A method and system for identifying and suppressing multi-source interference of in-vehicle millimeter wave radar

Abstract

The application relates to the technical field of vehicle-mounted millimeter wave radar signal processing, and discloses a vehicle-mounted millimeter wave radar multi-source interference identification and suppression method and system, wherein echo signals of each chirp of a millimeter wave radar are transformed to a distance-velocity two-dimensional space to obtain a two-dimensional data matrix; the phases of each detection point in a slow time dimension in the two-dimensional data matrix are subjected to coherent accumulation and variance statistics; time-frequency transformation is performed on the two-dimensional data matrix to extract time-frequency texture features of each detection point; detection points meeting preset interference judgment conditions are marked as interference candidates; different types of interference are subjected to corresponding suppression processing according to preset priorities, and the detection points after the suppression processing are output as a clean target candidate set, so that the problems of unclear processing sequence and poor adaptability of a fixed threshold value when multiple interferences are superimposed are solved, stable interference identification performance is ensured when working conditions change, and human physiological signals and non-physiological disturbances can be effectively distinguished.

Description

Technical Field

[0001] This invention relates to the field of vehicle-mounted millimeter-wave radar signal processing technology, specifically to a method and system for identifying and suppressing multi-source interference in vehicle-mounted millimeter-wave radar. Background Technology

[0002] With the development of intelligent cockpit technology, in-vehicle non-contact vital sign detection technology has received widespread attention. Millimeter-wave radar, with its characteristics of large bandwidth, high resolution, all-weather operation and harmlessness to the human body, has shown great application value in the field of in-vehicle vital sign detection. At the same time, regulations such as Euro NCAP have included systems such as child abandonment warning and occupant classification in their evaluation system, promoting the rapid deployment of in-vehicle millimeter-wave radar perception technology.

[0003] However, applying millimeter-wave radar to the enclosed cavity environment of a vehicle for vital sign detection faces severe interference challenges. Unlike external traffic scenarios, in-vehicle radar operates in a close-range, multi-reflection, and confined space environment, and its interference sources are diverse and cumulative. Multipath strong reflection interference: There are a large number of strong reflective surfaces inside the vehicle, including the windshield, central control screen, door inner panels and seat metal frame, etc. In addition to being reflected by the target, the radar signal may also be reflected twice by the above-mentioned reflective surfaces and enter the receiving antenna, forming a mirror false target. Studies have shown that the number of false targets caused by multipath reflection can be up to 25 times that of real targets, which seriously interferes with target identification. Non-physiological periodic disturbances: The airflow velocity of automotive air conditioning systems can reach 2m / s to 8m / s, creating turbulence in the enclosed cabin and modulating the phase and amplitude of millimeter-wave signals. During vehicle operation, engine vibration and road surface excitation are transmitted to the seats through the chassis, causing random phase fluctuations at the seats and nearby detection points. The spectral distribution of these disturbances covers the range of several hertz to tens of hertz, which differs from the frequency bands of human breathing and heartbeat. However, if these disturbances are not treated indiscriminately, they can easily lead to misjudgment. Inter-radar crosstalk: A car may be equipped with multiple millimeter-wave radars at the same time. If the transmitted waveforms of different radars are not strictly orthogonal, the transmitted signals may directly or indirectly enter the receiving channels of other radars, resulting in frequency sweep crosstalk. To address the above problems, the existing technology has the following main shortcomings: First, the coverage of interference types is incomplete. Existing methods mostly deal with a single type of interference. For example, they use a time median filter in the short-time Fourier transform domain to suppress sweep frequency crosstalk, or they eliminate false targets for multipath effects. There is a lack of a unified identification framework for multi-source interference in the cabin. Second, the ability to distinguish between physiological signals and non-physiological disturbances is insufficient. Existing methods are unable to effectively distinguish between human respiratory and heartbeat signals and non-rigid motion disturbances such as air conditioning airflow and seat vibration. Third, the processing order is unclear when multiple interferences are superimposed. When the same echo signal is affected by multiple interferences at the same time, there is a lack of reasonable hierarchical suppression strategy, which can easily cause excessive damage to the signal or the suppression effect cancels each other out. The aforementioned defects result in a high false alarm rate for the existing system under complex in-vehicle conditions, making it difficult to meet the high reliability requirements of in-vehicle vital sign detection. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for identifying and suppressing multi-source interference in in-vehicle millimeter-wave radar, in order to solve the technical problem that the existing technology lacks a unified identification framework for multi-source interference in vehicles and cannot simultaneously handle the superposition of four types of interference: stable strong reflection, non-physiological periodic disturbance, crosstalk between radar frequencies, and multipath false targets.

[0005] To solve the above-mentioned technical problems, the present invention specifically provides the following technical solution: A method and system for identifying and suppressing multi-source interference of in-vehicle millimeter-wave radar, comprising the following steps: The raw echo signal of the in-vehicle millimeter-wave radar is collected, and the echo signal of each chirp of the millimeter-wave radar is transformed into a range-velocity two-dimensional space to obtain a two-dimensional data matrix. The phase of each detection point in the two-dimensional data matrix is ​​coherently accumulated and variance is statistically analyzed in the slow time dimension. Based on the phase variance, the detection points are classified into stable reflectors and unstable reflectors. The two-dimensional data matrix is ​​subjected to time-frequency transformation to extract the time-frequency texture features of each detection point. Based on the time-frequency texture features, the sweep frequency crosstalk interference and non-physiological periodic disturbance interference are identified. The non-physiological periodic disturbances include air conditioning airflow disturbances and non-rigid motion caused by seat vibration. Based on the classification results of stability analysis and time-frequency texture features, detection points that meet the preset interference judgment conditions are marked as interference candidates, and the interference candidates are identified by multipath false target matching based on the spatial geometric distribution of the reflective surface inside the vehicle cabin. According to the preset priority, the corresponding suppression processing is performed on different types of interference. When the same detection point is marked as multiple interference types, the suppression is performed in the following order: multipath false target priority is higher than frequency sweep crosstalk priority, frequency sweep crosstalk priority is higher than non-physiological periodic disturbance priority, and non-physiological periodic disturbance priority is higher than stable reflector priority. The detection points after suppression processing are output as a clean target candidate set.

[0006] Furthermore, the stability analysis step includes: Extracting complex echo sequences of the same distance-velocity element over N consecutive chirp periods. ,in The expression for vector coherence accumulation in the slow time dimension is:

[0007] in, This represents the complex echo of the k-th chirp. This represents the echo amplitude of the k-th chirp, where j is the imaginary unit. This represents the echo phase of the k-th chirp. This represents the coherent accumulation result after vector summation of N chirp complex echo sequences for the same range-velocity cell; Phase values ​​for each chirp Perform phase unwrapping and calculate the phase variance, the expression of which is:

[0008] in, Indicates the phase mean; Set the first phase variance threshold TH1, when It is determined to be a stable reflector when It was determined to be an unstable reflector.

[0009] Furthermore, the identification method for the frequency sweep crosstalk interference is as follows: Perform a short-time Fourier transform on the difference frequency signal of each chirp to generate a time-frequency spectrum. A straight line detection algorithm is used in the time spectrum to detect oblique line trajectories with continuous energy distribution along the time axis and slope within a preset interference slope tolerance range; Detection points with the oblique trajectory and a duration exceeding a preset time window are marked as frequency sweep crosstalk interference.

