

[0012] Step 4: After performing phase difference on the unwound signal, pulse interference removal is performed on the phase difference signal; then, based on the frequency band range of the breathing and heartbeat time domain signals, bandpass filtering is performed on the phase difference signals after pulse interference removal to obtain the breathing and heartbeat time domain signals.

[0013] Step 5: Perform a Fourier transform on the respiratory time-domain signal; the frequency corresponding to the maximum peak value is the respiratory rate.

[0014] Step 6: After adaptively filtering the heartbeat signal based on the respiratory rate, perform Fourier transform, and extract the heartbeat frequency of the heartbeat signal using the peak value method.

[0015] The better ones also include:

[0016] Step 7: Using a sliding window approach, Kalman filtering is used to track and estimate the heart rate, and the heart rate result is output.

[0017] Preferably, in step 1, the radar transmits signals at fixed time intervals.

[0018] Preferably, in step 3, the phase is obtained by arctangenting the complex signal of the distance unit with the maximum energy variance, and then the phase sequence is unwrapped.

[0019] A more preferred method for phase unwinding is as follows:

[0020]

[0021] Among them, phase i The phase value before unwinding. This refers to the new phase value after unwinding.

[0022] Preferably, in step 4, the phase difference signal is subjected to ten-point median filtering to remove pulse interference.

[0023] Preferably, in step 4, the filtered phase difference signal is processed by a 4th-order Butterworth bandpass filter and an 8th-order Butterworth bandpass filter to obtain the time-domain signals of respiration and heartbeat.

[0024] Beneficial effects:

[0025] (1) This invention first mixes the echo signal and the transmitted signal, and obtains a complex signal through cross-sampling. Then, it uses the variance of the slow-time dimension energy based on the complex signal to select the optimal distance unit without destroying the target phase signal, ensuring that the optimal distance unit simultaneously contains human respiratory and heartbeat signals. Simultaneously, before estimating the heart rate, this invention first eliminates the high-order harmonics of the respiratory signal, and then outputs the final heart rate result based on the amplitude, resulting in high accuracy in heart rate estimation. This invention reduces the requirements for signal processing and estimation time while ensuring detection accuracy, enabling the method to achieve high-precision, rapid heart rate estimation, adapting to detection requirements in various scenarios, and truly realizing rapid detection of vital signs.

[0026] (2) The present invention uses a median filtering method with appropriate points to eliminate pulse interference. It removes pulse interference noise caused by local oscillator leakage and circuit interference without affecting the breathing and heartbeat waveform. The calculation is small and the effect is significant.

[0027] (3) The present invention also uses the sliding window method to continuously estimate vital signs and uses the Kalman filter algorithm to track and estimate heart rate, thereby further improving the detection accuracy. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of an embodiment of the present invention;

[0029] Figure 2 This is a schematic diagram of a data testing and verification scenario according to an embodiment of the present invention;

[0030] Figure 3 This is a schematic diagram of the main flow of signal processing according to the method of the present invention;

[0031] Figure 4 This is a schematic diagram of the sliding window detection method according to an embodiment of the present invention;

[0032] Figure 5 This is a schematic diagram comparing the results of vital sign detection using 5s data with the results of ECG detection simultaneously, according to an embodiment of the present invention.

[0033] Figure 6 This is a schematic diagram comparing the results of vital sign detection using 8s data with the results of simultaneous ECG detection in an embodiment of the present invention.

[0034] Figure 7 This is a schematic diagram showing the results of vital sign detection using 5s data in an embodiment of the present invention, including a comparison of the method of using Kalman filtering and averaging the first 2s of data to replace abnormal data with the method of no processing, and a comparison of the results of ECG detection at the same time.

[0035] Figure 8 This is a schematic diagram showing the results of vital sign detection using 5s data in an embodiment of the present invention, including a comparison of the method of using Kalman filtering and averaging the first 4s of data to replace abnormal data with the method of no processing, and a comparison of the results of ECG detection at the same time.

[0036] Figure 9 This is a schematic diagram showing the results of vital sign detection using 5s data in an embodiment of the present invention, including a comparison of the method of using Kalman filtering and averaging the first 8s of data to replace abnormal data with the method of no processing, and a comparison of the results of ECG detection at the same time.

[0037] Figure 10 This is a schematic diagram showing the comparison between the detection results of 5-second data and the ECG results detected simultaneously, which is the result of the detection of vital signs using the distance cell selection method based on energy variance, phase variance and energy maximum.

