A vibration detection method and terminal based on millimeter wave radar

Abstract

The present invention discloses a vibration detection method and terminal based on millimeter wave radar, and the steps are as follows: locate and identify the vibrating object to obtain the original signal sequence; adopt a curve segment fitting processing method on the IQ plane, use the static point as the curve dividing point to divide the original signal sequence into two segments on the IQ plane, fit them respectively, restore the positive and negative displacements, and obtain the original vibration signal; use a convolutional network with a jump connection layer to process the time-frequency graph of the original vibration signal, remove additive noise and multiplicative noise, and obtain an enhanced vibration signal; judge the enhanced vibration signal, and if an abnormality occurs, issue a processing notification. The present invention adopts non-invasive sensing technology, does not need to modify the machine, and realizes vibration sensing through millimeter wave wireless radio frequency signals.

Description

Technical Field

[0001] The present invention belongs to the field of millimeter wave radar passive sensing and industrial Internet technology, and specifically relates to a vibration detection method and terminal based on millimeter wave radar. Background Art

[0002] Vibration is one of the most common phenomena in industrial scenarios. Since equipment damage or abnormality will directly cause abnormal vibration characteristics of industrial equipment, the vibration of industrial equipment can usually clearly reflect the internal working status of the equipment. However, due to the complexity of actual industrial scenarios, it is difficult to detect machine failures in time by relying solely on regular inspections by staff using methods such as listening to sounds and measuring vibrations, which may lead to major production accidents. Therefore, we need an intelligent automatic machine vibration detection system to notify and warn staff in time when the machine produces abnormal vibrations.

[0003] The main machine vibration detection solutions currently include the following: 1) Sensor-based methods. This method mainly measures the vibration of the machine by installing professional sensors on the machine. Among them, the most widely used is the piezoelectric ceramic vibration sensor. The main disadvantage of this method is that it can only sense a single device, and its deployment requires additional power supply and transmission lines, which is an invasive sensing method. 2) Optical device-based methods. This method uses a laser vibrometer to sense vibration. Due to the characteristics of the laser, this method has extremely high displacement measurement accuracy. However, in actual deployment, the optical path needs to be direct and unobstructed, which is almost impossible to achieve in a complex factory environment. In addition, high-precision laser sensors are usually expensive, which limits their large-scale application.

[0004] Therefore, based on the above considerations, it is necessary to propose a new vibration detection method to achieve non-intrusive, high-precision, and low-cost instant machine vibration detection. Summary of the invention

[0005] In view of the above-mentioned deficiencies in the prior art, the purpose of the present invention is to provide a vibration detection method and terminal based on millimeter-wave radar, so as to solve the problem that in the existing machine vibration detection technology, it is difficult to detect and deal with unexpected situations in time only by relying on regular inspections by staff; the present invention adopts non-invasive sensing technology, does not require modification of the machine, and realizes vibration sensing through millimeter-wave wireless radio frequency signals.

[0006] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0007] A vibration detection method based on millimeter wave radar of the present invention comprises the following steps:

[0008] 1) Positioning and identifying vibrating objects: Coarse-grained positioning of objects is performed through distance Fourier transform, and then fine-grained positioning of objects after coarse-grained positioning is performed through Chirp-z transform. The positions of several objects on a two-dimensional plane identified by the arrival angle measurement method are screened through Doppler Fourier transform to select a set of vibration measurement target objects, and the original signal sequence is obtained for the objects in the set;

[0009] 2) For the original signal sequence obtained in step 1), a curve segment fitting processing method is adopted on the IQ (Inphase-Quadrature) plane, and the original signal sequence is divided into two segments on the IQ plane with the static point as the curve dividing point, and the fitting is performed separately to restore the positive and negative displacements and obtain the original vibration signal;

[0010] 3) using a convolutional network with a skip connection layer to process the time-frequency graph of the original vibration signal obtained in step 2), remove additive noise and multiplicative noise, and obtain an enhanced vibration signal;

[0011] 4) The enhanced vibration signal obtained in step 3) is judged, and if an abnormality occurs, a processing notification is issued.

