Vehicle-mounted vital sign monitoring millimeter wave radar signal processing method

Abstract

This invention discloses a millimeter-wave radar signal processing method for vehicle-mounted vital sign monitoring, belonging to the field of radar signal processing. The method simultaneously acquires millimeter-wave radar echo signals and vibration sensor signals to generate an equivalent vibration reference signal with the same vibration response characteristics as the radar. Based on the frequency domain analysis of this reference signal, the vehicle's operating conditions are identified, and the frequency domain interference detection threshold, mask attenuation intensity, and frequency band coverage are dynamically adjusted to construct a weighted frequency domain mask. The mask is multiplied by the radar echo spectrum to suppress interference frequencies, and residual vibrations overlapping with the vital sign frequency band are vector-cancelled using a complex domain interference cancellation model to reconstruct a high signal-to-noise ratio time-domain signal. Furthermore, chest cavity micro-motion information is extracted, and respiratory and heartbeat signals are separated and their frequencies estimated. This invention improves the stability and robustness of vital sign monitoring in complex vehicle environments.

Description

Technical Field

[0001] This invention relates to the fields of radar signal processing and intelligent vehicle technology; in particular, it relates to a method for processing millimeter-wave radar signals for vehicle-mounted vital signs monitoring. Background Technology

[0002] In the field of in-vehicle intelligent cockpit health monitoring, frequency modulated continuous wave (FMCW) millimeter-wave radar has become a core technology solution for monitoring driver vital signs due to its advantages such as non-contact sensing, high precision, and strong anti-interference capabilities. It can achieve continuous and imperceptible detection of heartbeat and breathing signals without interfering with driving behavior.

[0003] However, existing millimeter-wave radar vital sign monitoring technology faces severe challenges in real-world driving scenarios. Low-frequency vibrations (0.1-2Hz) caused by vehicle engine operation and uneven road surfaces highly overlap with respiratory signals (0.1-0.5Hz) and heartbeat signals (0.8-2Hz) in the frequency domain, severely masking weak physiological signals. Current technologies largely rely on passive compensation and correction using the radar's own signals, lacking a systematic modeling and suppression mechanism for external vibration interference, making it difficult to fundamentally solve the frequency overlap problem. In vehicle environments where vibration intensity and frequency dynamically change, the monitoring accuracy and robustness are insufficient, failing to meet practical application requirements. Summary of the Invention

[0004] Based on the above analysis, the present invention aims to disclose a millimeter-wave radar signal processing method for vehicle-mounted vital signs monitoring; to solve the problems of low monitoring accuracy and poor robustness caused by the overlap of vibration interference and physiological signal frequency domain in real driving scenarios in the prior art.

[0005] This invention discloses a method for processing millimeter-wave radar signals for vehicle-mounted vital signs monitoring, comprising the following steps: S1. Simultaneously acquire millimeter-wave radar echo signals reflected by human targets inside the car cabin, as well as vibration sensor signals used to characterize the vibration interference experienced by the radar, and transform the vibration sensor signals into equivalent vibration reference signals with the same vibration response characteristics as the radar. S2. Perform frequency domain analysis on the equivalent vibration reference signal to identify the vehicle operating conditions, and dynamically adjust the frequency domain interference detection threshold, mask attenuation intensity and frequency band coverage to construct a weighted frequency domain mask to suppress interference. S3. Perform a dot product operation between the weighted frequency domain mask and the slow time spectrum of the radar echo to suppress interference frequencies. For vibration interference frequencies that overlap with the frequency band of vital signs, construct a complex domain interference cancellation model using the frequency domain amplitude and phase characteristics of the equivalent vibration reference signal to perform vector cancellation, and reconstruct the time domain radar signal through inverse Fourier transform. S4. Perform phase unwrapping and differential processing on the reconstructed time-domain radar signal to extract chest cavity micro-motion information. Separate the respiratory signal and heartbeat signal through bandpass filtering and estimate their frequencies.

[0006] Furthermore, S1 includes: S101. Simultaneously collect millimeter-wave radar echo signals reflected by the human body inside the car cabin, as well as vibration sensor signals fixed on the radar mounting base to characterize the vibration interference experienced by the radar. S102. The multi-axis signal output by the vibration sensor is transformed by a pre-calibrated rotation matrix and combined with the unit vector projection of the radar beam main lobe direction to generate an equivalent vibration reference signal with the same vibration response characteristics as the radar. S103. After performing DC removal preprocessing on the radar echo signal and the original equivalent vibration reference signal, a radar signal with zero mean and an equivalent vibration reference signal are obtained.

