Millimeter wave radar physiological information echo signal detection method and system

Abstract

The invention relates to the technical field of millimeter-wave radar signal processing, and provides a millimeter-wave radar physiological information echo signal detection method and system, which are based on a radar receiver adopting a complex baseband receiver structure, and the method comprises the following steps: receiving an echo signal by a receiving antenna, dividing the echo signal into two paths, and mixing to obtain an intermediate frequency signal; a circle fitting technology based on a nonlinear least square method is combined with a Levenberg-Marquardt algorithm to calibrate the direct current bias of the intermediate frequency signal; extracting phase signals in a target distance unit by using an improved differential cross multiplication algorithm to form a physiological signal sequence of the target; separating the physiological signal by using a mode decomposition method based on symplectic geometry to obtain a respiration signal and a heartbeat signal; mUSIC spectrum estimation is carried out on the respiration signals and the heartbeat signals, and respiration and heartbeat frequencies are obtained. By means of the method, the anti-interference performance and the detection precision of millimeter wave radar physiological information echo signal detection are improved in the face of various complex environmental factors.

Description

Technical Field

[0001] The present invention relates to the technical field of millimeter-wave radar signal processing, and specifically to a method and system for detecting physiological information echo signals of a millimeter-wave radar. Background Art

[0002] With the continuous progress of society and the increasing attention of people to personal health, the monitoring of human vital signs has become an important issue. Respiration signals and heartbeat signals are important vital sign signals of the human body, which are directly related to multiple health indicators. In the medical field, detecting changes in the human heart rate can provide very valuable reference information for detecting diseases such as arterial obstruction, arrhythmia, and some neurological diseases. Therefore, fast and accurate measurement technologies have received extensive attention. At present, vital sign monitoring can be divided into two categories: contact type and non-contact type. Contact-type vital sign monitoring generally uses methods such as electrocardiogram, respiratory belt, and smart watch for measurement. These methods can obtain relatively accurate vital sign data, but there are some limitations. For example, for patients with large-area skin damage, limb trauma, or skin allergies, it is difficult for contact sensors to directly adhere to the skin, making measurement difficult. In addition, for special groups that require long-term monitoring, such as the elderly and infants, long-term wearing will cause discomfort to the human body, and at the same time will limit the human body's behavior activities and interfere with their normal sleep and daily life. In this context, non-contact vital sign monitoring technology is particularly important.

[0003] Compared with other methods, the non-contact measurement technology based on millimeter-wave radar has more advantages. On the one hand, radar signals have penetrability and can stably detect the human body through clothes without direct contact with the skin, reducing the discomfort of users and being unaffected by some common environmental factors such as light, temperature, humidity, etc.; on the other hand, in recent years, millimeter-wave technology has developed rapidly, hardware chip technology has also improved, and with the implementation of various signal processing technologies and the rapid development of deep learning algorithms, etc., the method based on millimeter-wave radar has received extensive attention in the monitoring of human vital signs and provides solutions for people who cannot use contact-type monitoring methods.

[0004] In the prior art, due to the interference of various factors during the normal operation of the radar system, including the minute jitter during human body measurement, the strong clutter interference reflected by stationary obstacles in the environment, and the possible deviation of the hardware circuit itself, the physiological information echo signal of the millimeter-wave radar has great randomness, resulting in a large error in the detection of the physiological information echo signal. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to improve the anti-interference performance and detection accuracy of millimeter-wave radar physiological information echo signal detection in the face of various complex environmental factors.

[0006] The present invention solves the above technical problems through the following technical means:

[0007] The present invention provides a method for detecting millimeter-wave radar physiological information echo signals, based on a radar receiver adopting a complex baseband receiver structure, including the following steps:

[0008] S1. The echo signal reflected by the target from the radar transmitted signal is received by the receiving antenna and divided into two paths for mixing to obtain an intermediate-frequency signal composed of an in-phase component of the I path and a quadrature component of the Q path;

[0009] S2. Calibrate the DC offset of the intermediate-frequency signal by using the circle fitting technology based on the non-linear least squares method in combination with the Levenberg–Marquardt algorithm;

[0010] S3. Perform fast Fourier transform on consecutive multiple frames of intermediate-frequency signals respectively to obtain the distance units of the target, use the improved differential cross-multiplication algorithm MDACM for phase demodulation, extract the phase signals within the distance units of multiple frames of the target, and form a physiological signal sequence of the target;

[0011] S4. Use the symplectic geometry-based modal decomposition method SGMD to separate the physiological signal into a respiration signal and a heartbeat signal;

[0012] S5. Perform MUSIC spectrum estimation on the respiration signal and the heartbeat signal to obtain the respiration and heartbeat frequencies.

[0013] Further, the intermediate-frequency signal composed of the in-phase component of the I path and the quadrature component of the Q path described in step S1 is specifically:

[0014] The millimeter-wave radar transmits a frequency-modulated continuous wave signal, and the echo signal reflected by the target object is received by the receiving antenna. The signal is mixed with the local oscillator signal and the signal obtained by shifting the local oscillator signal by 90°, respectively, to form two signals of the I / Q paths. Among them, the I path transmits the in-phase component of the intermediate-frequency signal, and the Q path transmits the quadrature component of the intermediate-frequency signal. The intermediate-frequency signal is a complex signal composed of the in-phase component of the I path and the quadrature component of the Q path.

[0015] The intermediate-frequency I / Q signals of the target are respectively expressed as:

[0016]

[0017] Among them, x(t) is the chest displacement caused by the respiration and heartbeat of the human target, λ is the signal wavelength, A I and A Q represent the amplitudes of the I / Q signals, θ vis the constant phase shift caused by signal propagation, DC I and DC Q are the DC offsets of the I / Q signals respectively.