[0038] Figure 11 This is a schematic diagram showing the comparison between the heart rate detection results and the simultaneously detected ECG results, which are obtained by using the phase variance before and after phase unwrapping to select distance cells in an embodiment of the present invention. Detailed Implementation

[0039] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0040] This invention provides a rapid and accurate method for detecting vital signs based on millimeter-wave radar.

[0041] like Figure 1 As shown, a millimeter-wave radar transmits radar signals and receives reflected echo signals from various targets in space. The received echo signals are then mixed with the transmitted signals using a mixer and orthogonally sampled to obtain I and Q intermediate frequency (IF) sampling signals. Range FFT is performed on each signal column to obtain a time-range image containing the target signal. The complex signal of each range cell is extracted, and its phase is calculated. After phase unwrapping, the optimal range cell is determined based on the phase variance in the slow time dimension. The phase of the optimal range cell is extracted, unwrapped, and then phase differential is performed to eliminate phase changes caused by the distance between the human body and the radar, thus obtaining accurate breathing and heartbeat changes. Next, a median filtering algorithm is used to filter the phase difference signal to eliminate impulse noise. Then, a bandpass filter is used to separate the time-domain signal containing respiratory and heartbeat signals. First, the respiratory signal frequency is estimated, and the respiratory signal frequency is obtained according to the peak-maximum method. Then, a notch filter is constructed using the respiratory signal frequency to remove respiratory harmonics from the heartbeat signal frequency band. After that, the heartbeat frequency is estimated again, and after peak detection, the x-coordinate of the point with the largest peak is selected as the heartbeat signal frequency. A sliding window method is used to continuously acquire the target's rapid vital signs, and a Kalman filter algorithm is used to track and estimate the heart rate results. This invention improves the accuracy of radar in rapidly detecting respiratory and heart rate by accurately selecting range cells, eliminating phase impulse noise, and using respiratory harmonic removal algorithms.

[0042] The present invention provides a rapid and accurate method for detecting vital signs based on millimeter-wave radar. The experimental scenario is as follows: Figure 2 As shown, the signal processing procedure is as follows: Figure 3 As shown, the specific steps are as follows:

[0043] Step 1: The frequency-modulated continuous wave radar transmits frequency-modulated continuous wave signals at fixed time intervals. This signal provides target range information and has the ability to simultaneously measure the vital signs of multiple targets. The transmitted frequency-modulated continuous wave signal S TX The expression for (t) is:

[0044]

[0045] Among them, A T It is the amplitude of the transmitted signal, f c T is the start frequency of the transmitted signal, t is the duration of the chirp signal, B is the bandwidth of the transmitted signal, and T is the base frequency of the transmitted signal. c The pulse duration is φ(t), the phase noise is k = B / T c is the frequency modulation slope of the chirp signal.

[0046] Step 2: The radar receives reflected echo signals from objects in space, including human bodies and other stationary clutter signals, and takes the echo signal S from one of the reflection points as an example. RX Taking (t) as an example, it can be expressed as:

[0047]

[0048] Where α is the reflection coefficient, which is related to factors such as the distance between the radar and the object, the object's RCS, and its material, t d =2d(t) / c, which represents the time delay of the echo signal of an object at a distance d(t) from the radar, where c is the speed of light.

[0049] Step 3: Mix the received echo signal and the transmitted signal, and then perform a low-pass filter on the mixed intermediate frequency (IF) signal to obtain the IF signal S. IF The expression for (t) is:

[0050]

[0051] Among them, A IF f is the amplitude of the intermediate frequency signal. b φ is the bandwidth of the intermediate frequency signal. b (t) represents the phase of the intermediate frequency signal.

[0052]

[0053] Because of t d Very small, The residual phase noise Δφ(t)=φ(t)-φ(t-2R / c) can be ignored.

[0054] The intermediate frequency signal is orthogonally sampled; the expression for the IQ signal y[n,m] after orthogonal sampling is:

[0055]

[0056] Where n is the number of sampling points in the fast time, and T f T is the ADC sampling rate, m is the number of sampling points in the slow time, and T is the sampling rate. s It is a slow sampling frequency.

[0057] Step 4: Combine the real and imaginary parts of the signal into a complex signal; perform a one-dimensional fast Fourier transform on the sampling matrix of each fast time dimension to obtain the target's distance-time map.

[0058] Step 5: Extract the complex signal from each distance cell and calculate its energy. The calculation formula is as follows:

[0059] energy i =I i 2+Q i 2

[0060] Among them, I i Q is the imaginary part of the complex signal. i This is the real part of the complex signal.