[0007] Furthermore, S2 includes: S201. Perform a fast Fourier transform on the zero-mean equivalent vibration reference signal to obtain its frequency domain representation, and extract the spectral entropy and impulse factor; wherein, the spectral entropy is used to measure the dispersion of vibration energy in the frequency domain, and the impulse factor is used to characterize the transient impact intensity during vehicle operation. S202. Based on the spectral entropy and pulse factor, adaptively identify the vehicle operating conditions and classify the vehicle operation into idling conditions, uniform and smooth conditions, or complex bumpy conditions. S203. Select frequency points whose amplitude exceeds the dynamic detection threshold from the spectrum to form a set of strong interference frequencies; The dynamic detection threshold is the sum of the product of the mean of the vibration spectrum amplitude and the adaptive factor and standard deviation; the value of the adaptive factor is dynamically determined based on the spectral entropy. S204. Construct a weighted frequency domain mask for the set of strong interference frequencies to suppress the slow-time spectrum corresponding to each range cell of the radar echo signal in the frequency domain.

[0008] Furthermore, the extracted spectral entropy: ; Extracted pulse factor: ; in, Based on spectral amplitude Define the normalized spectral energy distribution; ; ; Indicates frequency; The equivalent vibration reference signal with zero mean; This is the slow time sampling sequence number. This represents the number of sampling points.

[0009] Furthermore, the methods for classifying vehicle operating conditions include: Idle operating condition: When the following conditions are met Furthermore, when the main energy of the spectrum is concentrated in a fixed low-frequency harmonic region, the vehicle is determined to be in an idling condition. The first spectral entropy threshold; Uniform speed leveling condition: when the following conditions are met And the spectral energy fluctuation coefficient At that time, it was determined that the vehicle was in a uniform speed and flat working condition. The time-domain vibration intensity threshold, The frequency spectrum energy fluctuation coefficient, The threshold for spectral energy fluctuation; Complex and bumpy operating conditions: when the following conditions are met and At that time, it was determined that the vehicle was in a complex and bumpy operating condition. The pulse factor threshold; This is the second spectral entropy threshold.

[0010] Furthermore, adaptive factor According to spectral entropy Dynamically updated: ; in, This is the minimum value of the spectral entropy. The maximum value of the spectral entropy. This is the minimum value of the adaptive factor. This represents the maximum value of the adaptive factor.

[0011] Furthermore, targeting sets of strong interference frequencies Construct a weighted frequency domain mask with smooth edge features. : ;in, It is the attenuation factor; Attenuation factor According to pulse factor Dynamic adjustment; ; in, This represents the maximum value of the attenuation factor. This is the attenuation adjustment coefficient determined based on the statistical characteristics of the vibration spectrum.

[0012] Furthermore, S3 includes: S301. Perform a fast Fourier transform on the radar echo signal after DC removal to obtain the slow-time radar echo spectrum of each range cell. Perform a dot multiplication operation between the weighted frequency domain mask and the radar echo spectrum to obtain the first interference suppression spectrum that initially attenuates strong interference frequencies. S302. For the residual vibration interference frequencies that overlap with the vital signs frequency band contained in the first interference suppression spectrum, vector cancellation is performed using a complex domain interference cancellation model to obtain the second interference suppression spectrum. The complex domain interference cancellation model uses complex weighting coefficients to quantify the amplitude-phase mapping relationship between the equivalent vibration reference signal and the vibration interference component in the radar echo. By weighting the equivalent vibration reference signal according to the complex weighting coefficients and then subtracting it from the corresponding interference component in the first interference suppression spectrum, amplitude-phase joint cancellation is performed. S303. Perform an inverse slow-time Fourier transform on the second interference suppression spectrum, and reassemble the high signal-to-noise ratio slow-time signals recovered from each range cell into two-dimensional vibration-disrupted radar data with range cell and slow time as the dimensions according to the range cell index.

[0013] Furthermore, complex weighting coefficients The expression is: ; In the formula, This is the first interference suppression spectrum. The spectrum of the vibration reference signal with zero mean The conjugate term of the spectrum of the vibration reference signal with zero mean; Indicates frequency; This is the regularization factor.

[0014] Furthermore, S4 includes: The cumulative distribution of reflected energy of each range cell in slow time is calculated based on the vibration-disrupted two-dimensional radar data. The range cell of the chest cavity target is locked based on the maximum energy criterion combined with the constraint of the continuity of human body micro-motion. S402. Extract the slow-time complex signal phase sequence from the target distance unit, and perform inter-frame difference after unwinding to eliminate periodic jumps, thereby extracting thoracic cavity micro-motion information while suppressing background low-frequency drift. S403. Use a finite impulse response bandpass filter to separate the aforementioned thoracic cavity micro-motion signal into respiratory and cardiac frequency bands in the frequency domain. S404. The Welch method is used to estimate the power spectral density of the separated respiratory and heartbeat signals, and the maximum spectral peak frequency is searched within their respective frequency bands and converted into respiratory rate and heart rate.

