Micrometer-level Vibration Detection Method and System Based on Multi-antenna Millimeter-wave Sensing

Abstract

The present invention discloses a micron-level vibration detection method and system based on multi-antenna millimeter-wave sensing. The method of the present invention is as follows: Step 1: Millimeter-wave sensing of mechanical vibration; Step 2: Constellation DC offset correction; Step 3: Phase extraction and noise removal based on the AD algorithm; Step 4: Processing by the maximum ratio combining algorithm; Step 5: Multi-antenna selection. By modeling and intelligently analyzing the micro-Doppler echo signal of millimeter-wave sensing, the present invention solves the problems of the influence of DC offset, noise, clutter interference, etc. on the sensed echo signal, and realizes the goal of high-precision detection and estimation of mechanical vibration. The millimeter-wave micron-level vibration detection method and system based on multi-antenna selection optimization proposed by the present invention effectively improve the detection accuracy of the system. The present invention realizes non-contact continuous vibration detection based on low-cost, portable, and low-power hardware millimeter-wave sensing technology.

Description

Technical Field

[0001] The present invention belongs to the technical field of vibration detection and wireless sensing, and particularly relates to a micron-level vibration detection method and system based on multi-antenna millimeter wave (mmWave) sensing. Background Art

[0002] Millimeter wave refers to the frequency domain of 30 - 300 GHz (wavelength of 1 - 10 mm), which has characteristics such as short wavelength, wide bandwidth, high resolution, narrow antenna beam, strong penetration ability and anti-interference ability. Millimeter wave sensors can achieve high-resolution, high-sensitivity, and high-accuracy target perception and detection, so they are widely used in fields such as information security, medical health, autonomous driving, safety search and rescue, etc., and have very important practical value.

[0003] Mechanical vibration is the most common phenomenon in industry. Equipment damage or failure usually leads to changes in abnormal vibration characteristics of the equipment. By measuring vibration, that is, measuring the amplitude and frequency of vibration, the task operation status of machines in various industrial scenarios can be checked, abnormalities can be identified, and faults can be diagnosed. Existing vibration measurement technologies often rely on professional sensors directly installed on vibrating objects, such as piezoelectric sensors, or optical devices such as laser vibrometers. However, the deployment and maintenance of such designs are relatively complex and costly. Summary of the Invention

[0004] Aiming at the deficiencies in the prior art, the present invention proposes a micron-level vibration detection method and system based on multi-antenna millimeter wave sensing. By modeling and intelligently analyzing the micro-Doppler echo signal of millimeter wave sensing, the present invention aims to solve the problem of realizing the detection and estimation of high-precision mechanical vibration under the influence of DC offset, noise, and clutter interference on the sensing echo signal, and effectively improves the detection accuracy of the system. The present invention realizes non-contact continuous vibration detection based on low-cost, portable, and low-power hardware millimeter wave sensing technology.

[0005] The object of the present invention is achieved by the following technical solutions:

[0006] A micron-level vibration detection method based on multi-antenna millimeter wave sensing specifically includes the following steps:

[0007] Step 1: Millimeter-wave sensing of mechanical vibration. Specifically, a multi-antenna millimeter-wave sensor is used to detect the vibration of mechanical equipment. This detection is wireless and non-contact. A Frequency-Modulated Continuous Wave (FMCW) is transmitted to the mechanical equipment, and then the FMCW reflected by the mechanical equipment with mechanical vibration information is received. The millimeter-wave probe converts the periodic linear frequency modulation signal of the received FMCW into an Intermediate Frequency (IF) signal, and calculates the results of the fast Fourier transform of the IF signals of multiple linear frequency modulation periods through sampling and calculation to calculate the vibration displacement information.

[0008] Step 2: Constellation DC (Direct Current) offset correction. DC offset will generate more phase noise, causing phase distortion in the extracted phase in the target range image containing vibration information. Therefore, it is necessary to correctly estimate the DC bias to reduce the phase harmonic level. In order to move the offset constellation to the origin, the present invention preferably estimates the center of the constellation through a non-linear least squares algorithm.