[0061] Then, the energy variance s of the complex signal of that distance cell along the slow time dimension is calculated. 2 (energy), the calculation formula is:

[0062]

[0063]

[0064] Where M is the number of frames in the chirp signal; energy i This represents the energy of a single sampling point after the Fourier transform.

[0065] Step 6: Obtain the optimal distance cell according to the maximum variance criterion; after selecting the optimal distance cell, calculate the phase of the complex signal of the optimal distance cell and perform phase unwrapping to obtain the correct phase change.

[0066] In this embodiment, the phase is obtained by performing arctangent calculation on the complex signal. The calculation formula is as follows:

[0067]

[0068] The phase unwrapping of the phase sequence is calculated using the following formula:

[0069]

[0070] in, This represents the new phase value after untangling.

[0071] Step 7: Perform phase difference analysis on the unwound signal to obtain the changes in respiration and heart rate signals over time. Phase difference analysis can eliminate the influence of the distance between the radar and the human body on the phase and allows for first-order differentiation of the phase signal, thereby enhancing the amplitude of the heartbeat signal. The phase difference formula is:

[0072] Δphase i+1 =phase i+1 -phase i i = 1, 2, ... N

[0073] in x(i) is the distance between the radar and the target, and Δx(i) is the speed of chest rise and fall caused by breathing and heartbeat.

[0074] Step 8: Filter the phase difference signal to remove pulse interference noise and improve the signal-to-noise ratio of the respiratory and heartbeat signals. Methods such as cubic spline interpolation and Savitzky-Golay can be used. This embodiment uses ten-point median filtering, which removes pulse interference while preserving the original phase signal information to the greatest extent possible, and requires less computation.

[0075] Specifically, the ten-point median filter calculation formula in this embodiment is as follows:

[0076]

[0077] Step 9: Perform bandpass filtering on the median-filtered signal. Use filters corresponding to the frequency bands of the respiratory and heartbeat time-domain signals respectively to obtain the respiratory and heartbeat time-domain signals. The frequency band of the respiratory filter is [0.1Hz, 0.5Hz], and the frequency band of the heartbeat signal filter is [0.8Hz, 2Hz]. This embodiment uses a 4th-order Butterworth bandpass filter and an 8th-order Butterworth bandpass filter to obtain the respiratory and heartbeat time-domain signals. IIR filters are characterized by high efficiency and simple design.

[0078] Step 10: Perform a 1024-point Fourier transform on the respiratory time-domain signal, find the point with the highest peak in the spectrum using the peak search method, extract the x-coordinate of the point, and convert it into respiratory frequency.

[0079] Step 11: Based on the respiratory rate estimated in Step 10, perform adaptive filtering on the obtained heartbeat time-domain signal to filter out high-order harmonics of respiration. Then, perform a 1024-point Fourier transform on the heartbeat time-domain signal, select the maximum peak value according to the peak search method, extract the x-coordinate of the point, and convert it into heartbeat frequency.

[0080] Step 12: Continuously acquire the target's rapid vital signs using a sliding window method. The sliding window length is 5 seconds, and the step size is 1 second. Use Kalman filtering to track and estimate the heart rate, and output the final heart rate. The Kalman filtering algorithm process is as follows:

[0081]

[0082]

[0083]

[0084]

[0085]

[0086] in, and Let represent the posterior state estimates at time k-1 and time k, respectively. P represents the prior state estimate at time k, which is an intermediate calculation result of the filter. It is the result predicted at time k based on the optimal estimate at time k-1. k-1 and P k Let the posterior estimated covariances at time k-1 and time k be represented respectively. Let z represent the prior estimate covariance at time k, A represent the state transition matrix (set to 1 here), and Q represent the process excitation noise covariance (set to 0.01 here). k K represents the spectral estimate of heart rate. k Here, represents the Kalman gain, H represents the transformation matrix from state variables to measurements, and R represents the measurement noise covariance matrix, which is set to 0.25 here.

[0087] The following example demonstrates how to quickly detect vital signs using radar receiving a signal from a human target at a distance of 0.8m from the radar. The specific steps are as follows:

[0088] A. A person sits 0.8m away from the radar and faces the radar directly. The radar operates at 60GHz, using a one-to-one transmission and one-to-reception method, with a fast time sampling rate of 2500ksps, 128 fast time sampling points, a frame interval of 50ms, a frequency modulation slope of 60MHz / us, an actual bandwidth of 3.6GHz, and a corresponding frequency resolution of 4.17cm.