[0012] Furthermore, the coarse-grained positioning in step 1) specifically includes: detecting objects in the environment by using a millimeter-wave radar, and the transmission signal S of the millimeter-wave radar Tx (t) and the received signal S Rx (t) is expressed as follows:

[0013] S Tx (t) = exp[j(2πf c t+πKt 2 )]

[0014] S Rx (t) = αS Tx [t-2R(t) / c]

[0015] Transmit signal S Tx (t) and the received signal S Rx (t) is processed by the onboard mixer to obtain the intermediate frequency signal s(t), the formula is as follows:

[0016]

[0017] Where α is the attenuation constant, f c is the starting frequency of the continuous frequency modulation wave, K is the modulation slope of the continuous frequency modulation wave, R(t) is the distance between the object and the radar, Δt is the signal propagation time, c is the speed of light, j is the imaginary unit, Represents the signal S TxThe conjugate operation of (t);

[0018] The distance Fourier transform of the intermediate frequency signal s(t) is as follows:

[0019]

[0020] In the formula, X k represents the kth element of the Fourier transform spectrum, x n represents the nth element of the intermediate frequency signal s(t), and N represents the number of Fourier transform points;

[0021] Through the distance Fourier transform, each frequency peak on the spectrum corresponds to the actual position of an object in the environment, and the coarse-grained positioning of the object is completed through the distance Fourier transform.

[0022] Furthermore, the fine-grained positioning in step 1) specifically includes: the ideal frequency f corresponding to the real position of the object ideal The peak frequency f on the Fourier spectrum of the distance FFT There are the following relations: Get f ideal With f FFT There is a frequency deviation between By setting the parameters of Chirp-z transformation, the frequency deviation can be reduced to Therefore, through the relationship between frequency and distance, the distance d between the object and the millimeter-wave radar is obtained, completing the fine-grained positioning of the object.

[0023] Furthermore, the intermediate frequency signal s(t) obtained in step 1) is a signal set received by n antennas of the millimeter wave radar, that is, s(t)=[s 1 (t),s 2 (t),s 3 (t),……s n (t)], for a time t, the arrival angle θ is calculated by the phase difference of n antennas, and combined with the distance d, an object M is determined on the two-dimensional plane i =[d i ,θ i ].

[0024] Furthermore, the step 1) specifically includes: after calculating the distance d and the angle θ, a set of several objects M=[M 1 ,M 1 ,M 1 ,……,M n], for the objects in the set, obtain their original signal sequence, perform Doppler Fourier transform with a time window of 3 to 5 seconds, obtain the range-Doppler velocity spectrum, and filter out the stationary objects and select the vibration measurement target objects in the following way:

[0025] The absolute value of the Doppler velocity in several consecutive time windows is greater than the threshold, and the variance does not exceed half of the average value of the Doppler velocity in 10 windows;

[0026] In several consecutive time windows, the Doppler velocity is near the two peaks of v and -v;

[0027] For objects that meet the above conditions, we take them as vibration sensing targets and get the set of vibration measurement target objects: For each object in the set, an original signal sequence S′ is obtained.

[0028] Furthermore, the step 2) specifically includes: processing the original signal sequence S′ of the object obtained in step 1) on the IQ plane to restore the original vibration signal, as follows:

[0029] 21) Through S center = S′-Mean(S′) completes the signal centering and the stationary point P static The position moves to the origin - O; where S center is the original signal sequence after centering, Mean(S′) is the mean of the original signal sequence;