[0010] Furthermore, the method for identifying non-physiological periodic disturbances is as follows: Extract the slow-time dimension signal from each detection point and perform spectrum analysis; Calculate the proportion of spectral energy outside the human respiratory frequency band (0.1Hz~0.8Hz) and the heartbeat frequency band (0.8Hz~2.0Hz); If the proportion of energy outside the frequency band exceeds a preset threshold and the peak frequency of the spectrum does not exhibit periodic stability, it is determined to be non-physiological periodic disturbance interference.

[0011] Furthermore, the multipath false target matching and identification method includes: A three-dimensional coordinate system for the vehicle cockpit is established with the radar installation location as the origin, and the spatial plane equations of each reflector are obtained. Based on the principle of specular reflection, a spatial mapping relationship is established between the real target position and the position of the mirrored false target generated by each reflecting surface; Calculate the spatial distance between the spatial location of the detection point and the location of each mirrored false target in the mapping relationship; If the spatial distance is less than a preset distance threshold, the detection point is determined to be a multipath false target.

[0012] Furthermore, in the hierarchical suppression step, when the same detection point is marked as multiple interference types, the specific execution method is as follows: If a detection point is identified as a multipath false target, the detection point is directly removed and no further suppression operations are performed. If a detection point is not identified as a multipath false target but is marked as a frequency sweep crosstalk, perform median filtering of the time spectrum and re-evaluate the non-physiological periodic disturbance for the filtered detection point. If non-physiological periodic disturbance still exists, continue to perform notch filtering. If the detection point is only marked as a non-physiological periodic disturbance, perform notch filtering directly; If the detection point is only marked as a stable reflector, perform phase cancellation.

[0013] Furthermore, in the hierarchical suppression step, the suppression processing performed on each type of interference is as follows: For interference from stable reflectors, a static clutter estimate is constructed by temporal averaging of the echoes of the same range-velocity unit across multiple consecutive frames, and the static clutter estimate is subtracted from the echo of the current frame. For frequency sweep crosstalk, a median filter is performed in the time-frequency spectrum diagram with the marked point as the center and a sliding window of length L along the frequency sweep direction. The value of L ranges from 5 to 15 chirp units. For non-physiological periodic disturbances, the current rotation frequency of the air conditioner blower and the characteristic frequency of seat vibration are obtained, and these are used as the center frequency of the notch filter to construct a multi-notch filter to filter the slow time dimension signal. Multipath false target interference is directly eliminated.

[0014] Furthermore, the thresholds used for the determination are adaptively calculated in the following manner: The first phase variance threshold TH1 satisfies:

[0015] in, For radar intermediate frequency bandwidth, The standard deviation of the system phase noise. The empirical value range is 1.5 to 3.0; The threshold for the out-of-band energy ratio is determined by statistically analyzing the out-of-band energy ratio histogram of all detection points in the current frame and taking the 70% to 85% quantile of the peak value of the histogram. The spatial distance threshold TH2 satisfies:

[0016] in, Where is the speed of light, B is the bandwidth of the radar transmission signal, and γ ranges from 0.5 to 1.5.

[0017] Furthermore, in the stability analysis step, the total time length corresponding to N consecutive chirp cycles is not less than 0.5 seconds and not more than 2 seconds, so as to ensure sufficient phase tracking resolution for micro-motion signals in the 0.1Hz~2Hz frequency band.

[0018] A second aspect of the present invention provides an in-vehicle millimeter-wave radar multi-source interference identification and suppression system for implementing the method described in claims 1-9, comprising: The data acquisition module is used to receive the echo signal from the millimeter-wave radar and transform it into a range-velocity two-dimensional space; The stability analysis module is used to perform coherent accumulation and variance statistics on the phase of each detection point in the two-dimensional space in the slow time dimension, and to distinguish between stable reflectors and unstable reflectors. The texture analysis module is used to perform time-frequency transformation on each detection point in the two-dimensional space, extract time-frequency texture features, and identify sweep frequency crosstalk interference and non-physiological periodic disturbance interference. The classification and labeling module is used to label the detection points as interference candidates based on the output of the stability analysis module and the texture analysis module, and to perform multipath false target matching and identification based on the vehicle cabin mirror reflection model. The hierarchical suppression module is used to perform differentiated suppression processing on different types of interference according to the priority order of multipath false targets, frequency sweep crosstalk, non-physiological periodic disturbances, and stable reflectors, and output a clean target candidate set.

[0019] Compared with the prior art, the present invention has the following advantages: This invention classifies stable reflectors, non-physiological periodic disturbances, frequency sweep crosstalk, and multipath reflection false targets inside the vehicle by constructing dual-dimensional features of phase stability and short-time Fourier transform time-frequency texture. It establishes a hierarchical suppression strategy with the highest priority for multipath false targets and adopts an adaptive threshold based on radar parameters to solve the problems of unclear processing order and poor adaptability of fixed thresholds when multiple interferences are superimposed. The adaptive mechanism ensures that the algorithm maintains stable interference recognition performance without manual calibration when the working conditions change, such as air conditioning on / off, different vehicle models, and different numbers of occupants. At the same time, it can effectively distinguish between human physiological signals and non-physiological disturbances, and avoid misjudging conventional in-vehicle interferences such as air conditioning airflow and seat vibration as occupant targets. Attached Figure Description

[0020] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of the process for identifying and suppressing multi-source interference from in-vehicle millimeter-wave radar according to the present invention.

[0022] Figure 2 This is a block diagram of the in-vehicle millimeter-wave radar multi-source interference identification and suppression system of the present invention.

[0023] Figure 3 This is a schematic diagram of the distance-velocity two-dimensional data matrix of the present invention.

[0024] Figure 4 This invention provides a slow-time-dimensional phase sequence comparison.

[0025] Figure 5 This is a schematic diagram of the STFT time spectrum and the oblique trajectory of the frequency sweep crosstalk in this invention.

[0026] Figure 6 This is a schematic diagram of the slow-time spectrum of the non-physiological periodic perturbation of the present invention.

[0027] Figure 7 This invention relates to the three-dimensional cockpit coordinate system and reflector model of the in-vehicle millimeter-wave radar of the present invention.

[0028] Figure 8 This is a flowchart of the hierarchical interference suppression priority processing of the present invention. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0030] like Figures 1 to 8 As shown, this invention provides a method and system for identifying and suppressing multi-source interference in in-vehicle millimeter-wave radar. The raw echo signal of the in-vehicle millimeter-wave radar is collected, and the echo signal of each chirp of the millimeter-wave radar is transformed into a range-velocity two-dimensional space to obtain a two-dimensional data matrix. In this embodiment, the radar system configuration and data acquisition process is as follows: Taking a certain SUV model as the example, the millimeter-wave radar is installed directly behind the rearview mirror inside the vehicle, with the radar antenna array facing the passenger area inside the vehicle. The radar operating parameters are as follows: Operating frequency band: 77GHz~81GHz (center frequency 79GHz); Chirp period: T c =50μs; Number of chirps per frame: M=128; Signal bandwidth: B = 4 GHz (corresponding to a distance resolution of δ) r =c / (2B)≈0.0375m); ADC sampling rate: f s =2.5MHz; Transmit power: 10dBm; After the radar is started, the waveform generator generates a linear frequency modulated continuous wave (FMCW) signal, which is radiated into the cabin through the transmitting antenna. The receiving antenna receives the target reflected echo, which is amplified by a low-noise amplifier and mixed with the transmitted signal to output a difference frequency (IF) signal. The difference frequency signal is then sent to the digital signal processor after analog-to-digital conversion. Perform a distance-dimensional FFT (256 points) on the difference frequency signal of each chirp to obtain the distance spectrum; Perform a velocity-dimensional FFT on the range spectrum of 128 consecutive chirps (128 points) to obtain a two-dimensional range-velocity data matrix. Each cell in the matrix... It is a complex value, containing amplitude and phase information; The FMCW radar's transmitted and echo signals are mixed to obtain a difference frequency signal, the difference frequency being... The distance d from the target satisfies:

[0031] Where B is the sweep bandwidth of the radar transmitted signal (in Hz), which is taken as 4 GHz in this embodiment. The period of one chirp (in seconds) is taken in this embodiment. c is the speed of light (3 × 10⁻⁶) 8 (m / s), where d is the distance between the radar and the target.