[0018] Furthermore, the specific steps of step S2 are as follows:

[0019] Utilize the orthogonality of the two paths of the complex signal I / Q. Take the I path as the abscissa and the Q path as the ordinate to construct a complex plane. By using the non-linear least squares method to perform circular fitting on the sampling point set for the I / Q two-path signals on the orthogonal complex plane, and combining with the Levenberg–Marquardt algorithm, accurately estimate the coordinates of the center of the circle (DC I , DC Q ) after DC offset. The formula is as follows:

[0020]

[0021] where, (DC I , DC Q ) are the coordinates of the center of the circle, r is the radius, n I / Q data sample points are sampled, and (x i , y i ) are the coordinates of the i-th sample point on the complex plane;

[0022] By subtracting the estimated center position from each sample point, make DC I = 0, DC Q = 0, to complete the DC offset calibration. The formula of the I / Q signal after DC offset calibration is as follows:

[0023]

[0024] Furthermore, the use of the improved differential cross-multiplication algorithm MDACM for phase demodulation to extract the phase signal in the target range cell in step S3 is specifically as follows:

[0025] The phase of the range cell where the target is located is expressed as follows:

[0026]

[0027] Use the improved differential cross-multiplication algorithm MDACM for phase demodulation to extract the phase The formula is as follows:

[0028]

[0029] According to the linear relationship between the phase and the displacement x(m) of the micro-vibration of the human target's chest cavity, x(m) can be expressed as:

[0030]

[0031] Among them, x(m) is the discretized form of the chest displacement x(t) caused by the breathing and heartbeat of the human body target, m is the number of sampling points after discretization, k ∈ m, and λ is the signal wavelength.

[0032] Further, the step S4 includes the following steps:

[0033] S41. Calculate the power spectral density of the input signal x(m), obtain the maximum frequency f in the power spectral density, then calculate the ratio of the maximum frequency f to the signal sampling rate F, compare the ratio with the set threshold, and set the embedding dimension d to m / 3 or 1.2×(F / f) according to the judgment result; finally, construct the trajectory matrix X of x(m) according to the embedding dimension d and the time delay τ; max , and then calculate the maximum frequency f max and the signal sampling rate F s of the ratio, compare the ratio with the set threshold, and set the embedding dimension d to m / 3 or 1.2×(F s / f max ); finally, construct the trajectory matrix X of x(m) according to the embedding dimension d and the time delay τ;

[0034] S42. Obtain the covariance symmetric matrix A = XX through the autocorrelation operation of the trajectory matrix X, and use the matrix A to construct the Hamiltonian matrix M; define W = M, and construct the equation containing the symplectic orthogonal matrix G: T XX, and use the matrix A to construct the Hamiltonian matrix M; define W = M 2 , and construct the equation containing the symplectic orthogonal matrix G:

[0035]

[0036] Among them, the matrix B is an upper triangular matrix, and the eigenvalues of B are λ1, λ2,..., λ d , and according to the properties of the Hamiltonian matrix, it can be known that the eigenvalues of the A matrix are and the corresponding eigenvectors are Q j ;

[0037] S43. Multiply the transpose Q of the eigenvector Q j and the trajectory matrix X to obtain a component S of the coefficient matrix S, and the formula is as follows j T S j = Q

[0038] S j = Q j T X

[0039] The reconstruction matrix Z can be obtained through the coefficient matrix S, and the formula is as follows

[0040] z j = Q j S j

[0041] Among them, z jrepresents the initial single-component reconstruction matrix. The reconstruction matrix Z is composed of d initial single-component reconstruction matrices, that is, Z = z1+z2+...+z d ;

[0042] S44, using the diagonal average method to rearrange the initial single component reconstruction matrix z j , convert the single-component matrix into a one-dimensional signal of length m, and obtain d one-dimensional modal components;

[0043] S45, selecting components belonging to the same mode for superposition and merging to form a symplectic geometric component SGC; calculating the normalized mean square error NMSE of the remaining components and the input signal; when the mean square error is greater than a preset threshold, continuing the iteration, otherwise the decomposition ends;

[0044] S46. Classify the breathing signal and the heartbeat signal from the decomposed symplectic geometric modal components according to the different frequency characteristics of human breathing and heartbeat.

[0045] Furthermore, the step S45 is specifically as follows:

[0046] Select the components with high similarity and add them together to form the first symplectic geometric component SGC1. Eliminate the components involved in the reconstruction of SGC1 and sum the remaining signals, which is recorded as g 1 , calculate the mean square error between the residual component and the input signal, the formula is:

[0047]

[0048] Among them, g h Represents the sum of all components in the residual matrix, h is the number of iterations; set the termination threshold, when the mean square error is greater than the threshold, repeat the above calculation for the remaining signal to generate new symplectic geometric components until the mean square error is less than the threshold; obtain the final decomposition result, the formula is as follows:

[0049]

[0050] where D is the number of symplectic geometry components obtained.

[0051] Furthermore, the spatial spectrum function expression of the MUSIC spectrum in step S5 is:

[0052]

[0053] Among them, a(ω) is the frequency vector of the signal, ω is the normalized circular frequency, the range is [-π,π], G is the matrix composed of the eigenvectors of the noise subspace, a H (ω) is the conjugate transpose of a(ω), G H is the conjugate transpose of G.

[0054] The present invention also provides a millimeter-wave radar physiological information echo signal detection system. The system is based on a radar receiver adopting a complex baseband receiver structure and operates using the above method, including the following modules:

[0055] A signal receiving module, which is used to mix the echo signal reflected by the target from the radar transmitted signal received by the receiving antenna into two paths to obtain an intermediate frequency signal composed of an in-phase component of the I path and a quadrature component of the Q path;

[0056] A calibration module, which is used to calibrate the DC offset of the intermediate frequency signal by using the circle fitting technology based on the nonlinear least squares method in combination with the Levenberg–Marquardt algorithm;

[0057] A physiological signal output module, which is used to perform fast Fourier transform on consecutive multiple frames of intermediate frequency signals respectively to obtain the distance units of the target, use the improved differential cross multiplication algorithm MDACM for phase demodulation, extract the phase signals within the distance units of multiple frames of the target, and form a physiological signal sequence of the target;

[0058] A physiological signal separation module, which is used to separate the physiological signal by using the symplectic geometry-based modal decomposition method SGMD to obtain a respiration signal and a heartbeat signal;

[0059] A data output module, which is used to perform MUSIC spectral estimation on the respiration signal and the heartbeat signal to obtain the respiration and heartbeat frequencies.

