A total variation-based millimeter wave radar signal motion noise removal method

Abstract

The present application belongs to the field of non-contact vital sign detection, and particularly relates to a millimeter wave radar signal motion noise removing method based on total variation, which specifically comprises the following steps: S1: after chest cavity signals containing motion noise are acquired by using a millimeter wave radar, the signals are rearranged and converted into a gray scale image; S2: after chest cavity images containing noise are acquired, second-order gradient information of the images in different directions is calculated respectively; S3: according to the gradient information in different directions, the sparse characteristics of noise and underlying image structure are obtained, and a second-order regularization constraint term in the transverse direction and the longitudinal direction is respectively designed by introducing an Lp pseudo norm; meanwhile, a global sparse constraint term is designed based on the prior characteristics of noise; and S4: the denoised image is restored into a one-dimensional chest cavity signal. The method fully utilizes the sparse characteristics of motion noise and underlying image structure, and better describes the sparse characteristics by introducing an Lp pseudo norm, so that the motion noise is effectively removed, and the detail characteristics of the chest cavity signal are better preserved.

Description

Technical Field

[0001] This invention belongs to the field of non-contact vital sign detection technology, specifically a method for removing motion noise from millimeter-wave radar signals based on total variation. Background Technology

[0002] With increasing population, patients in hospitals have diverse conditions. For patients with skin diseases or extensive burns, contact-based heart rate monitoring is unsuitable. Therefore, research on non-contact heart rate signal detection is essential. When using millimeter-wave radar to acquire chest cavity signals, the human body is rarely completely still; there will always be involuntary or imperceptible movements, resulting in motion noise in the acquired signals. This motion noise degrades the quality of chest cavity signal data, severely impacting its subsequent applications in medical and rescue fields. Current motion noise removal algorithms can be broadly categorized into two types: filtering-based methods and signal decomposition-based methods, but both often have limitations. Filtering-based methods typically utilize various filters to remove motion noise, such as Chebyshev filters, wavelet decomposition, and wavelet-Chebyshev combined filters. Signal decomposition-based methods primarily extract useful components by finding the set of modal components and their respective center frequencies, such as empirical mode decomposition and variational mode decomposition. However, both of these methods fail to achieve satisfactory results when dealing with non-periodic irregular noise. Therefore, we propose a motion noise removal method for millimeter-wave radar signals based on total variation to solve the problem of motion noise interference. Summary of the Invention

[0003] (a) Technical problems to be solved

[0004] To address the shortcomings of existing technologies, this invention provides a method for removing motion noise from millimeter-wave radar signals based on total variation, thus solving the problems mentioned in the background section.

[0005] (II) Technical Solution

[0006] To achieve the above objectives, the present invention specifically adopts the following technical solution:

[0007] A method for removing motion noise from millimeter-wave radar signals based on total variation includes the following steps;

[0008] S1. After obtaining the millimeter-wave radar acquisition data matrix from the millimeter-wave radar echo signal dataset, static clutter filtering, unwrapping, and differential processing are performed to obtain the phase difference signal. The phase difference signal matrix is ​​then processed by extracting the row vectors of each period according to the period of the thoracic signal. These extracted row vectors are then placed into each row of the thoracic signal grayscale matrix. Considering the periodicity of the signal, the newly constructed thoracic signal grayscale matrix should be relatively smooth when converted into a grayscale image, and should also have certain sparsity characteristics in its structure.

[0009] S2, after acquiring the chest cavity signal image containing motion noise, calculate the second-order gradient information in different directions of the image to obtain its horizontal second-order gradient, vertical second-order gradient, and mixed second-order gradient. The calculation formula for the gradient information in different directions of the image is as follows:

[0010]

[0011] Where: I(x,y) represents the pixel values ​​of the image, G xx (x,y) represents the second-order gradient information in the horizontal direction of the image, G yy (x,y) represents the second-order gradient information in the vertical direction of the image, G xy (x,y) represents the second-order mixed gradient information of the image, and (x,y) represents the row and column index of the image pixels.

[0012] S3, based on the second-order gradient information in different directions, obtains the sparsity characteristics of noise and the underlying image structure, and designs second-order regularization constraint terms in different directions by introducing the Lp pseudo-norm; at the same time, based on the prior characteristics of motion noise, a global sparsity constraint term is designed, the relevant expression of which is:

[0013] R0=||YZ||1

[0014]

[0015]

[0016]

[0017] Where: Y represents an ideal image without noise, Z represents an image contaminated by noise, and L... x =[-1,1] represents the convolution kernel with lateral difference, L y =[-1,1] T R represents the convolution kernel for vertical difference, R0 represents the global sparse constraint term, and R xx R represents the level second-order regularization constraint term. yy R represents the vertical second-order regularization constraint term. xy This represents the mixed second-order regularization constraint term.