[0015] Compared with traditional methods, the present invention has the following technical advantages: This invention addresses the technical bottleneck of frequency domain overlap between vibration interference and vital sign signals in vehicle-mounted environments by proposing a millimeter-wave radar signal processing method for vehicle-mounted vital sign monitoring that integrates vibration spatial projection, adaptive operating condition identification, and dynamic frequency domain suppression. By introducing an equivalent vibration reference signal aligned with the radar beam direction and combining spectral entropy, impulse factor, and spectral energy fluctuation characteristics, adaptive vehicle operating condition identification is achieved. The system can dynamically adjust the frequency domain detection threshold, frequency domain mask attenuation intensity, and frequency band coverage according to different vibration states, achieving precise suppression of vibration interference under various operating conditions such as idling, constant speed, and complex bumps. This improves the stability and robustness of respiratory rate and heart rate monitoring in complex vehicle-mounted environments. Attached Figure Description

[0016] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts. Figure 1 This is a flowchart of the vehicle-mounted vital signs monitoring millimeter-wave radar signal processing method in an embodiment of the present invention. Detailed Implementation

[0017] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and, together with the embodiments of the present invention, serve to illustrate the principles of the present invention.

[0018] One embodiment of the present invention discloses a method for processing millimeter-wave radar signals for vehicle-mounted vital signs monitoring, such as... Figure 1 As shown, it includes: S1. Simultaneously acquire millimeter-wave radar echo signals reflected by human targets inside the car cabin, as well as vibration sensor signals used to characterize the vibration interference experienced by the radar, and transform the vibration sensor signals into equivalent vibration reference signals with the same vibration response characteristics as the radar. S2. Perform frequency domain analysis on the equivalent vibration reference signal to identify the vehicle operating conditions, and dynamically adjust the frequency domain interference detection threshold, mask attenuation intensity and frequency band coverage to construct a weighted frequency domain mask to suppress interference. S3. Perform a dot product operation between the weighted frequency domain mask and the radar slow-time echo spectrum to suppress interference frequencies. For vibration interference frequencies that overlap with the vital signs frequency band, construct a complex domain interference cancellation model using the frequency domain amplitude and phase characteristics of the equivalent vibration reference signal to perform vector cancellation, and reconstruct the time domain radar signal through inverse Fourier transform. S4. Perform phase unwrapping and differential processing on the reconstructed time-domain radar signal to extract chest cavity micro-motion information. Separate the respiratory signal and heartbeat signal through bandpass filtering and estimate their frequencies.

[0019] Specifically, S1 includes: S101 synchronously acquires millimeter-wave radar echo signals reflected by the human body inside the car cabin, as well as vibration sensor signals fixed on the radar mounting base to characterize the vibration interference experienced by the radar; and achieves dual-channel sampling time alignment through a unified clock source, laying the timing foundation for subsequent signal fusion.

[0020] This embodiment employs a millimeter-wave radar system in conjunction with a vibration sensor. The millimeter-wave radar operates in the 77–81 GHz frequency band and uses a multiple transmit multiple receive (MIMO) antenna structure, with the main lobe of the antenna pointing towards the driver's chest cavity to acquire the micro-motion echo signals from the chest cavity caused by the driver's breathing and heartbeat.

[0021] The vibration sensor is a triaxial vibration sensor, which is fixed to a mounting bracket that has a rigid connection with the millimeter-wave radar. There is no relative displacement between the sensor and the radar, ensuring that both are subjected to consistent vibration excitation during vehicle operation.

[0022] Synchronous acquisition is achieved through a unified clock source; the radar chirp trigger signal is also used as the sampling enable signal of the vibration sensor, or a timestamp alignment mechanism is adopted to control the sampling time deviation between the radar echo signal and the vibration sensor signal within 1 ms, so as to ensure the timing consistency of subsequent interference cancellation.

[0023] S102. The multi-axis signal output by the vibration sensor is transformed by a pre-calibrated rotation matrix and combined with the unit vector projection of the radar beam main lobe direction to generate an equivalent vibration reference signal with the same vibration response characteristics as the radar, so that its frequency and phase characteristics are consistent with the vibration interference actually coupled to the radar.