[0032] The phase of each detection point in the two-dimensional data matrix is ​​coherently accumulated and variance is statistically analyzed in the slow time dimension. Based on the phase variance, the detection points are classified into stable reflectors and unstable reflectors. The two-dimensional data matrix is ​​subjected to time-frequency transformation to extract the time-frequency texture features of each detection point. Based on the time-frequency texture features, the sweep frequency crosstalk interference and non-physiological periodic disturbance interference are identified. The non-physiological periodic disturbances include air conditioning airflow disturbances and non-rigid motion caused by seat vibration. Based on the classification results of stability analysis and time-frequency texture features, detection points that meet the preset interference judgment conditions are marked as interference candidates, and the interference candidates are identified by multipath false target matching based on the spatial geometric distribution of the reflective surface inside the vehicle cabin. This invention combines phase variance statistics (reflecting the target's micro-motion stability in the slow time dimension) with time-frequency texture feature extraction (reflecting the target's spectral distribution pattern and time-frequency domain morphology) to form a two-dimensional feature classification system. This system can accurately distinguish three types of key targets / interference: Stable reflector: extremely low phase variance ( <0.01~0.05rad 2 The temporal spectrum has no texture features, corresponding to a completely static structure; Human target: Moderate phase variance (0.01~0.05 rad) 2 ≤ <0.15~0.30rad 2 The spectral energy is concentrated in the range of 0.1Hz to 2.0Hz and is periodic, corresponding to the physiological signals of breathing and heartbeat; Non-physiological perturbation: High phase variance ( ≥0.15~0.30rad 2 () or although moderate, the proportion of energy outside the frequency band is high (>60%~80%), and the spectrum has no physiological periodicity, corresponding to air conditioning airflow and seat vibration.

[0033] This invention achieves effective differentiation between human physiological signals and non-physiological periodic disturbances through a two-dimensional progressive classification system that combines initial screening with phase variance thresholding with secondary confirmation using time-frequency texture features.

[0034] First dimension: Preliminary classification based on phase variance The periodic rise and fall of the human chest cavity with respiration and the micro-movements on the body surface caused by heartbeats produce quasi-periodic phase changes in the slow time dimension. The phase variance of human targets (including respiration and heartbeat) usually falls within the following range: Human respiratory dominance zone: ≈0.02~0.08rad 2 ; Human heart rate control area: ≈0.01~0.04rad 2 ; Human Respiration + Heartbeat Integration Zone: ≈0.02~0.10 rad 2 ; The phase fluctuations caused by air conditioning disturbances and seat vibrations are usually larger and more random, and their phase variances often deviate significantly from the typical range of human targets. Air conditioning airflow disturbance: usually ≥ 0.15 rad 2 (large random phase fluctuation caused by air turbulence); Seat vibration disturbance: usually ≥ 0.20 rad 2 (periodic but non-physiological phase fluctuation caused by mechanical vibration); Based on the above rules, the present invention sets double thresholds (TH1=0.01~0.05rad 2 to distinguish stable / non-stable; TH3=0.15~0.30rad 2 to distinguish non-stable / strongly non-stable); When is significantly higher than TH3 (e.g., ≥ 0.15~0.30 rad 2 ), it is directly determined as a strongly non-stable source, which corresponds to non-physiological disturbances such as air conditioning airflow or seat vibration with a high probability; When is between TH1 and TH3, it may be either a human target or a non-physiological disturbance with small amplitude (such as air conditioning in gentle wind gear or low-frequency seat vibration), and it is necessary to enter the second dimension for further screening.

[0035] Second dimension: accurate screening based on time-frequency texture features For detection points whose phase variance falls into the fuzzy interval (TH1 ≤ < TH3), the present invention further extracts the frequency domain features of the signals in the slow time dimension, and judges whether it is a human physiological signal through the distribution of spectral energy.

[0036] Specifically, human respiration and heartbeat signals have clear physiological frequency band constraints: Respiratory frequency band: 0.1 Hz ~ 0.8 Hz; Heartbeat frequency band: 0.8 Hz ~ 2.0 Hz; Within the above frequency bands, the human signal presents periodic spectral lines with stable peaks, and there is a frequency doubling relationship or separable frequency band distribution between the respiratory harmonic and the heartbeat fundamental frequency.

[0037] In contrast, the spectral distribution of air conditioning airflow disturbance and seat vibration presents the following characteristics: Difference in energy distribution: the rotation frequency of an air conditioning blower is usually 20 Hz ~ 150 Hz (corresponding to 1200 rpm ~ 9000 rpm), and the seat vibration frequency is usually 5 Hz ~ 30 Hz; the main energy of both types of disturbances is concentrated above 2.0 Hz, which is frequency-domain separated from the human physiological frequency band (0.1 Hz ~ 2.0 Hz); Periodic differences: Human breathing and heartbeat are quasi-periodic movements driven by physiological rhythms, and their spectra show stable fundamental frequency and harmonic peaks, with the peak frequencies remaining stable over multiple consecutive frames; the peak frequencies of air conditioning disturbances and seat vibrations are affected by factors such as changes in blower speed and road surface excitation, and their frequencies do not exhibit the physiological level (0.1Hz~2.0Hz) quasi-periodic stability characteristics. Based on the above differences, this invention calculates the proportion of energy outside the frequency band, and determines it to be a non-physiological periodic disturbance when the following two conditions are met simultaneously: Condition 1: >TH4 (TH4=60%~80%) indicates that the signal energy is mainly concentrated outside the respiratory and heartbeat frequency band, which does not conform to the energy distribution pattern of human physiological signals; Condition 2: The peak frequency of the spectrum does not exhibit periodic stability, that is, the peak frequency of the spectrum varies by more than ±15% within several consecutive frames (such as 5 to 10 consecutive frames), while the human respiratory rate is basically constant in the short term (the variation is usually less than ±5%).

[0038] The dual-dimensional classification system of this invention enables more refined identification of interference sources, directly supporting the accurate implementation of subsequent hierarchical suppression strategies and avoiding damage to the target signal caused by one-size-fits-all suppression.

[0039] According to the preset priority, the corresponding suppression processing is performed on different types of interference. When the same detection point is marked as multiple interference types, the suppression is performed in the following order: multipath false target priority is higher than frequency sweep crosstalk priority, frequency sweep crosstalk priority is higher than non-physiological periodic disturbance priority, and non-physiological periodic disturbance priority is higher than stable reflector priority. The detection points after suppression processing are output as a clean target candidate set.