[0018] Specifically, in the field of image processing, the Lp norm is generally defined as... Lp pseudonorm is defined as

[0019] S4 combines the second-order regularization constraints in different directions and the global sparsity constraints in S3 to obtain the complete total variational model expression for removing random motion noise:

[0020]

[0021] Where μ represents the coefficient of the second-order sparse regularization term.

[0022] S5, the alternating multiplier method is used to iteratively solve the noise-removed total variational model in S4, including the following steps:

[0023] S501, by introducing intermediate variables, transforms a constrained problem into an unconstrained problem and obtains the augmented Lagrangian function of the problem.

[0024] The augmented Lagrangian expression for the unconstrained problem is:

[0025]

[0026] Where: A0 = YZ, A1 = L x *L x *Y, A2 = L y *L y *Y, A3 = L x *L y *Y, where Λ0, Λ1, Λ2 and Λ3 represent Lagrange multipliers, γ0, γ1, γ2 and γ3 represent the coefficients of the penalty term, max represents taking the maximum value, and min represents taking the minimum value.

[0027] S502 solves the subproblems of each variable in the augmented Lagrangian function one by one, and obtains their iterative formulas.

[0028] (1) From the above augmented Lagrangian function, the subproblem of Y can be obtained as follows:

[0029]

[0030]

[0031]

[0032]

[0033] Using the convolution theorem and the fast Fourier transform, the iterative formula for Y is obtained as follows:

[0034]

[0035]

[0036]

[0037] In the formula: Operators representing Fourier transforms The operator representing the inverse Fourier transform, express Conjugate mappings, where · represents dot product and / represents dot division.

[0038] (2) From the above augmented Lagrangian function, the subproblems of the intermediate variables A0, A1, A2, and A3 can be obtained as follows:

[0039]

[0040]

[0041]

[0042]

[0043] Using the contraction theorem, the iterative formulas for the intermediate variables A0, A1, A2, and A3 are obtained as follows:

[0044]

[0045]

[0046]

[0047]

[0048] In the formula:

[0049] (3) Finally, using the gradient ascent method, the iterative formulas for the Lagrange multipliers Λ0, Λ1, Λ2, and Λ3 are obtained as follows:

[0050]

[0051]

[0052]

[0053]

[0054] Where λ is the learning rate.

[0055] S503, after multiple iterations of each variable until the relevant conditions or the maximum number of iterations are met, the optimal solution of Y can be obtained, that is, the chest cavity signal image after motion noise removal.

[0056] (III) Beneficial Effects

[0057] Compared with existing technologies, this invention provides a method for motion noise removal of millimeter-wave radar signals based on total variation, which has the following beneficial effects:

[0058] This method converts one-dimensional physiological signals acquired using millimeter-wave radar into two-dimensional image signals, and then uses noise and the sparsity of the underlying image structure for denoising. This not only effectively removes noise, but also better preserves the detailed features of the thoracic cavity signal, providing a new approach to denoising physiological signals.

[0059] This invention utilizes the second-order gradient information of the chest cavity signal image when solving the proposed total variational model for removing motion noise. The second-order gradient information of the image can better highlight noise and possesses a certain degree of sparsity. Compared to other norms, the Lp pseudo-norm has better absorption and characterization capabilities; therefore, the Lp norm is used in the second-order sparse regularization constraint term. Simultaneously, to ensure the convergence of the model, the L1 norm is used in the global sparse constraint term. This significantly reduces noise that was originally distributed throughout the entire frequency domain, and the signal is concentrated within the ideal chest cavity signal frequency range. Motion noise is essentially filtered out, demonstrating better application prospects in processing vital sign signals acquired by millimeter-wave radar. Attached Figure Description

[0060] Figure 1 This is a flowchart of a motion noise removal method for millimeter-wave radar signals based on total variation according to the present invention.

[0061] Figure 2 This is the time-domain image of the signal before denoising in this invention;

[0062] Figure 3 This is the signal spectrum before noise reduction in this invention;

[0063] Figure 4 This is a time-domain image of the denoised signal using the motion noise removal method of the present invention;

[0064] Figure 5 This is a spectrum of the denoised signal after using the motion noise removal method of the present invention. Detailed Implementation

[0065] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0066] Example

[0067] like Figure 1 As shown in the figure, an embodiment of the present invention proposes a motion noise removal method for millimeter-wave radar signals based on total variation, which includes the following steps:

[0068] S1. After obtaining the millimeter-wave radar acquisition data matrix from the millimeter-wave radar echo signal dataset, static clutter filtering, unwrapping, and differential processing are performed to obtain the phase difference signal. The phase difference signal matrix is ​​then processed by extracting the row vectors of each period according to the period of the thoracic signal. These extracted row vectors are then placed into each row of the thoracic signal grayscale matrix. Considering the periodicity of the signal, the newly constructed thoracic signal grayscale matrix should be relatively smooth when converted into a grayscale image, and should also have certain sparsity characteristics in its structure.