[0060] Furthermore, the intermediate frequency signal composed of the in-phase component of the I path and the quadrature component of the Q path in the signal receiving module is specifically:

[0061] The millimeter-wave radar transmits a frequency-modulated continuous wave signal. The echo signal reflected by the target object is received by the receiving antenna. The signal is mixed with the local oscillator signal and the signal obtained by shifting the local oscillator signal by 90° respectively to form two signals of the I / Q paths. Among them, the I path transmits the in-phase component of the intermediate frequency signal, and the Q path transmits the quadrature component of the intermediate frequency signal. The intermediate frequency signal is a complex signal composed of the in-phase component of the I path and the quadrature component of the Q path;

[0062] The intermediate frequency I / Q signals of the target are respectively expressed as:

[0063]

[0064] Among them, x(t) is the chest displacement caused by the respiration and heartbeat of the human target, λ is the signal wavelength, A I and A Q represent the amplitudes of the I / Q signals, θ v is the constant phase shift caused by signal propagation, DC I and DC Q are respectively the DC offsets of the I / Q signals;

[0065] Furthermore, the specific implementation method of the calibration module is as follows:

[0066] Utilize the orthogonality of the I / Q channels of the complex signal. Construct a complex plane with the I channel as the abscissa and the Q channel as the ordinate. By performing circle fitting on the sampling point set for the I / Q signals on the orthogonal complex plane using the non-linear least squares method, and combining with the Levenberg–Marquardt algorithm, accurately estimate the coordinates of the center of the circle (DC I , DC Q ) after DC offset. The formula is as follows:

[0067]

[0068] where, (DC I , DC Q ) are the coordinates of the center of the circle, r is the radius, n I / Q data sample points are sampled, and (x i , y i ) are the coordinates of the i-th sample point on the complex plane;

[0069] By subtracting the estimated center position from each sample point, make DC I = 0, DC Q = 0, and complete the DC offset calibration. The formula for the I / Q signal after DC offset calibration is as follows:

[0070]

[0071] Furthermore, the physiological signal output module uses the improved differential cross-multiplication algorithm MDACM for phase demodulation to extract the phase signals within multiple frames of target range cells specifically as follows:

[0072] The phase of the range cell where the target is located is expressed as follows:

[0073]

[0074] Use the improved differential cross-multiplication algorithm MDACM for phase demodulation to extract the phase The formula is as follows:

[0075]

[0076] According to the linear relationship between the phase and the displacement x(m) of the chest micro-motion of the human target, x(m) can be expressed as:

[0077]

[0078] Among them, x(m) is the discretized form of the chest displacement x(t) caused by the breathing and heartbeat of the human body target, m is the number of sampling points after discretization, k ∈ m, λ is the signal wavelength;

[0079] Further, the physiological signal separation module includes the following units:

[0080] The trajectory matrix unit is used to calculate the power spectral density of the input signal x(m), obtain the maximum frequency f in the power spectral density max , and then calculate the ratio of the maximum frequency f max to the signal sampling rate F s . Compare the ratio with a set threshold, and set the embedding dimension d to m / 3 or 1.2×(F s / f max ) according to the judgment result; finally, construct the trajectory matrix X of x(m) according to the embedding dimension d and the time delay τ;

[0081] The eigenvector unit is used to obtain the covariance symmetric matrix A = X T X through the autocorrelation operation of the trajectory matrix X, and construct the Hamilton matrix M using the matrix A; define W = M 2 , and construct an equation containing the symplectic orthogonal matrix G:

[0082]

[0083] Among them, the matrix B is an upper triangular matrix, and the eigenvalues of B are λ1, λ2,..., λ d . According to the properties of the Hamilton matrix, it can be known that the eigenvalues of the A matrix are , and the corresponding eigenvectors are Q j .

[0084] The reconstruction matrix unit is used to multiply the transpose Q j of the eigenvector Q j Y by the trajectory matrix X to obtain a component S j of the coefficient matrix S:

[0085] S j = Q j T X

[0086] The reconstruction matrix Z can be obtained through the coefficient matrix S:

[0087] z j = Q j S j

[0088] Among them, z jDenote the initial single-component reconstruction matrix. The reconstruction matrix Z is composed of d initial single-component reconstruction matrices, i.e., Z = z1 + z2 +... + z d .

[0089] Modal component unit, which is used to rearrange the initial single-component reconstruction matrix z by the method of diagonal averaging, j convert the single-component matrix into a one-dimensional signal with a length of m, and obtain d one-dimensional modal components;

[0090] Symplectic geometry component unit, which is used to select and superimpose the components belonging to the same mode to form a symplectic geometry component SGC; calculate the normalized mean square error NMSE between the sum of the remaining components and the input signal; when the mean square error is greater than the preset threshold, continue the iteration, otherwise the decomposition ends;

[0091] Output unit, which is used to classify the respiration signal and the heartbeat signal from the decomposed modal components according to the different frequency characteristics of human respiration and heartbeat.

[0092] The advantages of the present invention are as follows:

[0093] (1) The radar receiver of the present invention adopts a complex baseband receiver structure, divides the received echo signal into two paths for quadrature mixing, improves the anti-interference ability, and avoids the problem of detection blind spots generated by a single-channel receiver.

[0094] (2) The present invention adopts a circle fitting technology based on the non-linear least squares method combined with the Levenberg–Marquardt algorithm to accurately estimate the coordinates of the center of the circle, and realizes the DC bias calibration of the I / Q signal. In the case where the radar system is faced with various interference factors, including but not limited to the slight jitter during human body measurement, the strong clutter interference reflected by stationary obstacles in the environment, and the possible deviation of the hardware circuit itself, etc., the accuracy of phase signal extraction is improved.

[0095] (3) The phase demodulation algorithm of the present invention uses the MDACM algorithm, which can directly obtain the phase, avoids the phase jump problem of the commonly used arctangent demodulation algorithm, and omits the phase unwrapping step, thus simplifying the complexity of hardware implementation.

[0096] (4) In the signal separation stage, the present invention selects the symplectic geometry-based modal decomposition SGMD algorithm in the signal decomposition and reconstruction method, which can suppress modal aliasing, and the reconstructed signal components after decomposition have good anti-noise performance.