[0030] 22) Initialize two sampling point sets C far ,C near , set C far represents the set of sampling points away from the radar direction, C near Represents a set of sampling points close to the radar direction; the original signal sequence S after centering center Take the sampling points with positive Doppler velocity and far from the origin and put them into set C far For the sequence S center Traverse the point P in Make Time (generally set ), put point P into set C far Repeat the above process until there are no points to add to the set C. far Put the remaining points into set C near In this way, we can obtain the set C of sampling points on both sides of the stationary point far ,C near , each set is represented as an arc on the IQ plane;

[0031] 23) The two arcs are fitted by the least squares method, and the two corresponding centers O are obtained by fitting. far , O near ; Restore the vibration and restore the original vibration signal V through the following formula:

[0032]

[0033] Where V(t) is the original vibration information of the object at time t, λ is the central wavelength of the millimeter-wave radar, and ∠O near OP(t),∠O far OP(t) is the angle between the sampling point and the stationary point in the two sets at time t relative to their respective centers.

[0034] Furthermore, the step 3) specifically includes: for the original vibration signal V obtained in step 2), using a deep learning-based model to enhance it, the model uses a multi-layer convolution-deconvolution network to extract information and sets a jump connection layer at the corresponding position, wherein the relationship between the layers is expressed as:

[0035]

[0036]

[0037] in, is the output of the previous layer and the input of the current layer, is the parameter of the current layer network, is the bias of the current layer, f is the selected activation function, is the output of the convolutional layer, x i-relu is the output of the relu layer; the model selects cross entropy as the loss function; after the network training is completed, the time-frequency graph of the original vibration signal V to be enhanced is input into the network to obtain the enhanced time-frequency graph, and then restored to the enhanced vibration signal through inverse short-time Fourier transform.

[0038] Furthermore, the step 4) specifically includes:

[0039] 41) Taking a certain time as a time window, calculate the enhanced vibration signal in step 3) The vibration amplitude Amp and the vibration frequency fre;

[0040] 42) Take several time windows as one frame and calculate the average value μ of the vibration amplitude Amp With variance σ Amp , the average value of the vibration frequency μ fre and variance σ fre ;

[0041] 43) For each time window, compare it with the previous frame. When Amp>μ Amp +3σ Amp or fre>μ fre +3σ fre When the error occurs, it is considered abnormal and a notification is issued.

[0042] Further, the step 41) specifically includes: for the enhanced vibration signal obtained in step 3) The vibration amplitude Amp is calculated by the following formula:

[0043]

[0044] By enhancing the vibration signal After Fourier transform, the peak frequency with the highest intensity on the spectrum is the object's vibration frequency fre.

[0045] A vibration detection terminal, characterized in that it comprises:

[0046] one or more processors;

[0047] A memory for storing one or more programs;

[0048] When the one or more programs are executed by the one or more processors, the one or more processors are enabled to implement the vibration detection method based on the millimeter wave radar.

[0049] Beneficial effects of the present invention:

[0050] 1. High-precision real-time vibration detection: High-precision, real-time vibration detection of targets in industrial environments with an accuracy of up to millimeter level.

[0051] 2. Accident monitoring: When abnormal vibration is detected, a corresponding warning can be sent to medical staff.

[0052] 3. Low cost: Millimeter-wave radar is used for perception, which is low-cost and easy to deploy in large quantities.

[0053] 4. Contactless and non-invasive sensing: The contactless and non-invasive sensing technology can detect multiple targets at the same time without modifying the machine. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 is a flow chart of the method of the present invention;

[0055] Figure 2 This is a schematic diagram of the signal propagation of millimeter wave radar;

[0056] Figure 3 Schematic diagram of the vibration signal reinforcement network;

[0057] Figure 4 It is a signal segmentation method on the IQ plane;

[0058] Figure 5 This is a schematic diagram of double circle fitting. DETAILED DESCRIPTION

[0059] In order to facilitate the understanding of those skilled in the art, the present invention is further described below in conjunction with embodiments and drawings. The contents mentioned in the implementation modes are not intended to limit the present invention.

