A method and apparatus for radar health monitoring based on physiological signal reconstruction

Abstract

A method for radar health monitoring based on physiological signal reconstruction includes: receiving physiological signals corresponding to a target human body, performing delay domain processing on the physiological signals to remove the initial heartbeat signal and additive noise components, and generating a reconstructed respiratory signal; determining a residual signal based on the physiological signal and the reconstructed respiratory signal, performing delay domain processing on the residual signal to remove the residual respiratory signal and the additive noise components, and generating an estimated heartbeat signal; performing frequency band decomposition and filtering on the estimated heartbeat signal according to a preset frequency band to generate a reconstructed heartbeat signal; and generating a health monitoring result based on the reconstructed respiratory signal and the reconstructed heartbeat signal. By reconstructing the physiological signal through delay domain processing and frequency band decomposition and filtering, respiratory harmonics can be effectively suppressed, and accurate reconstruction of the heartbeat signal can be achieved, meeting the accuracy requirements for analyzing complex physiological states of the human body in daily scenarios.

Description

Technical Field

[0001] This invention relates to the field of health monitoring, and more specifically to a method and apparatus for radar health monitoring based on physiological signal reconstruction. Background Technology

[0002] With the rapid development of IoT technology, non-contact radar health monitoring systems have received widespread attention in the field of smart healthcare. Among them, the physiological signal reconstruction method refers to the method of reconstructing human breathing and heartbeat signals based on human physiological characteristics using radar echo signals, providing technical support for radar health monitoring systems.

[0003] Current methods for reconstructing physiological signals mainly include the following three types: time-domain analysis, which extracts physiological signals based on time-domain characteristics, such as coherent accumulation; frequency-domain analysis, which separates physiological signals based on the differences in the spectral distribution of respiratory and heartbeat signals, such as frequency-domain filtering; and time-frequency analysis, which uses nonlinear and non-stationary instantaneous signal processing methods to separate physiological signals, such as wavelet transform, short-time Fourier transform, and mode decomposition. However, none of these methods can effectively solve the problem of poor physiological signal separation quality caused by the spectral aliasing of respiratory harmonics and heartbeat components. Summary of the Invention

[0004] In view of the aforementioned problems, this application is made to provide a method and apparatus for radar health monitoring based on physiological signal reconstruction that overcomes or at least partially solves the aforementioned problems, comprising:

[0005] A method for radar health monitoring based on physiological signal reconstruction, wherein the method monitors human health by periodically detecting physiological signals of the human body using radar, wherein the physiological signals include an initial respiratory signal, an initial heartbeat signal, and an additive noise component; including:

[0006] The system receives physiological signals corresponding to the target human body, performs delay domain processing on the physiological signals, removes the initial heartbeat signal and additive noise components, and generates a reconstructed respiratory signal.

[0007] The residual signal is determined based on the physiological signal and the reconstructed respiratory signal, and the residual signal is processed in the delay domain to remove the residual respiratory signal and the additive noise component, thereby generating the predicted heartbeat signal.

[0008] The estimated heartbeat signal is decomposed and filtered according to a preset frequency band to generate a reconstructed heartbeat signal;

[0009] Health monitoring results are generated based on the reconstructed respiratory signal and the reconstructed heartbeat signal.

[0010] Further, the step of receiving physiological signals corresponding to the target human body includes:

[0011] Radar phase information is generated based on radar signals;

[0012] The chest wall change signal is determined based on the radar phase information;

[0013] Physiological signals are determined based on the chest wall change signals.

[0014] Further, the step of receiving physiological signals corresponding to the target human body, performing delay domain processing on the physiological signals to remove the initial heartbeat signal and additive noise components, and generating a reconstructed respiratory signal includes:

[0015] A high-dimensional phase space of the physiological signal is constructed based on preset parameters, including embedding dimension, sampling delay, projection dimension, and neighborhood radius.

[0016] Obtain the low-dimensional space corresponding to the respiratory signal, and project the high-dimensional phase space onto the low-dimensional space to generate a reconstructed respiratory signal.

[0017] Further, the step of determining the residual signal based on the physiological signal and the reconstructed respiratory signal, and performing delay domain processing on the residual signal to remove the residual respiratory signal and additive noise components, and generating the predicted heartbeat signal, includes:

[0018] A residual signal is generated based on the physiological signals and the reconstructed respiratory signals;

[0019] The residual signal is subjected to local projection noise reduction processing to obtain the estimated heartbeat signal.

[0020] Furthermore, the step of performing frequency band decomposition and filtering on the estimated heartbeat signal based on a preset frequency band to generate a reconstructed heartbeat signal includes:

[0021] Based on the preset wavelet packet decomposition level, the estimated heartbeat signal is divided into several equal bandwidth frequency bands;

[0022] Target frequency band nodes are determined based on the heartbeat fundamental frequency band;

[0023] The energy threshold is determined based on the total energy of the target node;

[0024] The summation signal is determined based on the energy threshold, and the summation signal is summed to obtain the reconstructed heartbeat signal.