[0097] (5) After separating the respiration and heartbeat signals, MUSIC spectral estimation is used to estimate the heartbeat frequency and respiration frequency, which improves the accuracy of frequency estimation in a low signal-to-noise ratio environment. Description of the Drawings

[0098] Figure 1 Schematic flow chart of a method for detecting physiological information echo signals of a millimeter-wave radar according to an embodiment of the present invention;

[0099] Figure 2 Schematic diagram of the structure of a complex baseband receiver according to an embodiment of the present invention;

[0100] Figure 3 Respiratory signal waveform diagram of separating physiological signals by the SGMD algorithm according to an embodiment of the present invention;

[0101] Figure 4 Heartbeat signal waveform diagram of separating physiological signals by the SGMD algorithm according to an embodiment of the present invention;

[0102] Figure 5 Respiratory frequency spectrum diagram of MUSIC spectrum estimation according to an embodiment of the present invention;

[0103] Figure 6 Heartbeat frequency spectrum diagram of MUSIC spectrum estimation according to an embodiment of the present invention. Detailed implementation manners

[0104] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0105] Embodiment 1

[0106] The embodiment of the present invention provides a method for detecting physiological information echo signals of a millimeter-wave radar, using a radar receiver with a complex baseband receiver structure, as Figure 1 shown, including the following steps:

[0107] S1. The echo signal reflected by the target from the radar transmitted signal is received by the receiving antenna and divided into two paths for mixing to obtain an intermediate-frequency signal composed of the in-phase component of the I path and the quadrature component of the Q path.

[0108] The radar receiver selects a complex baseband receiver structure, and its structure is as Figure 2 shown. The complex baseband receiver can avoid the demodulation zero-point problem and restore the target motion information with high precision. The millimeter-wave radar transmits a frequency-modulated continuous-wave signal, and the echo signal reflected by the target object is received by the receiving antenna. The signal is mixed with the local oscillator signal and the signal obtained by shifting the local oscillator signal by 90° respectively to form two signals of the I / Q paths. Among them, the I path transmits the in-phase component of the intermediate-frequency signal, and the Q path transmits the quadrature component of the intermediate-frequency signal. The intermediate-frequency signal is a complex signal composed of the in-phase component of the I path and the quadrature component of the Q path.

[0109] The intermediate-frequency I / Q signals of the target are respectively expressed as:

[0110]

[0111] where x(t) is the chest displacement caused by the breathing and heartbeat of the human target, λ is the signal wavelength, A I and A Q represent the amplitudes of the I / Q signals, θ v is the constant phase shift caused by signal propagation, DC I and DC Q are the DC offsets of the I / Q signals respectively.

[0112] S2. Implement DC bias calibration for the I / Q signals by using the circle fitting technique based on the nonlinear least squares method in combination with the Levenberg–Marquardt algorithm. The specific method is as follows:

[0113] Utilize the orthogonality of the I and Q channels of the complex signal to construct a complex plane with the I channel as the abscissa and the Q channel as the ordinate, and process the I / Q signals on the orthogonal complex plane. If the values of DC I and DC Q are 0, then the sampling points of the I / Q signals can be approximately represented as a circle centered at the origin. If the values of DC I and DC Q are not 0, it causes the received complex signal to deviate from the origin, and the origin is shifted from (0, 0) to (DC I , DC Q ). In this embodiment, the circle fitting of the sampling point set is performed by using the nonlinear least squares method, and the Levenberg–Marquardt algorithm can accurately estimate the coordinates of the center of the circle (DC I , DC Q ) after DC bias, and the formula is as follows:

[0114]

[0115] where (DC I , DC Q ) are the coordinates of the center of the circle, r is the radius, n I / Q data sample points are sampled, and (x i , y i ) are the coordinates of the i-th sample point on the complex plane;

[0116] By subtracting the estimated center position from each sample point, make DC I = 0, DC Q = 0, and perform DC bias calibration to avoid the influence of nonlinear distortion. The formula of the I / Q signal after DC bias calibration is as follows:

[0117]

[0118] S3. Perform fast Fourier transform on consecutive multiple frames of intermediate frequency signals to obtain the range cells of the target, use the improved differential cross multiplication algorithm MDACM for phase demodulation, extract the phase signals within the range cells of multiple frames of the target, and form the physiological signal sequence of the target.

[0119] In an embodiment of the present invention, perform fast Fourier transform on consecutive multiple frames of intermediate frequency signals to obtain the spectrum of the intermediate frequency signals. The position of its peak reflects the target position. Find the range cell corresponding to the human target, perform phase extraction, and form the physiological signal sequence of the target. The phase of the range cell where the target is located is expressed as follows:

[0120]

[0121] Use MDACM to perform phase demodulation to extract the phase The formula is as follows:

[0122]

[0123] According to the linear relationship between the phase and the displacement x(m) of the chest micro-motion of the human target, x(m) can be expressed as:

[0124]

[0125] Among them, x(m) is the discretized form of the chest displacement x(t) caused by the breathing and heartbeat of the human target, m is the number of sampling points after discretization, k ∈ m, and λ is the signal wavelength.

[0126] S4. Use the symplectic geometry-based modal decomposition method (SGMD) to separate the physiological signal into the respiration signal and the heartbeat signal. The specific decomposition steps are as follows:

[0127] S41. Calculate the power spectral density of the input signal x(m) to obtain the maximum frequency f max , and then calculate the maximum frequency f max and the ratio of the signal sampling rate F s . Compare the ratio with the set threshold, and set the embedding dimension d to m / 3 or 1.2×(F s / f max ) according to the judgment result. Construct the trajectory matrix X of x(m) according to the embedding dimension d and the time delay τ.

[0128] The trajectory matrix X can directly affect the decomposition effect by changing the embedding dimension d and the time delay τ. The time delay τ is set to 1, and the selection of the embedding dimension d depends on the power spectrum of the sampled signal, the maximum frequency f max and the signal sampling rate F sThe ratio is less than the threshold value of 10 -3 When, d = m / 3 can be obtained, otherwise d = 1.2×(F s / f max ).