[0060] Reference Figure 1 As shown, a vibration detection method based on millimeter wave radar of the present invention comprises the following steps:

[0061] 1) Positioning and identifying vibrating objects: Coarse-grained positioning of objects is performed through distance Fourier transform, and then fine-grained positioning of objects after coarse-grained positioning is performed through Chirp-z transform. The positions of several objects on a two-dimensional plane identified by the arrival angle measurement method are screened through Doppler Fourier transform to select a set of vibration measurement target objects, and the original signal sequence is obtained for the objects in the set;

[0062] Among them, the coarse-grained positioning specifically includes: detecting objects in the environment through millimeter-wave radar, and the transmission signal S of the millimeter-wave radar Tx (t) and the received signal S Rx (t) is expressed as follows:

[0063] S Tx (t) = exp[j(2πf c t+πKt 2 )]

[0064] S Rx (t) = αS Tx [t-2R(t) / c]

[0065] Transmit signal S Tx (t) and the received signal S Rx (t) is processed by the onboard mixer to obtain the intermediate frequency signal s(t), the formula is as follows:

[0066]

[0067] Where α is the attenuation constant, f c is the starting frequency of the continuous frequency modulation wave, K is the modulation slope of the continuous frequency modulation wave, R(t) is the distance between the object and the radar, Δt is the signal propagation time, c is the speed of light, j is the imaginary unit, Represents the signal S Tx The conjugate operation of (t);

[0068] The distance Fourier transform of the intermediate frequency signal s(t) is as follows:

[0069]

[0070] In the formula, X k represents the kth element of the Fourier transform spectrum, x n represents the nth element of the intermediate frequency signal s(t), and N represents the number of Fourier transform points;

[0071] Reference Figure 2 As shown, the transmitted signal reaches the receiving antenna after being reflected by the vibrating object. Since the vibrating object modulates the radio frequency signal, the received signal contains the vibration information of the object.

[0072] Through the distance Fourier transform, each frequency peak on the spectrum corresponds to the actual position of an object in the environment, and the coarse-grained positioning of the object is completed through the distance Fourier transform.

[0073] Among them, fine-grained positioning specifically includes: the ideal frequency f corresponding to the actual position of the object ideal The peak frequency f on the Fourier spectrum of the distance FFT There are the following relations: Get f ideal With f FFT There is a frequency deviation between By setting the parameters of Chirp-z transformation, the frequency deviation can be reduced to Therefore, through the relationship between frequency and distance, the distance d between the object and the millimeter-wave radar is obtained, completing the fine-grained positioning of the object.

[0074] The Chirp-z transform is as follows:

[0075]

[0076] Among them, x n is the nth element of the input, X k is the kth element of the spectrum after Chirp-z transformation, M is the number of points of Chirp-z transformation, θ 0 is the starting sampling angle on the z plane, is the sampling interval angle; by setting the parameters of the Chirp-z transform, it can make a refined transformation of M points within the range of the Fourier transform deviation, let:

[0077]

[0078]

[0079] in, where f lower-CZT is the starting frequency of the Chirp-z transform, f upper-CZT is the end frequency of Chirp-z transform, F s is the signal sampling rate, N is the number of Fourier transform points;

[0080] In addition, the intermediate frequency signal s(t) obtained in step 1) is a set of signals received by n antennas of the millimeter wave radar, that is, s(t)=[s 1 (t),s 2 (t),s 3 (t),……s n (t)], for a time t, the arrival angle θ is calculated by the phase difference of n antennas, and combined with the distance d, an object M is determined on the two-dimensional plane i =[d i ,θ i ].

[0081] After calculating the distance d and the angle θ, we get a set of several objects M = [M 1 ,M 1 ,M 1 ,……,M n ], for the objects in the set, obtain their original signal sequence, perform Doppler Fourier transform with a time window of 3 seconds, obtain the range-Doppler velocity spectrum, and filter out the stationary objects and select the vibration measurement target objects in the following way:

[0082] The absolute value of the Doppler velocity in 10 consecutive time windows is greater than the threshold, and the variance does not exceed half of the average value of the Doppler velocity in the 10 windows;

[0083] In 10 consecutive time windows, the Doppler velocity is near the two peaks of v and -v;

[0084] For objects that meet the above conditions, we take them as vibration sensing targets and get the set of vibration measurement target objects: For each object in the set, an original signal sequence S′ is obtained.