[0024] Vehicle vibrations are directional in space, while millimeter-wave radar is only sensitive to displacement changes along the main axis of its antenna beam. Therefore, the multi-axis signal output by the triaxial vibration sensor needs to be transformed into an equivalent vibration reference signal with the same vibration response characteristics as the radar.

[0025] Let the original signal output by the triaxial vibration sensor be: ; in, , , These are the three-axis vibration components in the sensor's own coordinate system. This is the slow time sampling sequence number.

[0026] The rotation matrix between the radar coordinate system and the vibration sensor coordinate system is obtained through pre-calibration. The rotation matrix is ​​calibrated using a standard vibration excitation source during the factory or vehicle installation phase, and is used to compensate for the installation angle error between the radar and the vibration sensor.

[0027] The triaxial vibration signal is transformed into spatial coordinates using a rotation matrix, and then projected using the radar beam main lobe direction vector to synthesize an equivalent vibration reference signal aligned with the radar beam direction; its expression is: ; in, This represents the unit vector of the direction of the radar beam's main lobe in the radar coordinate system. This process rotates the triaxial vibration signal from the sensor coordinate system to the radar coordinate system, and then projects it onto the radar beam direction. After this transformation, It characterizes the equivalent vibration displacement (or velocity / acceleration) along the radar-sensitive direction. Its amplitude-frequency and phase-frequency characteristics have the same vibration response characteristics as the vibration interference actually coupled to the radar, and can therefore be used as a reference for subsequent frequency domain interference cancellation.

[0028] S103. After performing DC removal preprocessing on the radar echo signal and the original equivalent vibration reference signal, a radar signal with zero mean and an equivalent vibration reference signal are obtained.

[0029] For equivalent vibration reference signal Radar raw echo signal The formula for DC removal is: ; ; in, This represents the number of sampling points.

[0030] In this embodiment, the time window length is set to 25 seconds, corresponding to 500 sampling points (sampling rate 20 Hz). The zero-mean radar signal and the equivalent vibration reference signal are denoted as follows: and This serves as the input for subsequent vehicle condition identification and frequency domain mask construction.

[0031] Specifically, S2 includes: S201. Perform a fast Fourier transform on the zero-mean equivalent vibration reference signal to obtain its frequency domain representation, and extract the spectral entropy and impulse factor; wherein, the spectral entropy is used to measure the dispersion of vibration energy in the frequency domain, and the impulse factor is used to characterize the transient impact intensity during vehicle operation.

[0032] For zero-mean equivalent vibration reference signal Performing a Fast Fourier Transform yields its frequency domain representation: ; in, Indicates frequency; Based on spectral amplitude Define normalized spectral energy distribution : ; Further calculation of spectral entropy: ; Spectral entropy Used to measure the dispersion of vibrational energy in the frequency domain: When vibration energy is concentrated at a few fixed frequencies (such as engine idling harmonics), The value is low; When vibrations exhibit a wide-band random distribution (such as in complex, bumpy road conditions), The value is relatively high.

[0033] To characterize the transient impact intensity during vehicle operation, an impulse factor is defined. : ; Among them: larger pulse factor The value indicates the presence of a significant road impact or transient vibration event; a small impulse factor. The value indicates that the vibration is relatively stable.

[0034] S202. Based on the spectral entropy and pulse factor, adaptively identify the vehicle operating conditions and classify the vehicle operation into idling conditions, uniform and smooth conditions, or complex bumpy conditions; in order to determine the subsequent interference suppression processing strategy.

[0035] This embodiment is based on spectral entropy. With pulse factor Adaptive identification of vehicle operating conditions is performed, and the conditions are categorized into the following three types: (1) Idle condition: When the following conditions are met Furthermore, when the main energy of the spectrum is concentrated in a fixed low-frequency harmonic region, the vehicle is determined to be in an idling condition. The first spectral entropy threshold; Under this operating condition, the vibration mainly originates from the periodic operation of the engine, with a stable frequency spectrum and a clear harmonic structure.

[0036] (2) Uniform speed leveling condition: when the following conditions are met And the spectral energy fluctuation coefficient At that time, it was determined that the vehicle was in a uniform speed and flat working condition; in, The time-domain vibration intensity threshold, The frequency spectrum energy fluctuation coefficient, The threshold for spectral energy fluctuation; Spectral energy fluctuation coefficient Used to reflect the time-varying severity of vibration; it is the ratio of the standard deviation to the mean of the energy of multiple consecutive frames of spectrum.

[0037] Under this condition, the road surface excitation is weak, the overall vibration level is low, and the signal-to-noise ratio of vital signs is high.