[0009] Step 3: Phase extraction and noise removal based on AD (Arctangent Demodulation). The radar signal phase generated due to the oscillating motion of the target is extracted through the AD algorithm, and the extracted phase is phase-unwrapped to obtain the true phase change of the target. At the same time, to reduce the noise impact brought by phase shift, it is removed through a noise cancellation method.

[0010] Step 4: Maximal-Ratio Combining (MRC) processing. In order to improve the signal-to-noise ratio of the vibration detection signal and suppress noise, the present invention proposes to optimize and combine the target vibration echo signals obtained from different channels based on multiple antennas using the MRC method.

[0011] Step 5: Multi-antenna selection. Since the weight vector of MRC depends on the cross-correlation matrix between input signals, achieving the best MRC performance depends on the selection of input signals. The present invention preferably performs multi-antenna selection through a weight calculation selection algorithm to identify and eliminate bad channels to improve MRC performance.

[0012] Preferably, the specific process of converting the echo signal into an intermediate frequency signal and calculating the vibration displacement in Step 1 is as follows:

[0013] The FWCM signal x T (t) sent by the millimeter-wave sensor is:

[0014]

[0015] where A Tis the amplitude of the millimeter wave and also the transmission power, f c is the starting frequency of the FWCM, B is the signal bandwidth, T c is the signal frequency modulation period, is the noise. Considering the use of multi-antenna transmission technology, the position of the m th th transmitting antenna is d m =(m - 1)d TX , d TX is the distance between two transmitting antennas. Thus, the transmitted signal of the m th th transmitting antenna is:

[0016]

[0017] where T r is the switching time between transmitting antennas, θ TX is the angle between the transmitting antenna and the target. At the n th th receiving antenna, the position is d n =(n - 1)d RX , d RX is the distance between two receiving antennas. At the n th th receiving antenna, the received echo signal is:

[0018]

[0019] where a n is the reflection coefficient of the target, t d =2R(t) / c is the time delay between the target at distance R(t), θ RX is the angle between the receiving antenna and the target. The expression of the IF signal obtained after mixing the received signal and the transmitted signal through I / Q is as follows:

[0020]

[0021] A mn is the received power of the (m th , n th ) antenna, f b =B[(m - 1)T r +t d / T c is the center frequency.

[0022] Based on formula (4), the intermediate frequency signal will be further sampled by an ADC (Analog-to-Digital Converter) to obtain the target range image matrix R through 2D-FFT transformation, obtaining the phase offset Im(t)+dc i and Re(t)+dc rare the real and imaginary parts of the range image, where dc i , dc r are the DC offsets of the real and imaginary parts respectively.

[0023] Preferably, the specific process of constellation DC offset correction in step two is as follows:

[0024] According to the FMCW radar, in order to perform DC compensation on the target range image, the IQ constellation needs to be moved to the origin. The center of the constellation is estimated using the Non-Linear Least Square (NLLS) algorithm, and through algebraic simplification, it can be converted to a linear least square estimation as follows:

[0025]

[0026] where A and b are estimation coefficients related to the sampled values, and y = [CIm(C)Re(C)] T , where C is the origin and radius of the IQ constellation, and Im(C) and Re(C) are the real and imaginary parts of the center.

[0027] Preferably, the specific process of AD-based phase extraction and noise removal in step three is as follows:

[0028] After the constellation DC offset correction of the vibrating target, the AD algorithm is used to perform phase unwrapping on the extracted phase to eliminate phase shifts greater than ±π.

[0029]

[0030] The un-unwrapped differential phase d(m) is affected by phase wrapping errors caused by noise, especially when the phase shift is close to ±π but does not exceed ±π. At this time, phase unwrapping cannot eliminate the influence of the phase shift. This kind of pulse-like noise can be removed by calculating the forward phase shift difference d(m) - d(m + 1) and the backward phase shift difference d(m) - d(m - 1) of each d(m). If the value exceeds a specific threshold, d(m) is replaced by the interpolation within d interp (m), and d interp (m) is obtained by the three-point Lagrange interpolation algorithm.