[0085] 2) For the original signal sequence obtained in step 1), a curve segment fitting processing method is adopted on the IQ (Inphase-Quadrature) plane, and the original signal sequence is divided into two segments on the IQ plane with the static point as the curve dividing point, and the fitting is performed separately to restore the positive and negative displacements and obtain the original vibration signal;

[0086] For the original signal sequence S′ of the object obtained in step 1), the original vibration signal is restored on the IQ plane as follows:

[0087] 21) Through S center = S′-Mean(S′) completes the signal centering and the stationary point P static The position moves to the origin - O; where S center is the original signal sequence after centering, Mean(S′) is the mean of the original signal sequence; Figure 5 The signal is shown as being centered on the IQ plane;

[0088] 22)Reference Figure 4 As shown, initialize two sampling point sets C far ,C near , set C far represents the set of sampling points away from the radar direction, C near Represents a set of sampling points close to the radar direction; the original signal sequence S after centering center Take the sampling points with positive Doppler velocity and far from the origin and put them into set C far For the sequence S center Traverse the point P in Make Time (generally set ), put point P into set C far Repeat the above process until there are no points to add to the set C. far Put the remaining points into set C near In this way, we can obtain the set C of sampling points on both sides of the stationary point far ,C near , each set is represented as an arc on the IQ plane;

[0089] 23) The two arcs are fitted by the least square method, and the optimization target S of the square error of the fitting method can be expressed as:

[0090]

[0091] In the formula, (x i ,y i ) are the coordinates of the feature points on the arcs on both sides of the stationary point on the IQ plane, i = 1, 2, ..., n, n is the number of feature points, (x 0 ,y 0 ) is the center of the circle to be fitted, and r is the radius to be fitted; after fitting, two corresponding center points O far , O near ; Restore the vibration and restore the original vibration signal V through the following formula:

[0092]

[0093] Where V(t) is the original vibration information of the object at time t, λ is the central wavelength of the millimeter-wave radar, and ∠O near OP(t),∠O far OP(t) is the angle between the sampling point and the stationary point in the two sets at time t relative to their respective centers.

[0094] 3) Using a convolutional network with a skip connection layer to process the time-frequency graph of the original vibration signal obtained in step 2), remove additive noise and multiplicative noise, and obtain an enhanced vibration signal; specifically:

[0095] For the original vibration signal V obtained in step 2), every 1 second of the signal is transformed into a 128x128 time-frequency diagram and enhanced using a deep learning-based model, such as Figure 3 As shown in the figure, the model uses a multi-layer convolution-deconvolution network to extract information and sets a jump connection layer at the corresponding position, where the relationship between the layers is expressed as:

[0096]

[0097]

[0098] in, is the output of the previous layer and the input of the current layer, is the parameter of the current layer network, is the bias of the current layer, f is the selected activation function, is the output of the convolutional layer, x i-relu is the output of the relu layer; the model selects cross entropy as the loss function; after the network training is completed, the time-frequency graph of the original vibration signal V to be enhanced is input into the network to obtain the enhanced time-frequency graph, and then restored to the enhanced vibration signal through inverse short-time Fourier transform.