[0129] S42. The covariance symmetric matrix A = X T X can be obtained through the autocorrelation operation of the trajectory matrix X. Using A, the Hamilton matrix M is constructed. From the definition of the Hamilton matrix, it is known that W = M 2 is also a Hamilton matrix. An equation containing the symplectic orthogonal matrix G is constructed:

[0130]

[0131] The matrix B is an upper triangular matrix, and the eigenvalues of B are λ1, λ2,..., λ d . According to the properties of the Hamilton matrix, the eigenvalues of the A matrix are The corresponding eigenvector is Q j .

[0132] S43. Multiply the transpose Q j of the eigenvector Q j T by the trajectory matrix X to obtain a component S j

[0133] S j = Q j T X

[0134] Furthermore, the reconstruction matrix Z can be obtained through the coefficient matrix S

[0135] z j = Q j S j

[0136] where, z j represents the initial single-component reconstruction matrix, and the reconstruction matrix Z is composed of d initial single-component reconstruction matrices, that is, Z = z1 + z2 +... + z d .

[0137] S44. Rearrange the initial single-component reconstruction matrix z j by using the method of diagonal averaging, and the single-component matrix can be converted into a one-dimensional signal with a length of m. There are d original reconstruction matrices Z j . After diagonal averaging, d one-dimensional modal components can be obtained.

[0138] S45. After diagonal averaging, there may be the same periodic and modal components. Select the components belonging to the same mode, superimpose and merge them to form a symplectic geometric component (SGC). Calculate the normalized mean square error (NMSE) between the sum of the remaining components and the input signal. When the mean square error is greater than the preset threshold, continue the iteration; otherwise, the decomposition ends. The specific method is as follows:

[0139] The evaluation index for screening components of the same mode is period similarity. Compare the period similarity of the first initial single component with the remaining initial single components, select the components with high similarity and add them together to form the first symplectic geometric component SGC1. Eliminate the components participating in the reconstruction of SGC1, sum the remaining signals, and denote it as g 1 , calculate the mean square error between the sum of the remaining components and the input signal. The formula is as follows:

[0140]

[0141] where g h represents the sum of all components in the remaining matrix, h is the number of iterations. Set an appropriate NMSE termination threshold. When the mean square error is greater than the threshold, it is determined that the decomposition is not completed. Continue the above calculation for the remaining signal to generate a new symplectic geometric component until the decomposition is completed to obtain the final decomposition result. The formula is as follows:

[0142]

[0143] where D is the number of symplectic geometric components obtained.

[0144] S46. Classify the respiratory signal and the heartbeat signal from the symplectic geometric modal components obtained by decomposition according to the different frequency characteristics of human respiration and heartbeat; the waveform diagrams of the respiratory signal and the heartbeat signal are as Figure 3 and Figure 4 shown.

[0145] S5. Perform MUSIC spectral estimation on the respiratory signal and the heartbeat signal to obtain the respiratory and heartbeat frequencies.

[0146] In an embodiment of the present invention, the effective component subspace is the physiological signal subspace, and the physiological signal X(m) can be the respiratory signal or the heartbeat signal separated by the symplectic geometric-based modal decomposition method. The specific process is as follows:

[0147] (1) Taking the physiological signal X(m) as the observation sample, calculate the covariance matrix T of the sample data. Assume that the order of the covariance matrix is M = 32. The higher the order of the covariance matrix, the more accurate the result, but the computational amount will also increase exponentially. The observation sample sequence is:

[0148] X(m) = [X(m), X(m - 1),..., X(m - M + 1)]

[0149] (2) Perform eigenvalue decomposition on the covariance matrix T to obtain the basis vectors of the noise subspace. In this embodiment, the effective components of the signal are mainly the respiration or heartbeat signals of vital signs, so the number of effective signals is 1. The eigenvalues related to noise are smaller than those of the effective components. After eigenvalue decomposition, the eigenvectors corresponding to the M - 1 smallest eigenvalues are a set of basis vectors of the noise subspace.

[0150] (3) The expression of the spatial spectrum function of the MUSIC algorithm is:

[0151]

[0152] where \(a(\omega)\) is the frequency vector of the signal, \(\omega\) is the normalized circular frequency, with a range of \([-\pi,\pi]\), \(G\) is the matrix composed of the eigenvectors of the noise subspace, \(a H (\omega)\) is the conjugate transpose of \(a(\omega)\), and \(G H is the conjugate transpose of \(G\).

[0153] (4) Conduct spectral peak search to determine the frequencies of respiration and heartbeat.

[0154] This embodiment also provides a verification process using the above - mentioned test method, specifically:

[0155] The test target sits in front of the radar. A three - lead electrocardiograph is used to monitor the heart rate and respiration. During the monitoring process, the body shakes slightly to create detection obstacles. The three - lead electrocardiograph monitors the respiration at 16 times per minute and the heart rate at 90 times per minute. The millimeter - wave radar monitoring data is as Figure 5 and Figure 6 shown. The target respiration frequency is 0.28 Hz, approximately 17 times per minute, and the heartbeat frequency is 1.54 Hz, approximately 92 times per minute. The comparison error between the two is small, proving that the accurate extraction of the physiological information echo signal through a series of algorithms improves the accuracy of the non - contact vital sign monitoring results.

[0156] Embodiment 2

[0157] It should be further noted that, based on the same inventive concept, the present invention also provides a millimeter - wave radar physiological information echo signal detection system. The system is based on a radar receiver using a complex baseband receiver structure and operates using the method described in Embodiment 1, including the following modules:

[0158] A signal receiving module, which is used to mix the echo signal reflected by the target from the radar transmitted signal received by the receiving antenna into two paths to obtain an intermediate - frequency signal composed of the in - phase component of the I channel and the quadrature component of the Q channel;

[0159] A calibration module, which is used to calibrate the DC offset of the intermediate frequency signal by using the circle fitting technology based on the non-linear least squares method in combination with the Levenberg–Marquardt algorithm;

[0160] A physiological signal output module, which is used to perform fast Fourier transform on consecutive multiple frames of intermediate frequency signals respectively to obtain the range cells of the target, use the improved differential cross multiplication algorithm MDACM for phase demodulation, extract the phase signals within the range cells of multiple frames of the target, and form a physiological signal sequence of the target;

[0161] A physiological signal separation module, which is used to separate the physiological signal by using the symplectic geometry based modal decomposition method SGMD to obtain the respiration signal and the heartbeat signal;

[0162] A data output module, which is used to perform MUSIC spectral estimation on the respiration signal and the heartbeat signal to obtain the respiration and heartbeat frequencies.