[0040] The stability analysis steps include: Extracting complex echo sequences of the same distance-velocity element over N consecutive chirp periods. ,in The expression for vector coherence accumulation in the slow time dimension is:

[0041] in, This represents the complex echo of the k-th chirp. This represents the echo amplitude of the k-th chirp, where j is the imaginary unit. This represents the echo phase of the k-th chirp. This represents the coherent accumulation result after vector summation of N chirp complex echo sequences for the same range-velocity cell; Phase values ​​for each chirp Perform phase unwrapping and calculate the phase variance, the expression of which is:

[0042] in, Indicates the phase mean; Set the first phase variance threshold TH1, when It is determined to be a stable reflector when It was determined to be an unstable reflector.

[0043] This invention extracts the complex echo sequence of the same distance-velocity unit within N consecutive chirp cycles based on a two-dimensional data matrix; In this embodiment, N=128 (corresponding to all chirps in one frame), and the slow time sampling interval is... =50μs, the total time length is 128×50μs=6.4ms. It is worth noting that this duration is used for short-time phase variance calculation within the frame, while the phase sequence used for vital sign signal tracking accumulates across multiple frames.

[0044] Extract the complex echo sequence of a range-velocity unit (i, j) from 128 consecutive chirps:

[0045] Perform vector coherent accumulation in the slow time dimension:

[0046] Coherent accumulation can improve the signal-to-noise ratio of weak target echoes by approximately 10 log₂. 10 (128)≈21dB, effectively enhancing the detection capability of human micro-motion signals; the method based on multi-chirp accumulation can significantly improve the signal-to-noise ratio. Phase values ​​for each chirp Perform phase unwrapping, the original phase is obtained through arg( ) is obtained, and its main value is restricted to Within the range; when the target undergoes a slight movement causing a phase change exceeding At that time, 2 will appear. Phase jump; This embodiment uses the Itoh algorithm for phase unwrapping: it compares point by point along the slow time dimension, and when the difference between adjacent phases exceeds ±π, it performs ±2π correction to restore the phase sequence to a continuous curve; Calculate the average phase value after unwrapping:

[0047] Calculate the phase variance:

[0048] In this embodiment, the system phase noise standard deviation Radar intermediate frequency bandwidth:

[0049] This invention calculates the first phase variance threshold using an adaptive formula: TH1 = 2.0 × (π × 1.07 × 0.005) 2 ≈2.0×(0.0168) 2 ≈5.65×10 -4 rad 2 That is, 0.000565 rad 2 Within the range of 0.01~0.05 rad given in the technical solution. 2 Within the range.

[0050] The test results are shown below: Seat surface detection points (1.0m from radar): The phase sequence remains almost constant between each chirp. ≈0.003rad 2 <TH1(0.01rad 2 It is determined to be a stable reflector; Human chest cavity detection point (1.5m from radar): exhibits periodic phase fluctuations in response to respiratory rhythm. ≈0.035rad 2 >TH1 is determined to be an unstable reflector; Detection point near the air conditioner vent (2.0m from the radar): The phase exhibits large random fluctuations. ≈0.22rad 2 >TH3 (0.15 rad) 2 It was determined to be a strongly unstable source.

[0051] The method for identifying the frequency sweep crosstalk interference is as follows: Perform a short-time Fourier transform on the difference frequency signal of each chirp to generate a time-frequency spectrum. A straight line detection algorithm is used in the time spectrum to detect oblique line trajectories with continuous energy distribution along the time axis and slope within a preset interference slope tolerance range; Detection points with the oblique trajectory and a duration exceeding a preset time window are marked as frequency sweep crosstalk interference.

[0052] In this invention, the time-frequency texture feature extraction process is as follows: Short-time Fourier transform (STFT) is performed on the difference frequency signals of each chirp. In this embodiment, a Hanning window is used with a window length L. w =64 sampling points, overlap rate 50%; frequency resolution is Δf=f s / L w=2.5MHz / 64≈39.06kHz; STFT is executed along a slow time dimension sliding window to generate a distance-Doppler-time three-dimensional time spectrum.

[0053] Furthermore, the frequency sweep crosstalk identification process is as follows: In the time spectrum, the transmitted signals of other vehicle radars (such as forward radar, corner radar, etc.) may couple into the receiving channel of this radar. If the frequency modulation slope of the interfering radar is different from that of this radar, a difference frequency signal that changes linearly with time will be generated after mixing, which will appear as a diagonal trajectory in the time spectrum.

[0054] This embodiment uses the Hough transform for line detection: The time-frequency spectrum (with time as the horizontal axis and frequency as the vertical axis) is binarized to extract pixels with energy higher than the noise floor +10dB. The votes are accumulated in the Hough parameter space (angle θ, distance ρ); Detect straight line segments with a slope matching tolerance of ±10%; Pixels whose line segments last for more than 10 chirp cycles are marked as sweep frequency crosstalk interference. For example, in one actual measurement, a slope k was detected at a distance of 1.2m from the element. int The oblique trajectory, with a frequency sweep crosstalk of approximately 0.15 MHz / chirp and lasting for 22 chirp cycles, was determined to be from the forward radar of a nearby vehicle.

[0055] The method for identifying non-physiological periodic disturbances is as follows: Extract the slow-time dimension signal from each detection point and perform spectrum analysis; Calculate the proportion of spectral energy outside the human respiratory frequency band (0.1Hz~0.8Hz) and the heartbeat frequency band (0.8Hz~2.0Hz); If the proportion of energy outside the frequency band exceeds a preset threshold and the peak frequency of the spectrum does not exhibit periodic stability, it is determined to be non-physiological periodic disturbance interference.

[0056] In this invention, the non-physiological periodic disturbance identification process is as follows: Extract the signal sequence along the slow time dimension of each detection point (data spanning multiple frames; in this embodiment, 512 consecutive frames are taken, corresponding to approximately 6.4 seconds of data), and perform spectrum analysis (FFT point count is 1024). The human respiratory frequency band is set to 0.1Hz~0.8Hz, and the heart rate frequency band is set to 0.8Hz~2.0Hz-6. Calculate the energy of each frequency band: Total spectral energy from 0Hz to 5Hz; Energy in the 0.1Hz~0.8Hz frequency band; : Energy in the 0.8Hz~2.0Hz frequency band; Calculating the proportion of energy outside the frequency band:

[0057] In this embodiment, the proportion of energy outside the frequency band is determined by counting the histogram of the proportion of energy outside the frequency band of all detection points in the current frame Critical value TH4; In a typical operation, statistics are collected for about 200 detection points in a frame distribution, the peak of the histogram appears at about 50%, and the 80% quantile is taken as TH4≈68%.

[0058] Under the working condition of air conditioner turned on, the blower operates at medium speed (rotating speed is about 3000rpm, corresponding to fundamental frequency f fan ≈50Hz), the characteristic frequency of seat vibration f vib ≈15Hz; in the slow-time spectrum of a detection point (1.8m away from the radar), except for the breathing and heartbeat frequency band, there is an obvious spectral peak at 50Hz, t=85%>TH4, and the peak frequency of the spectrum is not within the breathing and heartbeat frequency band, so it is determined as non-physiological periodic disturbance.