[0031] Preferably, the specific process of MRC processing in step four is as follows:

[0032] The signal vector of the multi-antenna received signal is X(t), and the correlation matrix is R XX = <X(t)X H (t)>, where the superscript H represents the complex conjugate transpose operator. Through eigenvalue decomposition, we can obtain

[0033] R XX = [v1v2…vN diag{σ1 σ2 … σ N} [v1 v2 … v N (7)

[0034] where σ1 σ2 … σ N are eigenvalues, and v i are the eigenvectors corresponding to the eigenvalues σ i . MRC uses the first eigenvector v1 as the weight vector w = v1, thereby obtaining where the eigenvector and the weight vector are normalized values |w| 2 = 1, |v i | 2 = 1.

[0035] Preferably, the specific process of multi - antenna selection in step five is as follows:

[0036] After obtaining the weight vector w, find the w with the largest absolute value |w| max , given the threshold w th = αw max , α = [0, 1], remove the i - th channel where |w i | < w th .

[0037] The present invention also discloses a micron - level vibration detection system based on multi - antenna millimeter - wave sensing, including the following modules connected in sequence:

[0038] Sensing module: used to sense the millimeter - wave of mechanical vibration and obtain the displacement information of the vibration;

[0039] Correction module: correct the displacement information of the vibration for constellation DC offset to obtain the center of the constellation;

[0040] Phase extraction and noise removal module: extract the phase and remove the noise using the AD algorithm;

[0041] Noise suppression module: suppress the noise using the maximum ratio combining algorithm to obtain the target vibration echo signal;

[0042] Multi - antenna selection module: perform multi - antenna selection through the weight calculation selection algorithm.

[0043] Preferably, the sensing module is specifically as follows:

[0044] The FWCM signal x T (t) emitted by the millimeter - wave sensor is:

[0045]

[0046] where A T is the amplitude of the millimeter - wave and also the transmission power, fc is the starting frequency of FWCM, B is the signal bandwidth, and T c is the signal frequency modulation period, is the noise; The multi-antenna transmission technology is adopted. The position of the m th th transmitting antenna is d m =(m - 1)d TX , d TX is the distance between two transmitting antennas. Thus, the transmitted signal of the m th th transmitting antenna is:

[0047]

[0048] where T r is the switching time between transmitting antennas, and θ TX is the angle between the transmitting antenna and the target; At the n th th receiving antenna, the position is d n =(n - 1)d RX , d RX is the distance between two receiving antennas. At the n th th receiving antenna, the received echo signal is:

[0049]

[0050] where a n is the reflection coefficient of the target, and t d =2R(t) / c is the time delay of the target at a distance of R(t), and θ RX is the angle between the receiving antenna and the target; The expression of the IF signal obtained after the received signal and the transmitted signal pass through I / Q mixing is as follows:

[0051]

[0052] A mn is the received power of the (m th , n th ) antenna, and f b =B[(m - 1)T r +t d / T c is the center frequency;

[0053] Based on formula (4), the intermediate frequency signal is further subjected to ADC sampling to obtain the target range image matrix R through 2D-FFT transformation, and the phase offset

[0054] Im(t)+dc i and Re(t)+dc r are the real part and the imaginary part of the range image respectively, where dc i , dcr DC offsets that are the real part and the imaginary part respectively.

[0055] Preferably, the correction module is as follows:

[0056] According to the FMCW radar, in order to perform DC compensation on the target range profile, it is necessary to move the IQ constellation to the origin, and use the non - linear least - squares estimation algorithm to estimate the center of the constellation. After algebraic simplification, it is converted to the linear least - squares estimation as follows:

[0057]

[0058] where A and b are estimation coefficients related to the sampling values, and y = [CIm(C)Re(C)] T , where C is the origin and radius of the IQ constellation, and Im(C) and Re(C) are the real part and the imaginary part of the center.