[0163] The intermediate frequency signal composed of the in-phase component of the I channel and the quadrature component of the Q channel described in the signal receiving module is specifically:

[0164] The millimeter wave radar transmits a frequency modulated continuous wave signal, and the echo signal reflected by the target object is received by the receiving antenna. The signal is mixed with the local oscillator signal and the signal with a 90° phase shift of the local oscillator signal respectively to form two I / Q signals. Among them, the I channel transmits the in-phase component of the intermediate frequency signal, and the Q channel transmits the quadrature component of the intermediate frequency signal. The intermediate frequency signal is a complex signal composed of the in-phase component of the I channel and the quadrature component of the Q channel;

[0165] The intermediate frequency I / Q signals of the target are respectively expressed as:

[0166]

[0167] Among them, x(t) is the chest displacement caused by the respiration and heartbeat of the human target, λ is the signal wavelength, A I and A Q represent the amplitudes of the I / Q signals, θ v is the constant phase shift caused by signal propagation, DC I and DC Q are the DC offset amounts of the I / Q signals respectively;

[0168] The specific execution manner of the calibration module is:

[0169] Utilize the orthogonality of the two I / Q channels of the complex signal. With the I channel as the abscissa and the Q channel as the ordinate, construct a complex plane. By performing circle fitting on the sampling point set for the I / Q signals on the orthogonal complex plane by using the non-linear least squares method, and combining with the Levenberg–Marquardt algorithm, accurately estimate the center of the circle after DC offset (DC I , DCQ ) coordinates, the formula is as follows:

[0170]

[0171] Among them, (DC I , DC Q ) is the center coordinate, r is the radius, n I / Q data sample points are sampled, and (x i , y i ) is the coordinate of the i-th sample point on the complex plane;

[0172] By subtracting the estimated center position from each sample point, DC I = 0, DC Q = 0, the DC bias calibration is completed, and the formula of the I / Q signal after DC bias calibration is as follows:

[0173]

[0174] The physiological signal output module uses the improved differential cross multiplication algorithm MDACM for phase demodulation to extract the phase signal in multiple frames of target range cells specifically as:

[0175] The phase of the range cell where the target is located is expressed as follows:

[0176]

[0177] Use the improved differential cross multiplication algorithm MDACM for phase demodulation to extract the phase The formula is as follows:

[0178]

[0179] According to the linear relationship between the phase and the displacement x(m) of the chest micro-motion of the human target, x(m) can be expressed as:

[0180]

[0181] Among them, x(m) is the discretized form of the chest displacement x(t) caused by the breathing and heartbeat of the human target, m is the number of discretized sampling points, k ∈ m, and λ is the signal wavelength;

[0182] The physiological signal separation module includes the following units:

[0183] The trajectory matrix unit is used to calculate the power spectral density of the input signal x(m), obtain the maximum frequency f max , and then calculate the maximum frequency f max and the signal sampling rate F sThe ratio is compared with a set threshold value, and according to the judgment result, the embedding dimension d is set to m / 3 or 1.2×(F s / f max ); Finally, according to the embedding dimension d and the time delay τ, the trajectory matrix X of x(m) is constructed;

[0184] An eigenvector unit, which is used to obtain the covariance symmetric matrix A = X T X through the autocorrelation operation of the trajectory matrix X, and construct the Hamiltonian matrix M by using the matrix A; Define W = M 2 , and construct an equation containing the symplectic orthogonal matrix G:

[0185]

[0186] where the matrix B is an upper triangular matrix, and the eigenvalues λ1, λ2,..., λ d of B. According to the properties of the Hamiltonian matrix, it can be known that the eigenvalues of the A matrix are , and the corresponding eigenvectors are Q j .

[0187] A reconstruction matrix unit, which is used to multiply the transpose Q j of the eigenvector Q j T by the trajectory matrix X to obtain a component S j of the coefficient matrix S:

[0188] S j = Q j T X

[0189] The reconstruction matrix Z can be obtained through the coefficient matrix S:

[0190] z j = Q j S j

[0191] where z j represents the initial single-component reconstruction matrix, and the reconstruction matrix Z is composed of d initial single-component reconstruction matrices, that is, Z = z1 + z2 +... + z d .

[0192] A modal component unit, which is used to rearrange the initial single-component reconstruction matrix z j by using the diagonal averaging method, convert the single-component matrix into a one-dimensional signal with a length of m, and obtain d one-dimensional modal components;

[0193] The symplectic geometric component unit is used to select and superimpose components belonging to the same mode to form a symplectic geometric component SGC; calculate the normalized mean square error NMSE between the sum of the remaining components and the input signal; when the mean square error is greater than the preset threshold, continue the iteration, otherwise the decomposition ends;

[0194] The output unit is used to classify the respiration signal and the heartbeat signal from the decomposed modal components according to the different frequency characteristics of human respiration and heartbeat.

[0195] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting the physiological information echo signal of a millimeter-wave radar, characterized in that Based on a radar receiver adopting a complex baseband receiver structure, it includes the following steps: S1. The echo signal reflected by the target from the radar transmitted signal is received by the receiving antenna and divided into two paths for mixing to obtain an intermediate frequency signal composed of the in-phase component of the I path and the quadrature component of the Q path; S2. Use the circle fitting technology based on the nonlinear least squares method combined with the Levenberg–Marquardt algorithm to calibrate the DC bias of the intermediate frequency signal; S3. Perform fast Fourier transform on consecutive multiple frames of intermediate frequency signals respectively to obtain the target range cells, use the improved differential cross multiplication algorithm MDACM for phase demodulation, extract the phase signals within multiple frames of target range cells, and form the physiological signal sequence of the target; S4. Use the symplectic geometry-based modal decomposition method SGMD to separate the physiological signal into a respiration signal and a heartbeat signal; S5. Perform MUSIC spectral estimation on the respiration signal and the heartbeat signal to obtain the respiration and heartbeat frequencies.