[0059] The multipath false target matching identification method comprises: A three-dimensional coordinate system of the vehicle cockpit is established with the radar installation position as the origin, and the spatial plane equations of each reflecting surface are obtained; Based on the principle of specular reflection, a spatial mapping relationship between the real target position and the position of the mirror false target generated by each reflecting surface is established; Calculating the spatial distance between the spatial position of the detection point and the position of each mirror false target in the mapping relationship; If the spatial distance is less than a preset distance threshold, the detection point is determined as a multipath false target.

[0060] In the present invention, the identification process of interference candidate marking and multipath false target is: Based on the classification results, interference candidates are marked according to the following rules: Rule 1: If a detection point <TH1 (stable reflector) and is marked as frequency sweep crosstalk or non-physiological periodic disturbance, it is marked as an interference candidate; Rule 2: If a detection point TH1≤ <TH3 (unstable reflector), and its time-frequency texture feature does not conform to the periodic characteristics of the human breathing and heartbeat frequency band, it is marked as an interference candidate; Multipath false target matching identification adopts a spatial matching method based on specular reflection model: Establish a three-dimensional cockpit coordinate system with the radar installation location (behind the rearview mirror) as the origin: The X-axis points towards the front of the vehicle (from the rear of the passengers to the front), the Y-axis points to the right side of the vehicle, and the Z-axis points vertically upward (towards the roof). For the SUV model in this embodiment, the following reflection surface plane equation is extracted from the vehicle CAD model: Windshield: (Including an inclination angle of approximately 60°); Central control screen: (Tilted at approximately 65°); Left door inner panel: ; For any real target location Its mirror image position with respect to the reflecting surface P is ,in For the Householder reflection matrix, For example, for the windshield plane, if the actual target is located at the driver's chest position. m, calculate the mirror position m (symmetrical position behind the glass). The spatial distance threshold TH2 = c / (2B)·γ = 0.0375 × 1.0 = 0.0375m ≈ 0.04m, which is rounded down to 0.05m, within the given range of 0.05~0.15m; If the spatial distance between a detection point and its mirror image is less than TH2 (0.05m), it is determined to be a multipath false target; In this embodiment, during actual testing, a target was detected at a distance of 1.2m from the radar and at an angle 15° to the left. Its spatial location was... m, and the driver's chest real target m is the mirror image of the windshield. The projection of m in the XY plane is close (approximately 0.045m), but the difference in the Z direction is significant, resulting in a failed match; while another detection point is located at... m, and the mirror position .

[0061] The spatial distance of m is approximately 0.02m, which is less than TH2. Therefore, it is determined to be a multipath false target generated by the windshield and marked.

[0062] This invention is the first to introduce the spatial plane equations of the radar installation location, each reflecting surface (windshield, central control screen, and door inner panel), the mirror reflection matrix (Householder reflection matrix), and the translation vector into the interference identification process. By establishing a spatial mapping relationship between real targets and mirrored false targets, it achieves accurate spatial matching and elimination of multipath false targets, thereby improving the accuracy of multipath false target identification.

[0063] In the hierarchical suppression step, when the same detection point is marked as multiple interference types, the specific execution method is as follows: If a detection point is identified as a multipath false target, the detection point is directly removed and no further suppression operations are performed. If a detection point is not identified as a multipath false target but is marked as a frequency sweep crosstalk, perform median filtering of the time spectrum and re-evaluate the non-physiological periodic disturbance for the filtered detection point. If non-physiological periodic disturbance still exists, continue to perform notch filtering. If the detection point is only marked as a non-physiological periodic disturbance, perform notch filtering directly; If the detection point is only marked as a stable reflector, perform phase cancellation.

[0064] In this invention, when the same detection point is marked as multiple interference types, suppression processing is performed according to the following priority: Priority 1 (highest): Multipath false target removal If a detection point is marked as a multipath false target, the detection point is directly deleted from the target candidate set, and no further suppression operations are performed.

[0065] Priority 2: Median filtering of the spectrum during frequency sweep crosstalk If it is not a multipath false target but is marked as a frequency sweep crosstalk, perform median filtering in the STFT time spectrum; take a sliding window of length L=9 chirp units along the frequency sweep direction with the marked point as the center, take the median of the energy value of each frequency channel in the sliding window, replace the center point value with the median, and after filtering, re-detect whether there is a non-physiological periodic disturbance in the slow time signal of the detection point; if it still exists, continue to perform notch filtering with priority 3. In the actual test, after median filtering, the signal energy at the detection point marked as frequency sweep crosstalk (at a distance of 1.2m) dropped from about -45dBm to -63dBm, the interference was suppressed by about 18dB, and the time-spectrum characteristics of the target signal were restored. Priority 3: Notch filtering for non-physiological periodic disturbances For detection points marked only as non-physiological periodic disturbances, read the current speed signal of the air conditioning blower from the vehicle CAN bus; In this embodiment, the air conditioner is set to medium speed, and the blower speed is 3000 rpm, corresponding to the base frequency f. fan =50Hz, obtain the seat vibration characteristic frequency f from vibration spectrum analysis. vib =15Hz; Using 50Hz, 100Hz (2nd harmonic), 150Hz (3rd harmonic) and 15Hz as notch center frequencies, and setting the notch bandwidth to 8% of each center frequency, an IIR multi-notch filter was constructed; notch filtering was performed on the slow time dimension signal within the marked interference period to filter out non-physiological frequency band signal components caused by air conditioning and vibration. Priority 4 (lowest): Stabilized reflector phase cancellation For detection points marked only as stable reflectors, a static clutter estimate is constructed by temporal averaging of the echoes from the same range-velocity cell over M=16 consecutive frames (corresponding to approximately 0.8 seconds):

[0066] From the current frame echo Subtracting static clutter from the estimate:

[0067] Dynamic target echo components are preserved, while strong static reflections (seat frame, center console structure, etc.) are effectively suppressed; After each suppression operation is completed, the signal-to-noise ratio of the current detection point is re-evaluated. If the improvement in signal-to-noise ratio is less than 0.5dB, the suppression operation is not performed to avoid excessive damage to the signal.

[0068] This invention establishes for the first time a priority system for interference types: multipath false targets (highest) > frequency sweep crosstalk > non-physiological periodic disturbances > stable reflectors (lowest). It also clearly defines the specific execution process for when the same detection point is marked as multiple interference types. If it is a multipath false target, it will be directly eliminated and no further suppression will be performed. If it is not a multipath false target but is frequency sweep crosstalk, it will first be median filtered and then re-detected for non-physiological disturbances. If it is only a non-physiological disturbance, it will be directly notched filtered. If it is only a stable reflector, it will be phase cancellation.

[0069] Multipath false targets are given the highest priority because they are spatially completely different from real targets, and their direct removal will not affect the signal quality of any real targets. Stable reflectors are given the lowest priority because they cause the least interference to dynamic targets and are the gentlest to cancel out with the least signal damage. This priority design avoids excessive signal damage or mutual cancellation of suppression effects caused by improper suppression order when various interferences exist simultaneously and superimpose each other, thus ensuring the output quality of a clean target candidate set.