[0038] (3) Complex bumpy working conditions: when the following conditions are met and At that time, it was determined that the vehicle was in a complex and bumpy operating condition. The pulse factor threshold; The second spectral entropy threshold; Under this operating condition, the vibration exhibits a distinct broadband random distribution, accompanied by transient impact characteristics; The above threshold parameters , , , as well as Adjustments can be made during the factory calibration stage, depending on the vehicle model, radar installation location, and sensor sensitivity.

[0039] S203. Select frequency points whose amplitude exceeds the dynamic detection threshold from the spectrum to form a set of strong interference frequencies; the dynamic detection threshold is the sum of the product of the mean of the vibration spectrum amplitude and the adaptive factor and standard deviation; the value of the adaptive factor is dynamically determined according to the spectrum entropy.

[0040] For vibration spectrum amplitude Calculate the mean with standard deviation Construct dynamic detection thresholds: ; in, As an adaptive factor, based on the spectral entropy Dynamically updated: ; in, This is the minimum value of the spectral entropy. The maximum value of the spectral entropy. This is the minimum value of the adaptive factor. This represents the maximum value of the adaptive factor. In this embodiment, ; ; When the spectral entropy is low, it indicates that the vibration spectrum is more concentrated. In this case, reducing k will improve the sensitivity of interference detection. When the spectral entropy is high, it indicates that the vibration spectrum is more dispersed. In this case, increasing k can help avoid false detections caused by broadband random noise.

[0041] When the following conditions are met: ; When the frequency f is identified as a strong interference frequency, it is included in the set of strong interference frequencies. .

[0042] Furthermore, this embodiment employs spectral energy constraints for auxiliary screening: ; in, The threshold for the proportion of spectrum energy. In this embodiment, the following is taken That is, only strong interference frequencies with a cumulative energy of more than 80% of the total spectrum energy are retained.

[0043] S204. A weighted frequency domain mask is constructed for the set of strong interference frequencies to suppress the slow-time spectrum corresponding to each range unit of the radar echo signal in the frequency domain; the weighted frequency domain mask has smooth edge characteristics, its attenuation factor is dynamically adjusted according to the pulse factor, and its neighborhood bandwidth is adaptively expanded according to the spectral energy fluctuation coefficient.

[0044] For sets of strong interference frequencies Construct a weighted frequency domain mask with smooth edge features. : ; in, As the attenuation factor, in this embodiment: ; Furthermore, attenuation factor Dynamically adjusted based on the pulse factor; ; This represents the maximum value of the attenuation factor. The attenuation adjustment coefficient is determined based on the statistical characteristics of the vibration spectrum. In this embodiment: ; .

[0045] When vehicle vibration and impact increase, reduce Increase vibration suppression strength; when vibration is relatively stable, moderately increase... To avoid excessive attenuation leading to loss of vital signs.

[0046] To adapt to broadband random vibration interference under complex and bumpy operating conditions, the frequency domain mask is extended from single-frequency point suppression to neighborhood frequency band suppression. For Its neighborhood bandwidth : ; in, Based on bandwidth, For expansion coefficient, The spectral energy fluctuation coefficient; Within the neighboring frequency band The mask values ​​are applied smoothly from the center frequency to the edge using a transition function (such as raised cosine roll-off or linear gradient). Gradually reduce to 1 to avoid the Gibbs effect introduced by spectral abrupt changes, thereby protecting the integrity of the vital signs spectral structure while ensuring strong interference suppression.

[0047] S3 includes: S301. Perform a fast Fourier transform on the radar echo signal after DC removal to obtain the slow-time radar echo spectrum of each range cell. Perform a dot multiplication operation between the weighted frequency domain mask and the radar echo spectrum to obtain the first interference suppression spectrum that initially attenuates strong interference frequencies.

[0048] The Fast Fourier Transform includes: Radar echo signal after DC removal Perform a range-dimensional Fast Fourier Transform (FFT) to obtain the slow-time complex signal sequence for each range cell; then perform a slow-time FFT on the slow-time complex signal sequence for each range cell to obtain the slow-time radar echo spectrum for each range cell. ; The weighted frequency domain mask generated in step S2 With radar echo spectrum Performing dot multiplication achieves initial attenuation of strong interference frequencies, yielding the first interference suppression spectrum. : ; This operation, within the frequency domain, assigns the located strong interference frequency components and their neighboring frequency bands to attenuation factors. Suppressing interference while preserving the original spectral structure of undisturbed frequency bands and vital sign frequency bands.

[0049] S302. For the residual vibration interference frequencies that overlap with the vital signs frequency band contained in the first interference suppression spectrum, vector cancellation is performed using a complex domain interference cancellation model to obtain the second interference suppression spectrum.