2. The method for detecting the physiological information echo signal of a millimeter-wave radar according to claim 1, wherein The intermediate frequency signal composed of the in-phase component of the I path and the quadrature component of the Q path described in step S1 is specifically: The millimeter-wave radar transmits a frequency-modulated continuous wave signal, and the echo signal reflected by the target object is received by the receiving antenna. The signal is mixed with the local oscillator signal and the signal with a 90° phase shift of the local oscillator signal respectively to form two signals of the I / Q paths. Among them, the I path transmits the in-phase component of the intermediate frequency signal, and the Q path transmits the quadrature component of the intermediate frequency signal. The intermediate frequency signal is a complex signal composed of the in-phase component of the I path and the quadrature component of the Q path; The intermediate frequency I / Q signals of the target are respectively expressed as: Among them, x(t) is the chest displacement caused by the breathing and heartbeat of the human body target, λ is the signal wavelength, A I and A Q represent the amplitudes of the I / Q signals, θ v is the constant phase shift caused by signal propagation, DC I and DC Q are the DC offsets of the I / Q signals respectively.

3. The method for detecting physiological information echo signals of a millimeter-wave radar according to claim 2, characterized in that The specific content of step S2 is: Using the orthogonality of the I / Q two channels of the complex signal, a complex plane is constructed with the I channel as the abscissa and the Q channel as the ordinate. By using the nonlinear least squares method to perform circle fitting on the sampling point set for the I / Q two-channel signals on the orthogonal complex plane, and combining with the Levenberg–Marquardt algorithm, the coordinates of the center of the circle after DC bias (DC I , DC Q ) are accurately estimated. The formula is as follows: Among them, (DC I , DC Q ) is the center coordinate, r is the radius, n I / Q data sample points are sampled, and (x i , y i ) is the coordinate of the i-th sample point on the complex plane; By subtracting the estimated center position from each sample point, DC I = 0, DC Q = 0, to complete the DC bias calibration. The I / Q signal formula after DC bias calibration is as follows:

4. The method for detecting the physiological information echo signal of the millimeter-wave radar according to claim 3, wherein The specific content of using the improved differential cross multiplication algorithm MDACM for phase demodulation and extracting the phase signals within multiple frames of target range cells described in step S3 is: The phase of the range cell where the target is located is expressed as follows: Use the improved differential cross - multiplication algorithm MDACM for phase demodulation to extract the phase The formula is as follows: According to the linear relationship between the phase and the displacement x(m) of the thoracic cavity micro-motion of the human target, x(m) can be expressed as: Among them, x(m) is the discretized form of the thoracic cavity displacement x(t) caused by the respiration and heartbeat of the human target, m is the number of discretized sampling points, k ∈ m, and λ is the signal wavelength.

5. A method for detecting physiological information echo signals of a millimeter-wave radar according to claim 4, characterized in that Step S4 includes the following steps: S41. Calculate the power spectral density of x(m), obtain the maximum frequency f in the power spectral density, max , and then calculate the ratio of the maximum frequency f max to the signal sampling rate F s . Compare the ratio with a set threshold, and set the embedding dimension d to m / 3 or 1.2×(F s / f max ) according to the judgment result; Finally, construct the trajectory matrix X of x(m) based on the embedding dimension d and the delay τ; S42. Obtain the covariance symmetric matrix A = X through the autocorrelation operation of the trajectory matrix X T X, and construct the Hamiltonian matrix M using the matrix A; define W = M 2 , and construct an equation containing the symplectic orthogonal matrix G: Among them, matrix B is an upper triangular matrix, and the eigenvalues of B are λ1, λ2,..., λ d . According to the properties of the Hamiltonian matrix, the eigenvalues of matrix A are , and the corresponding eigenvectors are q j ; S43. Transpose the eigenvector q j to obtain Q j T and multiply it by the trajectory matrix X to obtain a component S j of the coefficient matrix S. The formula is as follows S j = Q j T X The reconstruction matrix Z can be obtained through the coefficient matrix S, and the formula is as follows z j = q j S j Among them, z j represents the initial single-component reconstruction matrix. The reconstruction matrix Z is composed of d initial single-component reconstruction matrices, that is, Z = z1 + z2 +... + z d ; S44. Rearrange the initial single-component reconstruction matrix z by using the method of diagonal averaging j , convert the single-component matrix into a one-dimensional signal with a length of m, and obtain d one-dimensional modal components; S45. Select the components belonging to the same mode and superimpose and combine them to form a symplectic geometry component SGC; calculate the normalized mean square error NMSE between the sum of the remaining components and the input signal; when the mean square error is greater than the preset threshold, continue the iteration, otherwise the decomposition ends; S46. Separate the respiration signal and the heartbeat signal from the symplectic geometry mode components obtained by decomposition according to the different frequency characteristics of human respiration and heartbeat.

6. The method for detecting physiological information echo signals of a millimeter-wave radar according to claim 5, characterized in that, The specific content of step S45 is: Select the components with high similarity and add them together to form the first symplectic geometric component SGC1. Exclude the components involved in the reconstruction of SGC1, sum the remaining signals, and denote it as g 1 , calculate the mean square error between the sum of the remaining components and the input signal. The formula is: where g h represents the sum of all components in the residual matrix, h is the number of iterations; a termination threshold is set. When the mean square error is greater than the threshold, the above calculations are repeated for the residual signal to generate new symplectic geometric components until the mean square error is less than the threshold; the final decomposition result is obtained, and the formula is as follows: Among them, D is the number of symplectic geometry components obtained.

7. A method for detecting millimeter-wave radar physiological information echo signals according to claim 5, characterized in that, The expression of the spatial spectral function of the MUSIC spectrum described in step S5 is: Among them, \(a(\omega)\) is the frequency vector of the signal, \(\omega\) is the normalized circular frequency, with a range of \([-\pi,\pi]\), \(G\) is the matrix composed of the eigenvectors of the noise subspace, \(a H (\omega)\) is the conjugate transpose of \(a(\omega)\), \(G H is the conjugate transpose of \(G\).