[0059] Preferably, the phase extraction and noise removal module is as follows:

[0060] After the DC offset correction of the constellation of the vibrating target, the AD algorithm is used to perform phase unwrapping on the extracted phase to eliminate phase shifts greater than ±π.

[0061]

[0062] The un - unwrapped differential phase d(m) will be affected by the phase - wrapping error caused by noise. The noise is removed by calculating the forward phase - shift difference d(m)-d(m + 1) and the backward phase - shift difference d(m)-d(m - 1) of each d(m). If the value exceeds a specific threshold, d(m) is replaced by the interpolation within d interp (m), and d interp (m) is obtained by the three - point Lagrange interpolation algorithm. Preferably, the noise suppression module is as follows:

[0063] The signal vector of the multi - antenna received signal is X(t), and the correlation matrix is R XX =<X(t)X H (t)>, where the superscript H represents the complex conjugate transpose operator; through eigenvalue decomposition, we get

[0064] R XX =[v1v2…v N diag{σ1σ2…σ N}[v1v2…v N (7)

[0065] where σ1σ2…σ N are the eigenvalues, and v i are the eigenvectors corresponding to the eigenvalues σ iThe eigenvector; MRC uses the first eigenvector v1 as the weight vector w = v1, thus obtaining where the eigenvector and the weight vector are normalized values |w| 2 = 1, |v i | 2 = 1.

[0066] Preferably, the multi-antenna selection module is as follows:

[0067] After obtaining the weight vector w, find the w with the largest absolute value |w| max , given the threshold w th = αw max , α = [0,1], remove the i-th channel with |w i | < w th of the i-th channel.

[0068] The beneficial effects of the present invention are as follows:

[0069] The present invention proposes a method and system for micron-level vibration detection based on multi-antenna millimeter-wave sensing. By modeling and intelligently analyzing the micro-Doppler echo signal of millimeter-wave sensing, it solves the problem of the influence of DC offset, noise, and clutter interference on the sensing echo signal, and realizes the detection and estimation goal of high-precision mechanical vibration. The method and system for millimeter-wave micron-level vibration detection based on multi-antenna selection optimization proposed by the present invention can effectively improve the detection accuracy.

[0070] The present invention realizes non-contact continuous vibration detection based on low-cost, portable, and low-power hardware millimeter-wave sensing technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 is a flowchart of the method for micron-level vibration detection based on multi-antenna millimeter-wave sensing in Embodiment 1 of the present invention.

[0072] Figure 2 is a block diagram of the system for micron-level vibration detection based on multi-antenna millimeter-wave sensing in Embodiment 2 of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0073] The present invention will be described in detail below according to the drawings and preferred embodiments, and the purpose and effect of the present invention will become more apparent. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0074] Embodiment 1

[0075] As Figure 1 shown, a method for micron-level vibration detection based on multi-antenna millimeter-wave sensing in this embodiment is as follows:

[0076] Step 1. Use a millimeter-wave radar from Texas Instruments to collect vibration signals of the device. The specific process of converting the echo signal into an intermediate-frequency signal and calculating the vibration displacement is as follows:

[0077] The FWCM signal x T (t) emitted by the millimeter-wave sensor is:

[0078]

[0079] where A T is the amplitude of the millimeter wave and also the transmission power, f c is the starting frequency of the FWCM, B is the signal bandwidth, T c is the signal frequency modulation period, is the noise. Considering the use of multi-antenna transmission technology, the position of the m th th transmitting antenna is d m =(m - 1)d TX , d TX is the distance between two transmitting antennas. Thus, the transmission signal of the m th th transmitting antenna is:

[0080]

[0081] where T r is the switching time between transmitting antennas, θ TX is the angle between the transmitting antenna and the target. At the n th th receiving antenna, the position is d n =(n - 1)d RX , d RX is the distance between two receiving antennas. At the n th th receiving antenna, the received echo signal is:

[0082]

[0083] where a n is the reflection coefficient of the target, t d =2R(t) / c is the time delay of the target at a distance of R(t), θ RX is the angle between the receiving antenna and the target. The expression of the IF signal obtained after I / Q mixing of the received signal and the transmitted signal is as follows:

[0084]

[0085] A mn is the received power of the (m th , n th ) antenna, f b =B[(m - 1)T r +t d / Tc is the center frequency.