[0089] B. Perform a range-dimensional FFT on the echo signal matrix, extract the complex signal of each range cell along the slow time dimension, calculate its energy, and calculate the energy variance of each range cell along the slow time dimension.

[0090] C. Obtain the optimal distance cell according to the maximum variance criterion, extract the imaginary and real part signals respectively, use the arctangent to calculate the phase, and unwrap the phase signal.

[0091] D. Perform phase difference on the phase unwrapped signal to eliminate the influence of the distance between the radar and the human body on the phase, and perform first-order derivative on the phase signal to enhance the amplitude of the heartbeat signal.

[0092] E. Perform ten-point median filtering on the phase difference signal to suppress impulse noise without affecting the breathing and heartbeat waveforms.

[0093] F. The corresponding breathing and heartbeat waveforms are obtained by using fourth-order and eighth-order bandpass filters, respectively.

[0094] G. First, perform a 1024-point FFT operation on the respiratory signal, select the respiratory frequency based on the peak value of the spectrum, extract the x-coordinate of the point, and convert it into the respiratory frequency.

[0095] H. Adaptive filtering is performed on the heartbeat signal based on the respiratory rate to remove high-order harmonics of respiration. Then, the frequency of the heartbeat signal is estimated, and the maximum peak value is selected according to the peak search method. The x-coordinate of the point is extracted and converted into the heartbeat frequency.

[0096] I. A sliding window method is used, with a sliding window length of 5s and a step size of 1s. Kalman filtering is used to track and estimate the heart rate, and the heart rate result is output.

[0097] Depend on Figure 5 and Figure 6 It can be seen that the method proposed in this invention achieves a high detection accuracy even with a measurement time of only 5 seconds, proving the effectiveness of the proposed method. Figure 7 , Figure 8 and Figure 9 It can be seen that the method of smoothing the heart rate results using Kalman filtering achieves better heart rate estimation results.

[0098] Depend on Figure 10 and Figure 11 It can be seen that the method proposed in this invention can be used to select distance units more accurately, thereby achieving a more robust detection accuracy.

[0099] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A rapid and accurate method for detecting vital signs based on millimeter-wave radar, characterized in that, include: Step 1: The radar transmits a frequency-modulated continuous wave signal toward the target; The received echo signal and the transmitted signal are mixed. The mixed signal is low-pass filtered and then quadrature sampled to obtain a sampled complex signal. Step 2: Perform a fast Fourier transform on the sampled complex signal in the fast time dimension to obtain the distance-time graph; Step 3: Extract the energy variance in the slow time dimension of the sampled complex signal for each distance cell, specifically as follows: Extract the complex signal from each distance cell and calculate its energy using the following formula: in, For the imaginary part of the complex signal, This represents the real part of the complex signal; Then calculate the energy variance of the complex signal of that distance cell along the slow time dimension. The calculation formula is: Where M is the number of frames in the chirp signal; This represents the energy of a single sampling point after the Fourier transform. Extract the phase of the complex signal from the unit with the largest variance and perform phase unwrapping; Step 4: After performing phase difference on the unwound signal, perform ten-point median filtering on the phase difference signal to remove pulse interference; then, based on the frequency band range of the breathing and heartbeat time domain signals, perform bandpass filtering on the phase difference signals after removing pulse interference to obtain the breathing and heartbeat time domain signals. Step 5: Perform a Fourier transform on the respiratory time-domain signal; the frequency corresponding to the maximum peak value is the respiratory rate. Step 6: After adaptively filtering the heartbeat signal based on the respiratory rate, perform Fourier transform, and extract the heartbeat frequency of the heartbeat signal using the peak method for the maximum peak value. Step 7: Using a sliding window approach, Kalman filtering is used to track and estimate the heart rate, and the heart rate result is output.

2. The method as described in claim 1, characterized in that, In step 1, the radar transmits signals at fixed time intervals.

3. The method as described in claim 1, characterized in that, In step 3, the phase is obtained by arctangenting the complex signal of the distance unit with the maximum energy variance, and then the phase sequence is unwrapped.

4. The method as described in claim 3, characterized in that, Phase unwinding specifically involves: in, The phase value before unwinding. This refers to the new phase value after unwinding.

5. The method as described in claim 1, characterized in that, In step 4, the time-domain signals of respiration and heartbeat are obtained by applying a 4th-order Butterworth bandpass filter and an 8th-order Butterworth bandpass filter to the filtered phase difference signal, respectively.

Images