8. A millimeter-wave radar physiological information echo signal detection system, characterized in that, The system is based on a radar receiver adopting a complex baseband receiver structure. When running, it adopts the method described in any one of claims 1-7, and includes the following modules: A signal receiving module, which is used for the echo signal reflected by the target from the radar transmitted signal to be received by the receiving antenna and divided into two paths for mixing to obtain an intermediate frequency signal composed of the in-phase component of the I path and the quadrature component of the Q path; A calibration module, which is used to calibrate the DC bias of the intermediate frequency signal by using the circle fitting technology based on the nonlinear least squares method in combination with the Levenberg–Marquardt algorithm; A physiological signal output module, which is used to perform fast Fourier transform on consecutive multiple frames of intermediate frequency signals respectively to obtain the range cells of the target, use the improved differential cross multiplication algorithm MDACM for phase demodulation, extract the phase signals within the range cells of multiple frames of the target, and form a physiological signal sequence of the target; A physiological signal separation module, which is used to separate the physiological signal into a respiration signal and a heartbeat signal by using the symplectic geometry based modal decomposition method SGMD; A data output module, which is used to perform MUSIC spectral estimation on the respiration signal and the heartbeat signal to obtain the respiration and heartbeat frequencies.

9. The millimeter-wave radar physiological information echo signal detection system according to claim 8, characterized in that, The intermediate frequency signal composed of the in-phase component of the I channel and the quadrature component of the Q channel described in the signal receiving module is specifically: The millimeter wave radar emits a frequency modulated continuous wave signal, and the echo signal reflected by the target object is received by the receiving antenna. The signal is mixed with the local oscillator signal and the signal with a 90° phase shift of the local oscillator signal respectively to form two signals of the I / Q channels. Among them, the I channel transmits the in-phase component of the intermediate frequency signal, and the Q channel transmits the quadrature component of the intermediate frequency signal. The intermediate frequency signal is a complex signal composed of the in-phase component of the I channel and the quadrature component of the Q channel; The intermediate frequency I / Q signals of the target are respectively expressed as: where x(t) is the chest displacement caused by the breathing and heartbeat of the human target, λ is the signal wavelength, A I and A Q represent the amplitudes of the I / Q signals, θ v is the constant phase shift caused by signal propagation, DC I and DC Q are the DC offsets of the I / Q signals, respectively; The specific execution mode of the calibration module is: Using the orthogonality of the I / Q two channels of the complex signal, a complex plane is constructed with the I channel as the abscissa and the Q channel as the ordinate. By performing circle fitting on the sampling point set for the I / Q two-channel signals on the orthogonal complex plane using the nonlinear least squares method, combined with the Levenberg–Marquardt algorithm, the coordinates of the center of the circle after DC bias (DC I , DC Q ) are accurately estimated. The formula is as follows: Among them, (DC I , DC Q ) is the center coordinate, r is the radius, and n I / Q data sample points are sampled. (x i , y i ) is the coordinate of the i-th sample point in the complex plane; By subtracting the estimated center position from each sample point, DC I = 0, DC Q = 0, the DC bias calibration is completed. The I / Q signal formula after DC bias calibration is as follows:

10. The millimeter-wave radar physiological information echo signal detection system according to claim 9, wherein, The use of the improved differential cross multiplication algorithm MDACM for phase demodulation and the extraction of the phase signals within the range cells of multiple frames of the target described in the physiological signal output module are specifically: The phase of the range cell where the target is located is expressed as follows: Use the improved differential cross - multiplication algorithm MDACM for phase demodulation to extract the phase The formula is as follows: According to the linear relationship between the phase and the displacement x(m) of the thoracic cavity micro-motion of the human target, x(m) can be expressed as: Where x(m) is the discretized form of the thoracic cavity displacement x(t) caused by the respiration and heartbeat of the human target, m is the number of discretized sampling points, k ∈ m, and λ is the signal wavelength; The physiological signal separation module includes the following units: A trajectory matrix unit for calculating the power spectral density of the input signal x(m), obtaining the maximum frequency f in the power spectral density max , then calculating the maximum frequency f max and the ratio of the signal sampling rate F s . Comparing the ratio with a set threshold, and setting the embedding dimension d to m / 3 or 1.2×(F s / f max ) according to the judgment result; finally, constructing the trajectory matrix X of x(m) according to the embedding dimension d and the delay τ; An eigenvector unit for obtaining a covariance symmetric matrix A = X through autocorrelation operation of a trajectory matrix X T X, constructing a Hamiltonian matrix M using matrix A; defining W = M 2 , constructing an equation containing a symplectic orthogonal matrix G: Among them, matrix B is an upper triangular matrix, and the eigenvalues of B are λ1, λ2,..., λ d , according to the properties of the Hamilton matrix, the eigenvalues of matrix A are The corresponding eigenvector is q j ; A reconstructed matrix unit for multiplying the transpose Q j of the eigenvector q j T by the trajectory matrix X to obtain a component S j of the coefficient matrix S S j = Q j T X The reconstruction matrix Z can be obtained through the coefficient matrix S: Z j = q j s j Among them, z j represents the initial single-component reconstruction matrix, and the reconstruction matrix Z is composed of d initial single-component reconstruction matrices, that is, Z = z1 + Z2 +... + Z d ; Modal component unit, which is used to rearrange the initial single-component reconstruction matrix Z by means of diagonal averaging j , convert the single-component matrix into a one-dimensional signal with a length of m, and obtain d one-dimensional modal components; The symplectic geometry component unit is used to select the components belonging to the same mode, superimpose and combine them to form a symplectic geometry component SGC; calculate the normalized mean square error NMSE between the sum of the remaining components and the input signal; when the mean square error is greater than the preset threshold, continue the iteration, otherwise the decomposition ends; The output unit is used to classify the respiration signal and the heartbeat signal from the modal components obtained by decomposition according to the different frequency characteristics of human respiration and heartbeat.

Images