[0050] For sets of strong interference frequencies Frequency components overlap with those in the vital signs frequency range (respiratory signals 0.1–0.5 Hz, heartbeat signals 0.8–2.5 Hz). Amplitude attenuation using frequency domain masks alone is insufficient to completely eliminate vibration interference while fully preserving weak physiological signals. Therefore, a frequency domain representation of an equivalent vibration reference signal is introduced. A complex domain interference cancellation model is constructed, and vector cancellation is performed by aligning the amplitude and phase: , ; in, This represents a subset of vibration interference frequencies that overlap with the frequency bands of vital signs. The complex weighting coefficient is used to characterize the amplitude-phase mapping relationship of vibration interference to radar echo, and its expression is: ; In the formula, It is the conjugate term of the vibration reference signal spectrum; This is a regularization factor used to avoid numerical instability when the denominator approaches zero. In this embodiment, it is taken as... .

[0051] This complex domain cancellation model utilizes the phase prior of the vibration reference signal to vector-align and subtract the vibration interference component in the radar echo from the equivalent vibration reference signal within the overlapping frequency band. This achieves directional cancellation of interference energy in the frequency domain, rather than simply zeroing the amplitude, significantly reducing damage to the spectral structure of vital signs.

[0052] For strong interference frequencies that do not fall into the overlapping region (i.e.) middle The complement of the set, and the non-strong interference frequencies, are directly processed using the masking results: , .

[0053] S303. Perform an inverse slow-time Fourier transform on the second interference suppression spectrum, and reassemble the high signal-to-noise ratio slow-time signals recovered from each range cell into two-dimensional vibration-disrupted radar data with range cell and slow time as dimensions, as the reconstructed time-domain radar signal.

[0054] The spectrum after coarse suppression by frequency domain masking and fine suppression by complex domain Perform an inverse fast Fourier transform to recover the high signal-to-noise ratio slow-time signal corresponding to each distance cell; By recombining the slow-time signals of each range cell according to the range cell index, vibration-free two-dimensional radar data is obtained. , represented as a distance unit (Distance-oriented index) and slow time A two-dimensional function of (frame number): ; in, For distance cell index, Indicates the first The second interference suppression spectrum of a distance unit.

[0055] In the reconstructed two-dimensional radar data, random disturbances caused by vehicle vibration are effectively suppressed in the slow time dimension of each range cell, while the phase modulation information of chest cavity micro-movements caused by breathing and heartbeat is preserved.

[0056] S4 includes: S401: Calculate the cumulative distribution of reflected energy of each range cell in slow time on the two-dimensional radar data to remove vibration interference, and lock the range cell of the chest cavity target based on the maximum energy criterion and the constraint of the continuity of human body micro-motion; establish the spatial sampling position for subsequent phase extraction.

[0057] Calculate the reflected energy distribution for each distance cell: ; in, For distance cell index, For slow time frame number ( =500).

[0058] Select the range cell with the highest energy as the target range cell. The strongest reflex point corresponding to the chest cavity: .

[0059] To further eliminate false targets (such as stationary strong reflectors like seats and steering wheels), the continuity constraint of the target motion trajectory is combined: if the reflection energy fluctuation of a certain distance unit matches the micro-motion characteristics of the human chest cavity (amplitude on the order of approximately 0.1–0.5 mm) within multiple consecutive frames, it is determined to be a valid target distance unit.

[0060] S402. Extract the slow-time complex signal phase sequence from the target distance unit, and perform inter-frame difference after unwinding to eliminate periodic jumps, thereby extracting thoracic cavity micro-motion information while suppressing low-frequency background drift.

[0061] Extract target range unit Complex signals varying with slow time frames ,in For amplitude information, This is phase information.

[0062] phase The untangling process is performed to eliminate periodic jumps, resulting in the untangled phase. : ; in, For integers, satisfying .

[0063] Chest cavity micro-motion information is extracted by inter-frame phase difference operation, and the differential phase signal is obtained by suppressing low-frequency background trends (such as slow swaying of the driver's body). : ; The differential operation is equivalent to high-pass filtering, which enhances the high-frequency micro-motion components caused by breathing and heartbeat.

[0064] S403. The aforementioned thoracic micro-motion signal is separated into respiratory and cardiac frequency bands using a finite impulse response bandpass filter.

[0065] A finite impulse response (FIR) bandpass filter is used for the phase difference signal. Perform frequency domain separation to obtain respiratory signals. and heartbeat signals ; The passband for the respiratory signal is set to 0.1–0.5 Hz, and the passband for the heartbeat signal is set to 0.8–2.5 Hz. The filter order is designed to be 128 based on the sampling rate (20 Hz in this embodiment) to ensure linear phase characteristics and avoid waveform distortion. ; .