[0070] In the hierarchical suppression step, the suppression processing performed on each type of interference is as follows: For interference from stable reflectors, a static clutter estimate is constructed by temporal averaging of the echoes of the same range-velocity unit across multiple consecutive frames, and the static clutter estimate is subtracted from the echo of the current frame. For frequency sweep crosstalk, a median filter is performed in the time-frequency spectrum diagram with the marked point as the center and a sliding window of length L along the frequency sweep direction. The value of L ranges from 5 to 15 chirp units. For non-physiological periodic disturbances, the current rotation frequency of the air conditioner blower and the characteristic frequency of seat vibration are obtained, and these are used as the center frequency of the notch filter to construct a multi-notch filter to filter the slow time dimension signal. Multipath false target interference is directly eliminated.

[0071] In this invention, after hierarchical suppression processing, a clean target candidate set is output. Each target in the candidate set contains four-dimensional information: distance, velocity, angle, and signal-to-noise ratio. This embodiment is validated on a test set containing 1500 frames of actual test data. The test scenarios include: Driver breathing normally (single person), front driver + front passenger (two people), four passengers in the whole vehicle (four people), air conditioning on (low / medium / high three levels), and vehicle stationary and idling vibration conditions; The false alarm rate is defined as FP / (FP+TN), where FP is the number of interferences that are mistakenly reported as targets, and TN is the number of interferences that are correctly identified and eliminated. In 1500 frames of test data, the total number of detection points was approximately 18,000 (an average of 12 detection points per frame), of which approximately 3,600 were actual interference points. After processing by this method, the number of interference points correctly identified and eliminated was TN=3592, and FP=8 (interference was falsely reported as the target). The false alarm rate was 8 / (3592+8)≈0.22%, which is better than the target requirement of 0.1%. In typical operating condition tests, when the air conditioner is turned on at medium speed, the accuracy rate of non-physiological cycle disturbance recognition reaches 94.5%; the accuracy rate of multipath false target recognition reaches 91.2%; and in a four-occupant full-seat scenario, the signal-to-noise ratio of vital signs signals is improved by about 15~20dB compared with the original signal.

[0072] The thresholds used for the determination are adaptively calculated in the following manner: The first phase variance threshold TH1 satisfies:

[0073] in, For radar intermediate frequency bandwidth, The standard deviation of the system phase noise. The empirical value range is 1.5 to 3.0; The threshold for the out-of-band energy ratio is determined by statistically analyzing the out-of-band energy ratio histogram of all detection points in the current frame and taking the 70% to 85% quantile of the peak value of the histogram. The spatial distance threshold TH2 satisfies:

[0074] in, Where is the speed of light, B is the bandwidth of the radar transmission signal, and γ ranges from 0.5 to 1.5.

[0075] In the stability analysis step, the total time length corresponding to N consecutive chirp cycles is not less than 0.5 seconds and not more than 2 seconds, so as to ensure sufficient phase tracking resolution for micro-motion signals in the 0.1Hz~2Hz frequency band.

[0076] This invention employs adaptive thresholds in multiple decision-making stages, including: a first phase variance threshold TH1 (adaptively calculated based on radar intermediate frequency bandwidth and phase noise), an out-of-band energy proportion threshold TH4 (dynamically determined based on 70%~85% quantiles of the peak values ​​of the histograms of all detection points in the current frame), and a spatial distance threshold TH2 (adaptively based on radar bandwidth). The above adaptive mechanisms ensure that the algorithm maintains stable interference recognition performance without manual calibration when operating conditions change, such as air conditioning being turned on / off, different vehicle models, and different numbers of occupants.

[0077] In this embodiment, by establishing an interference type priority system and defining a clear execution flow, the problems of unclear processing order and poor adaptability of fixed thresholds when multiple interferences are superimposed are solved. The processing flow is as follows: Based on the nature of the impact of each type of interference on the target signal, this invention sets a priority order from high to low: First priority (highest): Multipath false target removal. Multipath false targets are completely different from real targets in spatial location (they usually deviate from the real target location by more than 0.05m to 0.15m). They are essentially false echoes that do not exist in space. Directly removing this type of interference will not affect the signal quality of any real target. Moreover, if other suppression operations (such as median filtering or notch filtering) are performed first, it may change the signal characteristics of the multipath false target, blurring its feature boundary with the real target, which will reduce the accuracy of subsequent matching and identification. Therefore, this invention places multipath false target removal as the highest priority to ensure that identification and removal are completed before the signal characteristics are changed by any other processing. Second priority: Median filtering of frequency sweep crosstalk in the time spectrum. Frequency sweep crosstalk appears as a diagonal energy trajectory along the time axis in the time spectrum, which is fundamentally different from the distribution pattern of normal targets in the time-frequency domain (stable and continuous horizontal energy distribution). Median filtering is performed along the frequency sweep direction, which can eliminate the diagonal interference with almost no change to the time-frequency characteristics of normal targets. If notch filtering or phase cancellation is performed first, it may destroy the diagonal texture characteristics of frequency sweep crosstalk, making it difficult for subsequent median filtering to accurately identify and locate the interference. Therefore, this invention places frequency sweep crosstalk as the second priority. Third priority: Notch filtering for non-physiological periodic disturbances. The spectral components of non-physiological periodic disturbances (such as the 50Hz fundamental frequency and harmonics of an air conditioner blower, and the 15Hz characteristic frequency of seat vibration) have a clear frequency division boundary with the frequency band of human physiological signals (0.1Hz~2.0Hz). Notch filtering only filters out specific non-physiological frequency points and has almost no effect on human physiological signals. If phase cancellation is performed first, it may change the amplitude and phase characteristics of non-physiological disturbances and affect the accurate estimation of notch frequency. Therefore, this invention places non-physiological periodic disturbances in the third priority. Fourth priority (lowest): Phase cancellation of stable reflectors. The echoes of stable reflectors (seat frame, center console structure, etc.) have a constant phase in the slow time dimension. They can be subtracted after constructing a clutter estimate by time-domain averaging. This operation has the mildest impact on all signals and does not depend on the results of other suppression operations. Therefore, this invention places stable reflectors in the lowest priority.

[0078] The specific execution flow of this invention for multiple interferences superimposed at the same detection point is as follows: When the same detection point is marked as multiple interference types, follow the procedure below: Scenario 1: If a detection point is identified as a multipath false target, the detection point is directly removed and no further suppression operations are performed. Reason: Multipath false targets are false targets in themselves, and retaining their signals is of no value. Performing other suppression operations first may consume computational resources and may change the signal characteristics, affecting subsequent judgments.

[0079] Scenario 2: The detection point is not identified as a multipath false target, but is marked as sweep frequency crosstalk. First, perform median filtering of the time spectrum to eliminate the pollution of the signal time-frequency characteristics by the sweep frequency crosstalk. Then, the non-physiological periodic disturbances are re-evaluated for the filtered detection points. This is because the energy of the sweep frequency crosstalk may increase the energy estimate of the non-physiological frequency band, leading to the sweep frequency crosstalk being mistakenly marked as a non-physiological periodic disturbance. Re-evaluation after filtering can avoid over-suppression. If non-physiological periodic disturbances still exist after re-evaluation, notch filtering will continue; otherwise, notch filtering will no longer be performed.

[0080] Scenario 3: If the detection point is only marked as a non-physiological periodic disturbance (not marked as a multipath false target or frequency sweep crosstalk), notch filtering is directly performed to filter out non-physiological frequency band components; Case 4: The detection point is only marked as a stable reflector, and phase cancellation is performed to eliminate static strong reflection.