[0099] 4) judging the enhanced vibration signal obtained in step 3), and issuing a processing notification if an abnormality occurs;

[0100] 41) Taking a certain time as a time window, calculate the enhanced vibration signal in step 3) The vibration amplitude Amp and the vibration frequency fre;

[0101] 42) Take several time windows as one frame and calculate the average value μ of the vibration amplitude Amp With variance σ Amp , the average value of the vibration frequency μ fre and variance σ fre ;

[0102] 43) For each time window, compare it with the previous frame. When Amp>μ Amp +3σ Amp or fre>μ fre +3σ fre When the error occurs, it is considered abnormal and a notification is issued;

[0103] For the enhanced vibration signal obtained in step 3) The vibration amplitude Amp is calculated by the following formula:

[0104]

[0105] By enhancing the vibration signal After Fourier transform, the peak frequency with the highest intensity on the spectrum is the object's vibration frequency fre.

[0106] The present invention has many specific application paths. The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements can be made without departing from the principle of the present invention. These improvements should also be regarded as the protection scope of the present invention.

Claims

1. A vibration detection method based on millimeter wave radar, characterized in that: Here are the steps: 1) Positioning and identifying vibrating objects: Coarse-grained positioning of objects is performed through distance Fourier transform, and then fine-grained positioning of objects after coarse-grained positioning is performed through Chirp-z transform. The positions of several objects on a two-dimensional plane identified by the arrival angle measurement method are screened through Doppler Fourier transform to select a set of vibration measurement target objects, and the original signal sequence is obtained for the objects in the set; 2) For the original signal sequence obtained in step 1), a curve segment fitting processing method is adopted on the IQ plane, and the original signal sequence is divided into two segments on the IQ plane with the static point as the curve dividing point, and fitting is performed separately to restore the positive and negative displacements to obtain the original vibration signal; 3) using a convolutional network with a skip connection layer to process the time-frequency graph of the original vibration signal obtained in step 2), remove additive noise and multiplicative noise, and obtain an enhanced vibration signal; 4) judging the enhanced vibration signal obtained in step 3), and issuing a processing notification if an abnormality occurs; The coarse-grained positioning in step 1) specifically includes: detecting objects in the environment by using a millimeter-wave radar, and the transmission signal S of the millimeter-wave radar Tx (t) and the received signal S Rx (t) is expressed as follows: S Tx (t)=exp[j(2πf c t+πKt 2 )] S Rx (t)=αS Tx [t-2R(t) / c] Transmit signal S Tx (t) and the received signal S Rx (t) is processed by the onboard mixer to obtain the intermediate frequency signal s(t), the formula is as follows: Where α is the attenuation constant, f c is the starting frequency of the continuous frequency modulation wave, K is the modulation slope of the continuous frequency modulation wave, R(t) is the distance between the object and the radar, Δt is the signal propagation time, c is the speed of light, j is the imaginary unit, Represents the signal S Tx The conjugate operation of (t); The distance Fourier transform of the intermediate frequency signal s(t) is as follows: Where, X k represents the kth element of the Fourier transform spectrum, x n represents the nth element of the intermediate frequency signal s(t), and N represents the number of Fourier transform points; Through the distance Fourier transform, each frequency peak on the spectrum corresponds to the real position of an object in the environment, and the coarse-grained positioning of the object is completed through the distance Fourier transform; The fine-grained positioning in step 1) specifically includes: the ideal frequency f corresponding to the real position of the object ideal The peak frequency f on the Fourier spectrum of the distance FFT There are the following relations: Get f ideal With f FFT There is a frequency deviation between By setting the parameters of Chirp-z transformation, the frequency deviation can be reduced to Therefore, through the relationship between frequency and distance, the distance d between the object and the millimeter wave radar is obtained, and the fine-grained positioning of the object is completed; The step 4) specifically includes: 41) Taking a certain time as a time window, calculate the enhanced vibration signal in step 3) The vibration amplitude Amp and the vibration frequency fre; 42) Take several time windows as one frame and calculate the average value μ of the vibration amplitude Amp With variance σ Amp , the average value of the vibration frequency μ fre and variance σ fre ; 43) For each time window, compare it with the previous frame. When Amp>μ Amp +3σ Amp or fre>μ fre +3σ fre When the error occurs, it is considered abnormal and a notification is issued.