[0066] S404. The Welch method was used to separate the respiratory signals. With heartbeat signals Power spectral density estimation is performed, and the maximum spectral peak frequency is searched within each frequency band and converted into respiratory rate and heart rate.

[0067] The Welch method was used to estimate the power spectral density (PSD) of the separated signals, with a window length of 256 and an overlap of 50%. Respiratory signal power spectrum estimation: ; Heartbeat signal power spectrum estimation: ; Find the maximum spectral peak frequency within the corresponding frequency band and convert it to obtain respiratory rate (RR) and heart rate (HR): ; ; in, The unit is "times / minute". The unit is "bpm (bpm)".

[0068] In summary, the millimeter-wave radar signal processing method for vehicle-mounted vital signs monitoring disclosed in this embodiment introduces an equivalent vibration reference signal consistent with the radar beam direction, and combines spectral entropy, pulse factor, and spectral energy fluctuation characteristics to achieve adaptive identification of vehicle operating conditions. The system can dynamically adjust the frequency domain detection threshold, frequency domain mask attenuation intensity, and frequency band coverage according to different vibration states, thereby achieving precise suppression of vibration interference under different operating conditions such as idling, constant speed, and complex bumps, thus improving the stability and robustness of respiratory rate and heart rate monitoring in complex vehicle environments.

[0069] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for processing millimeter-wave radar signals for vehicle-mounted vital signs monitoring, characterized in that, Includes the following steps: S1. Simultaneously acquire millimeter-wave radar echo signals reflected by human targets inside the car cabin, as well as vibration sensor signals used to characterize the vibration interference experienced by the radar, and transform the vibration sensor signals into equivalent vibration reference signals with the same vibration response characteristics as the radar. S2. Perform frequency domain analysis on the equivalent vibration reference signal to identify the vehicle operating conditions, and dynamically adjust the frequency domain interference detection threshold, mask attenuation intensity and frequency band coverage to construct a weighted frequency domain mask to suppress interference. S3. Perform a dot product operation between the weighted frequency domain mask and the slow time spectrum of the radar echo to suppress interference frequencies. For vibration interference frequencies that overlap with the frequency band of vital signs, construct a complex domain interference cancellation model using the frequency domain amplitude and phase characteristics of the equivalent vibration reference signal to perform vector cancellation, and reconstruct the time domain radar signal through inverse Fourier transform. S4. Perform phase unwrapping and differential processing on the reconstructed time-domain radar signal to extract chest cavity micro-motion information. Separate the respiratory signal and heartbeat signal through bandpass filtering and estimate their frequencies.

2. The method for processing millimeter-wave radar signals for vehicle-mounted vital signs monitoring according to claim 1, characterized in that, S1 includes: S101. Simultaneously collect millimeter-wave radar echo signals reflected by the human body inside the car cabin, as well as vibration sensor signals fixed on the radar mounting base to characterize the vibration interference experienced by the radar. S102. The multi-axis signal output by the vibration sensor is transformed by a pre-calibrated rotation matrix and combined with the unit vector projection of the radar beam main lobe direction to generate an equivalent vibration reference signal with the same vibration response characteristics as the radar. S103. After performing DC removal preprocessing on the radar echo signal and the original equivalent vibration reference signal, a radar signal with zero mean and an equivalent vibration reference signal are obtained.

3. The method for processing millimeter-wave radar signals for vehicle-mounted vital signs monitoring according to claim 2, characterized in that, S2 includes: S201. Perform a fast Fourier transform on the zero-mean equivalent vibration reference signal to obtain its frequency domain representation, and extract the spectral entropy and impulse factor; wherein, the spectral entropy is used to measure the dispersion of vibration energy in the frequency domain, and the impulse factor is used to characterize the transient impact intensity during vehicle operation. S202. Based on the spectral entropy and pulse factor, adaptively identify the vehicle operating conditions and classify the vehicle operation into idling conditions, uniform and smooth conditions, or complex bumpy conditions. S203. Select frequency points whose amplitude exceeds the dynamic detection threshold from the spectrum to form a set of strong interference frequencies; The dynamic detection threshold is the sum of the product of the mean of the vibration spectrum amplitude and the adaptive factor and standard deviation; the value of the adaptive factor is dynamically determined based on the spectral entropy. S204. Construct a weighted frequency domain mask for the set of strong interference frequencies to suppress the slow-time spectrum corresponding to each range cell of the radar echo signal in the frequency domain.