[0081] After each suppression operation is completed, the present invention also re-evaluates the signal-to-noise ratio of the current detection point. If the improvement in signal-to-noise ratio is less than 0.5dB, it indicates that the suppression operation has limited effect on improving the signal quality of the current detection point. The risk of performing the operation (potential damage to the effective signal) outweighs the benefit, so the suppression operation is not performed to avoid excessive damage to the signal. This mechanism further ensures the processing safety in scenarios with blurred interference boundaries.

[0082] Second embodiment: An in-vehicle millimeter-wave radar multi-source interference identification and suppression system, used to implement the method described in claims 1-9, comprising: The data acquisition module is used to receive the echo signal from the millimeter-wave radar and transform it into a range-velocity two-dimensional space; The stability analysis module is used to perform coherent accumulation and variance statistics on the phase of each detection point in the two-dimensional space in the slow time dimension, and to distinguish between stable reflectors and unstable reflectors. The texture analysis module is used to perform time-frequency transformation on each detection point in the two-dimensional space, extract time-frequency texture features, and identify sweep frequency crosstalk interference and non-physiological periodic disturbance interference. The classification and labeling module is used to label the detection points as interference candidates based on the output of the stability analysis module and the texture analysis module, and to perform multipath false target matching and identification based on the vehicle cabin mirror reflection model. The hierarchical suppression module is used to perform differentiated suppression processing on different types of interference according to the priority order of multipath false targets, frequency sweep crosstalk, non-physiological periodic disturbances, and stable reflectors, and output a clean target candidate set.

[0083] This embodiment provides an in-vehicle millimeter-wave radar multi-source interference identification and suppression system. Its system hardware architecture is based on the TIIWR6843 millimeter-wave radar SoC. The IWR6843 integrates 3 transmit channels and 4 receive channels, operates in the 60GHz~64GHz frequency band (or 77GHz~81GHz model), and has a built-in ARMCortex-R4F main processor and DSP subsystem. The data acquisition module acquires the difference frequency signal of each chirp at a sampling rate of 2.5MHz using an ADC. It performs a distance-dimensional FFT on 256 sampling points of each chirp and a velocity-dimensional FFT on the distance spectrum of 128 consecutive chirps to construct a distance-velocity two-dimensional data matrix. This module is also responsible for the buffer management of the raw data and sends the two-dimensional data matrix into the subsequent processing pipeline.

[0084] The stability analysis module includes a phase variance calculation unit. It performs slow-time coherent accumulation on the complex echo sequence of the same distance-velocity unit over N=128 consecutive chirp periods, employs a phase unwrapping algorithm (Itoh algorithm) to eliminate 2π jumps, and calculates the phase variance. Set TH1 = 0.01 rad 2 TH3 = 0.15 rad 2 Output the phase stability classification label (stable reflector / unstable reflector / strongly unstable source) for each detection point. The texture analysis module includes a time-frequency transformation unit, which performs STFT (Hanning window, window length 64 points, overlap rate 50%) on each chirp difference frequency signal to generate a time-frequency spectrum. In this module, sweep frequency crosstalk texture recognition (Hough transform line detection) and non-physiological periodic disturbance texture recognition (spectrum analysis + out-of-band energy proportion histogram threshold determination) are realized. The classification and labeling module receives the classification results from the stability analysis module and the texture analysis module, and makes a comprehensive judgment according to the interference candidate labeling rules (rule 1 and rule 2). This module has a built-in database of pre-calibrated vehicle cabin mirror reflection models for the target vehicle model, which includes the planar equation parameters and Householder reflection matrices of each reflecting surface (windshield, central control screen, door inner panel). Translation vector For detection points marked as interference candidates, multipath false target spatial matching and identification are performed based on the specular reflection model. The layered suppression module performs layered suppression processing in the following order: multipath false target removal (first priority), frequency sweep crosstalk median filtering (second priority), non-physiological periodic disturbance time domain notch filtering (third priority), and stable reflector phase cancellation (fourth priority). The module has a built-in IIR multi-notch filter bank and can dynamically update the notch center frequency by reading the air conditioning level information from the vehicle CAN bus.

[0085] In this embodiment, considering the differences in cabin geometry and reflective surface position among different vehicle models (sedans, SUVs, MPVs), this embodiment describes the system's adaptation method among different vehicle models; Reflector model calibration process: For the target vehicle model, a 3D laser scanner is used to acquire spatial point cloud data of various reflective surfaces in the cabin (inner surface of the windshield, surface of the central control screen, and metal surfaces of the four door inner panels); Plane fitting is performed on the point cloud of each reflecting surface to obtain the plane equation coefficients (a, b, c, d); A three-dimensional cockpit coordinate system is established with the radar installation location as the origin, and the plane equations of each reflector are entered into the vehicle parameter library of the classification and marking module. Adaptive threshold configuration: Due to differences in reflective surface materials and curvature, the multipath reflection intensity varies among different vehicle models, and the γ value of TH2 can be adjusted accordingly. Passenger cars (with a large windshield tilt angle, approximately 65°~70°): γ=0.8; SUVs (with a smaller windshield tilt angle, approximately 55°~60°): γ=1.0; MPV (with a more upright windshield, approximately 50°~55°): γ=1.2; System phase noise standard deviation for each vehicle model The phase noise of the radar transceiver is determined by the local oscillator noise, which is independent of the vehicle model and can be obtained from the radar chip datasheet. This embodiment was verified on an MPV model. The plane equation of the windshield differs from that of an SUV. After updating the vehicle parameters, the accuracy of multipath false target recognition increased from 89.5% (using the default SUV model) to 93.7%, verifying the effectiveness of the parameter adaptation method.

[0086] This invention classifies stable reflectors, non-physiological periodic disturbances, frequency sweep crosstalk, and multipath reflection false targets inside the vehicle by constructing dual-dimensional features of phase stability and short-time Fourier transform time-frequency texture. It establishes a hierarchical suppression strategy with the highest priority for multipath false targets and adopts an adaptive threshold based on radar parameters to solve the problems of unclear processing order and poor adaptability of fixed thresholds when multiple interferences are superimposed. The adaptive mechanism ensures that the algorithm maintains stable interference recognition performance without manual calibration when the working conditions change, such as air conditioning on / off, different vehicle models, and different numbers of occupants. At the same time, it can effectively distinguish between human physiological signals and non-physiological disturbances, and avoid misjudging conventional in-vehicle interferences such as air conditioning airflow and seat vibration as occupant targets.

[0087] The technical solution of the present invention also provides a computer-readable storage medium storing computer program instructions thereon, which, when executed by a processor, implement all the steps of the above-described in-vehicle millimeter-wave radar multi-source interference identification and suppression method and system.

[0088] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and not to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. All such modifications or substitutions should be covered within the protection scope of this application, and should not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application. Therefore, the protection scope of this application should be determined by the protection scope of the claims.