[0086] Based on Equation (4), the intermediate-frequency signal will be further sampled by ADC to obtain the target range image matrix R through 2D-FFT transformation, and the phase offset is obtained

[0087] Im(t)+dc i and Re(t)+dc r are the real and imaginary parts of the range image respectively, where dc i , dc r are the DC offsets of the real and imaginary parts respectively.

[0088] Step 2. Perform constellation DC offset correction on the echo signal. According to the FMCW radar, in order to perform DC compensation on the target range image, the IQ constellation needs to be moved to the origin. The nonlinear least squares estimation NLLS algorithm is used to estimate the center of the constellation, which can be converted to the linear least squares estimation after algebraic simplification as follows:

[0089]

[0090] where A and b are the estimation coefficients related to the sampled values, and y = [CIm(C)Re(C)] T , where C is the origin and radius of the IQ constellation, and Im(C) and Re(C) are the real and imaginary parts of the center.

[0091] Step 3. Perform AD-based phase extraction and noise removal. After the constellation DC offset correction of the vibrating target, the AD algorithm is used to unwrap the extracted phase to eliminate the phase shift greater than ±π.

[0092]

[0093] The unrolled differential phase d(m) will be affected by the phase wrapping error caused by noise. Especially when the phase shift is close to ±π but does not exceed ±π, the phase unwrapping cannot eliminate the influence of the phase shift. This kind of pulse-like noise can be removed by calculating the forward phase shift difference d(m)-d(m + 1) and the backward phase shift difference d(m)-d(m - 1) of each d(m). If the value exceeds a specific threshold, d(m) is replaced by the interpolation within d interp (m), and d interp (m) is obtained by the three-point Lagrange interpolation algorithm.

[0094] Step 4. To further improve the signal-to-noise ratio of the echo, perform MRC processing on the echo signal. Let the signal vector of the multi-antenna received signal be X(t), and the correlation matrix be R XX =<X(t)X H(t), where the superscript H represents the complex conjugate transpose operator. Through eigenvalue decomposition, we can obtain

[0095] R XX = [v1 v2 … v N diag{σ1 σ2 … σ N} [v1 v2 … v N (7)

[0096] where σ1 σ2 … σ N are the eigenvalues, and v i are the eigenvectors corresponding to the eigenvalues σ i . MRC uses the first eigenvector v1 as the weight vector w = v1, thereby obtaining where the eigenvector and the weight vector are normalized values |w| 2 = 1, |v i | 2 = 1.

[0097] Step 5. Perform multi-antenna selection. After obtaining the weight vector w, find the w with the largest absolute value |w| max , given the threshold w th = αw max , α = [0,1], and remove the i-th channel where |w i | < w th .

[0098] By implementing the above steps, the vibration signal of the device is obtained, thereby achieving an accurate assessment of the device state.

[0099] Embodiment 2

[0100] As Figure 2 shown, this embodiment discloses a micron-level vibration detection system based on multi-antenna millimeter-wave sensing, including the following modules connected in sequence:

[0101] Sensing module: used to sense the millimeter-wave of mechanical vibration and obtain the displacement information of the vibration; specifically as follows:

[0102] The FWCM signal x T (t) emitted by the millimeter-wave sensor is:

[0103]

[0104] where A T is the amplitude of the millimeter-wave and also the transmission power, f c is the starting frequency of the FWCM, B is the signal bandwidth, T c is the signal frequency modulation period, is the noise; adopting the multi-antenna transmission technology, the position of the m th th transmitting antenna is dm =(m - 1)d TX , d TX is the distance between two transmitting antennas, whereby the transmission signal of the m th th transmitting antenna is: x m (t) =