4. The method for processing millimeter-wave radar signals for vehicle-mounted vital signs monitoring according to claim 3, characterized in that, Extracted spectral entropy: ; Extracted pulse factor: ; in, Based on spectral amplitude Define the normalized spectral energy distribution; ； ; Indicates frequency; The equivalent vibration reference signal with zero mean; This is the slow time sampling sequence number. This represents the number of sampling points.

5. The method for processing millimeter-wave radar signals for vehicle-mounted vital signs monitoring according to claim 4, characterized in that, The methods for classifying vehicle operating conditions include: Idle operating condition: When the following conditions are met Furthermore, when the main energy of the spectrum is concentrated in a fixed low-frequency harmonic region, the vehicle is determined to be in an idling condition. The first spectral entropy threshold; Uniform speed leveling condition: when the following conditions are met And the spectral energy fluctuation coefficient At that time, it was determined that the vehicle was in a uniform speed and flat working condition. The time-domain vibration intensity threshold, The frequency spectrum energy fluctuation coefficient. The threshold for spectral energy fluctuation; Complex and bumpy operating conditions: when the following conditions are met and At that time, it was determined that the vehicle was in a complex and bumpy operating condition. The pulse factor threshold; This is the second spectral entropy threshold.

6. The method for processing millimeter-wave radar signals for vehicle-mounted vital signs monitoring according to claim 4, characterized in that, Adaptive factor According to spectral entropy Dynamically updated: ； in, This is the minimum value of the spectral entropy. The maximum value of the spectral entropy. This is the minimum value of the adaptive factor. This represents the maximum value of the adaptive factor.

7. The method for processing millimeter-wave radar signals for vehicle-mounted vital signs monitoring according to claim 4, characterized in that, For sets of strong interference frequencies Construct a weighted frequency domain mask with smooth edge features. : ;in, It is the attenuation factor; Attenuation factor According to pulse factor Dynamic adjustment; ； in, This represents the maximum value of the attenuation factor. This is the attenuation adjustment coefficient determined based on the statistical characteristics of the vibration spectrum.

8. The method for processing millimeter-wave radar signals for vehicle-mounted vital signs monitoring according to claim 1, characterized in that, S3 includes: S301. Perform a fast Fourier transform on the radar echo signal after DC removal to obtain the slow-time radar echo spectrum of each range cell. Perform a dot multiplication operation between the weighted frequency domain mask and the radar echo spectrum to obtain the first interference suppression spectrum that initially attenuates strong interference frequencies. S302. For the residual vibration interference frequencies that overlap with the vital signs frequency band contained in the first interference suppression spectrum, vector cancellation is performed using a complex domain interference cancellation model to obtain the second interference suppression spectrum. The complex domain interference cancellation model uses complex weighting coefficients to quantify the amplitude-phase mapping relationship between the equivalent vibration reference signal and the vibration interference component in the radar echo. By weighting the equivalent vibration reference signal according to the complex weighting coefficients and then subtracting it from the corresponding interference component in the first interference suppression spectrum, amplitude-phase joint cancellation is performed. S303. Perform an inverse slow-time Fourier transform on the second interference suppression spectrum, and reassemble the high signal-to-noise ratio slow-time signals recovered from each range cell into two-dimensional vibration-disrupted radar data with range cell and slow time as dimensions, as the reconstructed time-domain radar signal.

9. The method for processing millimeter-wave radar signals for vehicle-mounted vital signs monitoring according to claim 8, characterized in that, Complex weighting coefficients The expression is: ； In the formula, This is the first interference suppression spectrum. The spectrum of the vibration reference signal with zero mean The conjugate term of the spectrum of the vibration reference signal with zero mean; Indicates frequency; This is the regularization factor.

10. The method for processing millimeter-wave radar signals for vehicle-mounted vital signs monitoring according to claim 1, characterized in that, S4 includes: S401. Calculate the cumulative distribution of reflected energy of each range cell in slow time based on the vibration-disrupting two-dimensional radar data, and lock the chest cavity target range cell based on the maximum energy criterion and the constraint of human body micro-motion continuity. S402. Extract the slow-time complex signal phase sequence from the target distance unit, and perform inter-frame difference after unwinding to eliminate periodic jumps, thereby extracting thoracic cavity micro-motion information while suppressing background low-frequency drift. S403. Use a finite impulse response bandpass filter to separate the aforementioned thoracic cavity micro-motion signal into respiratory and cardiac frequency bands in the frequency domain. S404. The Welch method is used to estimate the power spectral density of the separated respiratory and heartbeat signals, and the maximum spectral peak frequency is searched within their respective frequency bands and converted into respiratory rate and heart rate.

Images