[0069] S2, after acquiring the chest cavity signal image containing motion noise, calculate the second-order gradient information in different directions of the image to obtain its horizontal second-order gradient, vertical second-order gradient, and mixed second-order gradient. The calculation formula for the gradient information in different directions of the image is as follows:

[0070]

[0071] Where: I(x,y) represents the pixel values ​​of the image, G xx (x,y) represents the second-order gradient information in the horizontal direction of the image, G yy (x,y) represents the second-order gradient information in the vertical direction of the image, G xy (x,y) represents the second-order mixed gradient information of the image, and (x,y) represents the row and column index of the image pixels.

[0072] S3, based on the second-order gradient information in different directions, obtains the sparsity characteristics of noise and the underlying image structure, and designs second-order regularization constraint terms in different directions by introducing the Lp pseudo-norm; at the same time, based on the prior characteristics of motion noise, a global sparsity constraint term is designed, the relevant expression of which is:

[0073] R0 = ||YZ||1

[0074]

[0075]

[0076]

[0077] Where: Y represents an ideal image without noise, Z represents an image contaminated by noise, and L... x =[-1,1] represents the convolution kernel with lateral difference, L y =[-1,1] T R represents the convolution kernel for vertical difference, R0 represents the global sparse constraint term, and R xx R represents the level second-order regularization constraint term. yy R represents the vertical second-order regularization constraint term. xy This represents the mixed second-order regularization constraint term.

[0078] Specifically, in the field of image processing, the Lp norm is generally defined as... Lp pseudonorm is defined as

[0079] S4 combines the second-order regularization constraints in different directions and the global sparsity constraints in S3 to obtain the complete total variational model expression for removing random motion noise:

[0080]

[0081] Where μ represents the coefficient of the second-order sparse regularization term.

[0082] S5, the alternating multiplier method is used to iteratively solve the noise-removed total variational model in S4, including the following steps:

[0083] S501, by introducing intermediate variables, transforms a constrained problem into an unconstrained problem and obtains the augmented Lagrangian function of the problem.

[0084] The augmented Lagrangian expression for the unconstrained problem is:

[0085]

[0086] Where: A0 = YZ, A1 = L x *L x *Y, A2 = L y *L y *Y, A3 = L x *L y *Y, where Λ0, Λ1, Λ2 and Λ3 represent Lagrange multipliers, γ0, γ1, γ2 and γ3 represent the coefficients of the penalty term, max represents taking the maximum value, and min represents taking the minimum value.

[0087] S502 solves the subproblems of each variable in the augmented Lagrangian function one by one, and obtains their iterative formulas.

[0088] (1) From the above augmented Lagrangian function, the subproblem of Y can be obtained as follows:

[0089]

[0090] Using the convolution theorem and the fast Fourier transform, the iterative formula for Y is obtained as follows:

[0091]

[0092]

[0093]

[0094] In the formula: Operators representing Fourier transforms The operator representing the inverse Fourier transform, express Conjugate mappings, where · represents dot product and / represents dot division.

[0095] (2) From the above augmented Lagrangian function, the subproblems of the intermediate variables A0, A1, A2, and A3 can be obtained as follows:

[0096]

[0097]

[0098]

[0099]

[0100] Using the contraction theorem, the iterative formulas for the intermediate variables A0, A1, A2, and A3 are obtained as follows:

[0101]

[0102]

[0103]

[0104]

[0105] In the formula:

[0106] (3) Finally, using the gradient ascent method, the iterative formulas for the Lagrange multipliers Λ0, Λ1, Λ2, and Λ3 are obtained as follows:

[0107]

[0108]

[0109]

[0110]

[0111] Where λ is the learning rate.

[0112] S503, after multiple iterations of each variable until the relevant conditions or the maximum number of iterations are met, the optimal solution of Y can be obtained, that is, the chest cavity signal image after motion noise removal.