Claims

1. A method and system for identifying and suppressing multi-source interference in in-vehicle millimeter-wave radar, characterized in that, Includes the following steps: The raw echo signal of the in-vehicle millimeter-wave radar is collected, and the echo signal of each chirp of the millimeter-wave radar is transformed into a range-velocity two-dimensional space to obtain a two-dimensional data matrix. The phase of each detection point in the two-dimensional data matrix is ​​coherently accumulated and variance is statistically analyzed in the slow time dimension. Based on the phase variance, the detection points are classified into stable reflectors and unstable reflectors. The two-dimensional data matrix is ​​subjected to time-frequency transformation to extract the time-frequency texture features of each detection point. Based on the time-frequency texture features, the sweep frequency crosstalk interference and non-physiological periodic disturbance interference are identified. The non-physiological periodic disturbances include air conditioning airflow disturbances and non-rigid motion caused by seat vibration. Based on the classification results of stability analysis and time-frequency texture features, detection points that meet the preset interference judgment conditions are marked as interference candidates, and the interference candidates are identified by multipath false target matching based on the spatial geometric distribution of the reflective surface inside the vehicle cabin. According to the preset priority, the corresponding suppression processing is performed on different types of interference. When the same detection point is marked as multiple interference types, the suppression is performed in the following order: multipath false target priority is higher than frequency sweep crosstalk priority, frequency sweep crosstalk priority is higher than non-physiological periodic disturbance priority, and non-physiological periodic disturbance priority is higher than stable reflector priority. The detection points after suppression processing are output as a clean target candidate set.

2. The method for identifying and suppressing multi-source interference of in-vehicle millimeter-wave radar according to claim 1, characterized in that, The stability analysis steps include: Extracting complex echo sequences of the same distance-velocity element over N consecutive chirp periods. ,in The expression for vector coherence accumulation in the slow time dimension is: in, This represents the complex echo of the k-th chirp. This represents the echo amplitude of the k-th chirp, where j is the imaginary unit. This represents the echo phase of the k-th chirp. This represents the coherent accumulation result after vector summation of N chirp complex echo sequences for the same range-velocity cell; Phase values ​​for each chirp Perform phase unwrapping and calculate the phase variance, the expression of which is: in, Indicates the phase mean; Set the first phase variance threshold TH1, when It is determined to be a stable reflector when It was determined to be an unstable reflector.

3. The method for identifying and suppressing multi-source interference of in-vehicle millimeter-wave radar according to claim 1, characterized in that, The method for identifying the frequency sweep crosstalk interference is as follows: Perform a short-time Fourier transform on the difference frequency signal of each chirp to generate a time-frequency spectrum. A straight line detection algorithm is used in the time spectrum to detect oblique line trajectories with continuous energy distribution along the time axis and slope within a preset interference slope tolerance range; Detection points with the oblique trajectory and a duration exceeding a preset time window are marked as frequency sweep crosstalk interference.

4. The method for identifying and suppressing multi-source interference of in-vehicle millimeter-wave radar according to claim 1, characterized in that, The method for identifying non-physiological periodic disturbances is as follows: Extract the slow-time dimension signal from each detection point and perform spectrum analysis; Calculate the proportion of spectral energy outside the human respiratory frequency band (0.1Hz~0.8Hz) and the heartbeat frequency band (0.8Hz~2.0Hz); If the proportion of energy outside the frequency band exceeds a preset threshold and the peak frequency of the spectrum does not exhibit periodic stability, it is determined to be non-physiological periodic disturbance interference.

5. The method for identifying and suppressing multi-source interference of in-vehicle millimeter-wave radar according to claim 1, characterized in that, The multipath false target matching and identification method includes: A three-dimensional coordinate system for the vehicle cockpit is established with the radar installation location as the origin, and the spatial plane equations of each reflector are obtained. Based on the principle of specular reflection, a spatial mapping relationship is established between the real target position and the position of the mirrored false target generated by each reflecting surface; Calculate the spatial distance between the spatial location of the detection point and the location of each mirrored false target in the mapping relationship; If the spatial distance is less than a preset distance threshold, the detection point is determined to be a multipath false target.

6. The method for identifying and suppressing multi-source interference of in-vehicle millimeter-wave radar according to claim 1, characterized in that, In the hierarchical suppression step, when the same detection point is marked as multiple interference types, the specific execution method is as follows: If a detection point is identified as a multipath false target, the detection point is directly removed and no further suppression operations are performed. If a detection point is not identified as a multipath false target but is marked as a frequency sweep crosstalk, perform median filtering of the time spectrum and re-evaluate the non-physiological periodic disturbance for the filtered detection point. If non-physiological periodic disturbance still exists, continue to perform notch filtering. If the detection point is only marked as a non-physiological periodic disturbance, perform notch filtering directly; If the detection point is only marked as a stable reflector, perform phase cancellation.

7. The method for identifying and suppressing multi-source interference of in-vehicle millimeter-wave radar according to claim 1, characterized in that, In the hierarchical suppression step, the suppression processing performed on each type of interference is as follows: For interference from stable reflectors, a static clutter estimate is constructed by temporal averaging of the echoes of the same range-velocity unit across multiple consecutive frames, and the static clutter estimate is subtracted from the echo of the current frame. For frequency sweep crosstalk, a median filter is performed in the time-frequency spectrum diagram with the marked point as the center and a sliding window of length L along the frequency sweep direction. The value of L ranges from 5 to 15 chirp units. For non-physiological periodic disturbances, the current rotation frequency of the air conditioner blower and the characteristic frequency of seat vibration are obtained, and these are used as the center frequency of the notch filter to construct a multi-notch filter to filter the slow time dimension signal. Multipath false target interference is directly eliminated.

8. A method for identifying and suppressing multi-source interference in in-vehicle millimeter-wave radar according to claims 2, 4, and 5, characterized in that, The thresholds used for the determination are adaptively calculated in the following manner: The first phase variance threshold TH1 satisfies: in, For radar intermediate frequency bandwidth, The standard deviation of the system phase noise. The empirical value range is 1.5 to 3.0; The threshold for the out-of-band energy ratio is determined by statistically analyzing the out-of-band energy ratio histogram of all detection points in the current frame and taking the 70% to 85% quantile of the peak value of the histogram. The spatial distance threshold TH2 satisfies: in, Where is the speed of light, B is the bandwidth of the radar transmission signal, and γ ranges from 0.5 to 1.

5.

9. The method for identifying and suppressing multi-source interference of in-vehicle millimeter-wave radar according to claim 1, characterized in that, In the stability analysis step, the total time length corresponding to N consecutive chirp cycles is not less than 0.5 seconds and not more than 2 seconds, so as to ensure sufficient phase tracking resolution for micro-motion signals in the 0.1Hz~2Hz frequency band.

10. An in-vehicle millimeter-wave radar multi-source interference identification and suppression system, used to implement the method described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to receive the echo signal from the millimeter-wave radar and transform it into a range-velocity two-dimensional space; The stability analysis module is used to perform coherent accumulation and variance statistics on the phase of each detection point in the two-dimensional space in the slow time dimension, and to distinguish between stable reflectors and unstable reflectors. The texture analysis module is used to perform time-frequency transformation on each detection point in the two-dimensional space, extract time-frequency texture features, and identify sweep frequency crosstalk interference and non-physiological periodic disturbance interference. The classification and labeling module is used to label the detection points as interference candidates based on the output of the stability analysis module and the texture analysis module, and to perform multipath false target matching and identification based on the vehicle cabin mirror reflection model. The hierarchical suppression module is used to perform differentiated suppression processing on different types of interference according to the priority order of multipath false targets, frequency sweep crosstalk, non-physiological periodic disturbances, and stable reflectors, and output a clean target candidate set.

Images