[0105]

[0106] where T r is the switching time between transmitting antennas, and θ TX is the angle between the transmitting antenna and the target; at the n th th receiving antenna, the position is d n =(n - 1)d RX , d RX is the distance between two receiving antennas. At the n th th receiving antenna, the received echo signal is:

[0107]

[0108] where a n is the reflection coefficient of the target, t d = 2R(t) / c is the time delay of the target at a distance of R(t), and θ RX is the angle between the receiving antenna and the target; the expression of the IF signal obtained after I / Q mixing of the received signal and the transmitted signal is as follows:

[0109]

[0110] A mn is the received power of the (m th , n th ) antenna, and f b = B[(m - 1)T r + t d / T c is the center frequency;

[0111] Based on formula (4), the intermediate frequency signal is further subjected to ADC sampling to obtain the target range image matrix R through 2D-FFT transformation, and the phase offset

[0112] Im(t) + dc i and Re(t) + dc r are the real and imaginary parts of the range image respectively, where dc i , dc r are the DC offsets of the real and imaginary parts respectively.

[0113] Correction module: Correct the DC offset of the constellation with the displacement information of the vibration to obtain the center of the constellation. Specifically as follows:

[0114] According to the FMCW radar, in order to perform DC compensation on the target range profile, it is necessary to move the IQ constellation to the origin, and use the non-linear least squares estimation algorithm to estimate the center of the constellation. After algebraic simplification, it is converted into a linear least squares estimation as follows:

[0115]

[0116] where A and b are estimation coefficients related to the sampling values, and y = [CIm(C)Re(C)] T , where C is the origin and radius of the IQ constellation, and Im(C) and Re(C) are the real and imaginary parts of the center.

[0117] Phase extraction and noise removal module: Use the AD algorithm to extract the phase and remove the noise. Specifically as follows:

[0118] After correcting the DC offset of the constellation of the vibrating target, use the AD algorithm to perform phase unwrapping on the extracted phase to eliminate phase shifts greater than ±π.

[0119]

[0120] The un-unwrapped differential phase d(m) will be affected by phase wrapping errors caused by noise. The noise is removed by calculating the forward phase shift difference d(m)-d(m + 1) and the backward phase shift difference d(m)-d(m-1) of each d(m). If the value exceeds a specific threshold, d(m) is replaced by the interpolation within d interp (m), and d interp (m) is obtained by the three-point Lagrange interpolation algorithm.

[0121] Noise suppression module: Use the maximum ratio combining algorithm to suppress the noise and obtain the target vibration echo signal. Specifically as follows:

[0122] The signal vector of the multi-antenna received signal is X(t), and the correlation matrix is R XX =<X(t)X H (t)>, where the superscript H represents the complex conjugate transpose operator; obtain by eigenvalue decomposition

[0123] R XX =[v1v2…v N diag{σ1σ2…σ N}[v1v2…v N (7)

[0124] where σ1σ2…σ N are the eigenvalues, and v i is the eigenvector corresponding to the eigenvalue σi Eigenvector; MRC uses the first eigenvector v1 as the weight vector w = v1, thereby obtaining where the eigenvector and the weight vector are normalized values |w| 2 = 1, |v i | 2 = 1.

[0125] Multi-antenna selection module: performs multi-antenna selection through a weight calculation and selection algorithm; specifically as follows:

[0126] After obtaining the weight vector w, find the w with the largest absolute value |w| max , given the threshold w th = αw max , α = [0, 1], remove the i-th channel with |w i | < w th .

[0127] Those of ordinary skill in the art can understand that the above are only preferred examples of the invention and are not used to limit the invention. Although the invention has been described in detail with reference to the foregoing examples, for those skilled in the art, they can still modify the technical solutions described in the foregoing examples, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, etc. made within the spirit and principle of the invention shall be included within the protection scope of the invention.