[0113] This method transforms one-dimensional physiological signals acquired using millimeter-wave radar into two-dimensional image signals. It then utilizes the noise and sparsity of the underlying image structure for denoising, effectively removing noise while better preserving the detailed features of the chest cavity signal, providing a new approach to physiological signal denoising. Simultaneously, considering motion noise and the sparsity of the underlying image structure, the second-order gradient information of the chest cavity signal image is used when solving the proposed motion-noise-removing total variational model. The second-order gradient information of the image can better highlight noise and possesses a certain degree of sparsity. Compared to other norms, the Lp pseudo-norm has better absorption and characterization capabilities; therefore, the Lp norm is used in the second-order sparse regularization constraint term, while the L1 norm is used in the global sparse constraint term to ensure model convergence. This significantly reduces noise that was originally distributed throughout the entire frequency domain, and the signal is concentrated within the ideal chest cavity signal frequency range, with motion noise being largely filtered out, thus improving the quality of the chest cavity signal. This can contribute to the subsequent application of vital sign signals acquired by millimeter-wave radar in rescue, medical, and other fields; Figure 2 , Figure 3 The text displays signals containing motion noise. Figure 4 , Figure 5 The figure shows the results of removing motion noise using the method presented in this paper.

[0114] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for removing motion noise from millimeter-wave radar signals based on total variation, characterized in that, Specifically, the following steps are included: S1: After obtaining the data matrix from the millimeter-wave radar echo signal dataset, static clutter filtering, unwrapping and differential processing are performed to obtain the phase difference signal matrix. The phase difference signal matrix is ​​rearranged to obtain the thoracic cavity signal matrix. The thoracic cavity signal matrix is ​​used to create a grayscale image of the thoracic cavity signal sample. S2: After obtaining the grayscale image of the chest cavity signal sample containing motion noise, calculate the second-order gradient and second-order mixed gradient information of the horizontal and vertical directions of the image respectively; S3: Based on the second-order gradient information in different directions, the sparsity characteristics of motion noise and underlying image structure are obtained. Second-order sparse regularization constraints in different directions are designed by introducing the Lp pseudo-norm. At the same time, a global sparsity constraint is designed based on the prior characteristics of motion noise. S4: Combining the second-order sparse regularization constraints in different directions and the global sparse constraints in S3, we obtain a complete total variational model that removes motion noise. S5: The alternating multiplier method is used to iteratively solve the total variational model in S4 to remove motion noise, and finally the motion noise-free chest cavity signal image can be obtained; the chest cavity signal image is converted into a one-dimensional chest cavity signal.

2. The method for removing motion noise from millimeter-wave radar signals based on total variation as described in claim 1, characterized in that: The formula for calculating gradient information in different directions of an image is: Where: I(x,y) represents the pixel values ​​of the image, G xx (x,y) represents the second-order gradient information in the horizontal direction of the image, G yy (x,y) represents the second-order gradient information in the vertical direction of the image, G xy (x,y) represents the second-order mixed gradient information of the image, and (x,y) represents the row and column index of the image pixels.

3. The method for removing motion noise from millimeter-wave radar signals based on total variation as described in claim 1, characterized in that: The global constraint term and the second-order sparse regularization constraint terms in different directions are as follows: R0=||YZ||1 Where: Y represents an ideal image without noise, Z represents an image contaminated by noise, and L... x =[-1,1] represents the convolution kernel with lateral difference, L y =[-1,1] T R represents the convolution kernel for vertical difference, R0 represents the global sparse constraint term, and R xx R represents the level second-order regularization constraint term. yy R represents the vertical second-order regularization constraint term. xy This represents the mixed second-order regularization constraint term.

4. The method for removing motion noise from millimeter-wave radar signals based on total variation as described in claim 1, characterized in that: The complete total variational model expression for noise removal is as follows: Where μ represents the coefficient of the second-order sparse constraint regularization term.

5. The method for removing motion noise from millimeter-wave radar signals based on total variation as described in claim 1, characterized in that: Solving the total variational model using the alternating direction multiplier method includes the following steps: S501, by introducing intermediate variables, the constrained problem is transformed into an unconstrained problem, and the augmented Lagrangian function of the problem is obtained; The augmented Lagrangian function for this problem is: Where: A0 = YZ, A1 = L x *L x *Y, A2 = L y *L y *Y, A3 = L x *L y *Y, where Λ0, Λ1, Λ2 and Λ3 represent Lagrange multipliers, γ0, γ1, γ2 and γ3 represent the coefficients of the penalty term, max represents taking the maximum value, and min represents taking the minimum value; S502, solve the subproblems of each variable in the augmented Lagrangian function one by one, and introduce auxiliary variables to obtain their iterative formulas respectively; S503, after multiple iterations of each variable, obtains the optimal solution for Y, which is the chest cavity signal image after noise removal.

Images