2. The vibration detection method based on millimeter wave radar according to claim 1, characterized in that: The intermediate frequency signal s(t) obtained in step 1) is a set of signals received by n antennas of the millimeter wave radar, that is, s(t)=[s1(t),s2(t),s3(t),...s n (t)], for a time t, the arrival angle θ is calculated by the phase difference of n antennas, and combined with the distance d, an object M is determined on the two-dimensional plane i =[d i ,θ i ].

3. The vibration detection method based on millimeter wave radar according to claim 2, characterized in that: The step 1) specifically includes: after calculating the distance d and the angle θ, a set of several objects M = [M1, M1, M1, ..., M n ], for the objects in the set, obtain their original signal sequence, perform Doppler Fourier transform with a time window of 3 to 5 seconds, obtain the range-Doppler velocity spectrum, and filter out the stationary objects and select the vibration measurement target objects in the following way: The absolute value of the Doppler velocity in several consecutive time windows is greater than the threshold, and the variance does not exceed half of the average value of the Doppler velocity in 10 windows; In several consecutive time windows, the Doppler velocity is near the two peaks of v and -v; For objects that meet the above conditions, we take them as vibration sensing targets and get the set of vibration measurement target objects: For each object in the set, an original signal sequence S′ is obtained.

4. The vibration detection method based on millimeter wave radar according to claim 3 is characterized in that: The step 2) specifically includes: processing the original signal sequence S′ of the object obtained in step 1) on the IQ plane to restore the original vibration signal, as follows: 21) Through S center = S′-Mean(S′) completes the signal centering and the stationary point P static The position moves to the origin - O; where S center is the original signal sequence after centering, Mean(S′) is the mean of the original signal sequence; 22) Initialize two sampling point sets C far ,C near , set C far represents the set of sampling points away from the radar direction, C near Represents a set of sampling points close to the radar direction; the original signal sequence S after centering center Take the sampling points with positive Doppler velocity and far from the origin and put them into set C far For the sequence S center Traverse the point P in Make When , put point P into set C far Repeat the above process until there are no points to add to the set C. far Put the remaining points into set C near In this way, we can obtain the set C of sampling points on both sides of the stationary point far ,C near , each set is represented as an arc on the IQ plane; 23) The two arcs are fitted by the least squares method, and the two corresponding center points O are obtained by fitting. far , O near ; Restore the vibration and restore the original vibration signal V through the following formula: Where V(t) is the original vibration information of the object at time t, λ is the central wavelength of the millimeter-wave radar, and ∠O near OP(t),∠O far OP(t) is the angle between the sampling point and the stationary point in the two sets at time t relative to their respective centers.

5. The vibration detection method based on millimeter wave radar according to claim 4, characterized in that: The step 3) specifically includes: for the original vibration signal V obtained in step 2), using a model based on deep learning to enhance it, the model uses a multi-layer convolution-deconvolution network to extract information and sets a jump connection layer at the corresponding position, wherein the relationship between the layers is expressed as: in, is the output of the previous layer and the input of the current layer, is the parameter of the current layer network, is the bias of the current layer, f is the selected activation function, is the output of the convolutional layer, x i-relu is the output of the relu layer; the model selects cross entropy as the loss function; after the network training is completed, the time-frequency graph of the original vibration signal V to be enhanced is input into the network to obtain the enhanced time-frequency graph, and then restored to the enhanced vibration signal through inverse short-time Fourier transform.

6. The vibration detection method based on millimeter wave radar according to claim 1, characterized in that: The step 41) specifically includes: for the enhanced vibration signal obtained in step 3) The vibration amplitude Amp is calculated by the following formula: By enhancing the vibration signal After Fourier transform, the peak frequency with the highest intensity on the spectrum is the object's vibration frequency fre.

7. A vibration detection terminal, characterized in that: include: one or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors are enabled to implement the method according to any one of claims 1 to 6.

Images