Claims

1. A micron-level vibration detection method based on multi-antenna millimeter-wave sensing, characterized in that It includes the following steps: Step 1: Millimeter-wave sensing of mechanical vibration; Step 2: Constellation DC offset correction; Step 3: Phase extraction and noise removal based on the AD algorithm; Step 4: Processing by the maximum ratio combining algorithm; Step 5: Multi-antenna selection; The specific steps of the millimeter-wave sensing of mechanical vibration in Step 1 are as follows: The FWCM signal x T (t) emitted by the millimeter-wave sensor is as follows: Among them, A T is the amplitude of the millimeter wave and also the transmission power, f c is the starting frequency of the FWCM, B is the signal bandwidth, T c is the signal frequency modulation period, is the noise; The multi-antenna transmission technology is adopted. The position of the m th th transmitting antenna is d m =(m - 1)d TX , d TX is the distance between two transmitting antennas. Thus, the transmission signal of the m th th transmitting antenna is: x m (t) = where T r is the switching time between the transmitting antennas, and θ TX is the angle between the transmitting antenna and the target; at the nth th the position of the receiving antenna is d n = (n - 1)d RX where d RX is the distance between two receiving antennas, and at the nth th the received echo signal of the receiving antenna is: where a n is the reflection coefficient of the target, and t d = 2R(t) / c is the time delay of the target between distances R(t), and θ RX is the angle between the receiving antenna and the target; the expression of the IF signal obtained after I / Q mixing of the received signal and the transmitted signal is as follows: A mn is the received power of the antenna for (m th , n th ), and f b = B[(m - 1)T r + t d / T c is the center frequency; Based on Equation (4), the intermediate frequency signal is further sampled by ADC to obtain the target range image matrix R through 2D-FFT transformation, and the phase offset is obtained. and Re(t)+dc r are the real part and the imaginary part of the range image respectively, where dc i , dc r are the DC offsets of the real part and the imaginary part respectively.

2. The micron-level vibration detection method based on multi-antenna millimeter-wave sensing according to claim 1, wherein The specific steps of the constellation DC offset correction in Step 2 are as follows: According to the FMCW radar, in order to perform DC compensation on the target range image, the IQ constellation needs to be moved to the origin, and the nonlinear least squares estimation algorithm is used to estimate the center of the constellation. After algebraic simplification, it is converted to the linear least squares estimation as follows: where A and b are estimation coefficients related to the sampling values, and y = [C Im(C) Re(C)] T , where C is the origin and radius of the IQ constellation, and Im(C) and Re(C) are the real and imaginary parts of the center respectively.

3. The micron-level vibration detection method based on multi-antenna millimeter-wave sensing according to claim 2, wherein The specific process of the phase extraction and noise removal based on AD in Step 3 is as follows: After the constellation DC offset correction of the vibrating target, the AD algorithm is used to perform phase unwrapping on the extracted phase to eliminate phase shifts greater than ±π. The unrolled differential phase d(m) is affected by phase-wrapping errors caused by noise. The noise is removed by calculating the forward phase-shift difference d(m) - d(m+1) and the backward phase-shift difference d(m) - d(m-1) for each d(m). If the value exceeds a specific threshold, d(m) is replaced by an interpolated value within d interp (m), and d interp (m) is obtained by a three-point Lagrange interpolation algorithm.

4. The method for micron-level vibration detection based on multi-antenna millimeter-wave sensing according to claim 3, wherein The specific process of the processing by the maximum ratio combining algorithm in Step 4 is as follows: The signal vector of the multi-antenna received signal is X(t), and the correlation matrix is R XX =<X(t)X H (t)>, where the superscript H represents the complex conjugate transpose operator; obtained by eigenvalue decomposition R XX = [v1 v2 … v N diag{σ1 σ2 … σ N} [v1 v2 … v N (7) where σ1σ2…σ N are the eigenvalues, and v i is the eigenvector corresponding to the eigenvalue σ i ; the maximum ratio combining algorithm uses the first eigenvector v1 as the weight vector w = v1, thereby obtaining where the eigenvector and the weight vector are normalized values |w| 2 = 1, |v i | 2 = 1.

5. The method for micron-level vibration detection based on multi-antenna millimeter-wave sensing according to claim 4, wherein The specific process of the multi-antenna selection in Step 5 is as follows: After obtaining the weight vector w, find the w with the largest absolute value |w| max , given the threshold w th = αw max , α = [0,1], remove the i-th channel of |w i |< w th ​ 6. A micron-level vibration detection system based on multi-antenna millimeter-wave sensing, characterized in that, It includes the following modules: Sensing module: Used to sense the millimeter-wave of mechanical vibration and obtain the displacement information of the vibration; Correction module: Correct the constellation DC offset of the displacement information of the vibration to obtain the center of the constellation; Phase extraction and noise removal module: Use the AD algorithm to extract the phase and remove the noise; Noise suppression module: Suppress the noise by the maximum ratio combining algorithm to obtain the target vibration echo signal; Multi-antenna selection module: Perform multi-antenna selection through the weight calculation selection algorithm; The sensing module is as follows: The FWCM signal x T (t) emitted by the millimeter-wave sensor is as follows: Among them, A T is the amplitude of the millimeter wave and also the transmission power, f c is the starting frequency of the FWCM, B is the signal bandwidth, T c is the signal frequency modulation period, is the noise; The multi-antenna transmission technology is adopted. The position of the m th th transmitting antenna is d m =(m - 1)d TX , d TX is the distance between two transmitting antennas. Thus, the transmission signal of the m th th transmitting antenna is: x m (t) = where T r is the switching time between transmitting antennas, and θ TX is the angle between the transmitting antenna and the target; at the n th the position of the receiving antenna is d n = (n - 1)d RX where d RX is the distance between two receiving antennas, and at the n th the received echo signal of the receiving antenna is: where a n is the reflection coefficient of the target, and t d = 2R(t) / c is the time delay between the target at a distance R(t), and θ RX is the angle between the receiving antenna and the target; the expression of the IF signal obtained after the received signal and the transmitted signal are mixed by I / Q is as follows: A mn is the received power of the antenna for (m th , n th ), f b = B[(m - 1)T r + t d / T c is the center frequency; Based on Equation (4), the intermediate-frequency signal is further sampled by ADC to obtain the target range image matrix R through 2D-FFT transformation, and the phase offset is obtained. Im(t)+dc i and Re(t)+dc r are the real part and the imaginary part of the range image respectively, where dc i , dc r are the DC offsets of the real part and the imaginary part respectively.

7. The micron-level vibration detection system based on multi-antenna millimeter-wave sensing according to claim 6, characterized in that, The correction module is as follows: According to the FMCW radar, in order to perform DC compensation on the target range image, the IQ constellation needs to be moved to the origin, and the nonlinear least squares estimation algorithm is used to estimate the center of the constellation. After algebraic simplification, it is converted to the linear least squares estimation as follows: Among them, A and b are estimated coefficients related to the sampling value, and y = [C Im(C) Re(C)] T , where C is the origin and radius of the IQ constellation, and Im(C) and Re(C) are the real and imaginary parts of the center respectively.

8. The micron-level vibration detection system based on multi-antenna millimeter-wave sensing according to claim 7, wherein The phase extraction and noise removal module is as follows: After the constellation DC offset correction of the vibrating target, the AD algorithm is used to perform phase unwrapping on the extracted phase to eliminate phase shifts greater than ±π. The unwrapped differential phase d(m) is affected by phase wrapping errors caused by noise, which is removed by calculating the forward phase shift difference d(m) - d(m+1) and the backward phase shift difference d(m) - d(m-1) for each d(m). If the value exceeds a specific threshold, d(m) is replaced by an interpolated value within d interp (m), and d interp (m) is obtained by a three-point Lagrange interpolation algorithm.

Images