[0025] Further, the step of generating health monitoring results based on the reconstructed respiratory signal and the reconstructed heartbeat signal includes:

[0026] Extract the maximum points of the reconstructed heartbeat signal to generate a candidate set of R peaks;

[0027] The R-peaks in the candidate R-peak set are filtered and supplemented according to preset filtering conditions;

[0028] Heart rate data are determined based on the number of R peaks in a preset time period;

[0029] Extract the respiratory rate data from the reconstructed respiratory signal;

[0030] Based on the heart rate data and respiratory rate data, a health monitoring result containing physiological parameter values ​​and trend analysis is generated.

[0031] Furthermore, the step of filtering and completing the R-peaks in the candidate R-peak set according to preset filtering conditions includes:

[0032] Based on the average RR interval and the minimum RR interval sampling points, the candidate set of R peaks is screened and the missed R peaks are filled in; wherein, the initial value of the average RR interval is the reciprocal of the estimated heart rate fundamental frequency, and the minimum RR interval sampling point is 0.6 multiplied by the average RR interval and then multiplied by the signal sampling frequency; when the number of R peak points is greater than 5, the average RR interval and the minimum RR interval sampling points are updated in real time.

[0033] A radar health monitoring device based on physiological signal reconstruction, wherein the device monitors human health by periodically detecting physiological signals of the human body using radar, wherein the physiological signals include an initial respiratory signal, an initial heartbeat signal, and an additive noise component; comprising:

[0034] The respiratory signal reconstruction module is used to receive physiological signals corresponding to the target human body, perform delay domain processing on the physiological signals, remove the initial heartbeat signal and additive noise components, and generate a reconstructed respiratory signal.

[0035] The heartbeat prediction signal module is used to determine the residual signal based on the physiological signal and the reconstructed respiratory signal, and to perform delay domain processing on the residual signal to remove the residual respiratory signal and the additive noise component, thereby generating the heartbeat prediction signal.

[0036] The reconstructed heartbeat signal module is used to perform frequency band decomposition and filtering on the estimated heartbeat signal according to a preset frequency band to generate a reconstructed heartbeat signal.

[0037] The result generation module is used to generate health monitoring results based on the reconstructed respiratory signal and the reconstructed heartbeat signal.

[0038] A device for radar health monitoring based on physiological signal reconstruction includes a processor, a memory, and a computer program stored in the memory and capable of running on the processor. When the computer program is executed by the processor, it implements any of the methods for radar health monitoring based on physiological signal reconstruction as described above.

[0039] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the radar health monitoring method based on physiological signal reconstruction as described above.

[0040] This application has the following advantages:

[0041] In the embodiments of this application, in contrast to the problem of poor physiological signal separation quality caused by the spectral aliasing of respiratory harmonics and heartbeat components in the prior art, this application provides a solution for physiological signal reconstruction using local projection denoising and wavelet packet decomposition. Specifically, it is a radar health monitoring method based on physiological signal reconstruction. The method monitors the health of the human body by periodically detecting physiological signals of the human body using radar. The physiological signals include an initial respiratory signal, an initial heartbeat signal, and an additive noise component. The method includes: receiving physiological signals corresponding to the target human body and performing delay domain processing on the physiological signals to remove the initial heartbeat signal and the additive noise component, generating a reconstructed respiratory signal; determining residual signals based on the physiological signals and the reconstructed respiratory signals, and performing delay domain processing on the residual signals to remove the residual respiratory signals and the additive noise component, generating an estimated heartbeat signal; performing frequency band decomposition and filtering on the estimated heartbeat signal according to a preset frequency band to generate a reconstructed heartbeat signal; and generating a health monitoring result based on the reconstructed respiratory signals and the reconstructed heartbeat signals. By reconstructing physiological signals through time-domain processing and frequency band decomposition and screening, respiratory harmonics can be effectively suppressed, and accurate reconstruction of heartbeat signals can be achieved. The estimation of respiratory rate and heart rate shows high accuracy and good robustness in actual observation datasets in different scenarios, meeting the accuracy requirements for analyzing complex physiological states of the human body in daily scenarios. Attached Figure Description

[0042] To more clearly illustrate the technical solution of this application, the drawings used in the description of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is a flowchart illustrating the steps of a radar health monitoring method based on physiological signal reconstruction according to an embodiment of this application;

[0044] Figure 2 This is a radar signal reception diagram of a radar health monitoring method based on physiological signal reconstruction provided in an embodiment of this application;

[0045] Figure 3 This is a wavelet packet decomposition diagram of a radar health monitoring method based on physiological signal reconstruction provided in an embodiment of this application;

[0046] Figure 4 This is a frequency domain distribution diagram of a respiratory signal reconstructed by Algorithm 1 according to an embodiment of this application;

[0047] Figure 5 This is a frequency domain distribution diagram of a heartbeat signal reconstructed by Algorithm 1 according to an embodiment of this application;

[0048] Figure 6 This is a waveform of the reconstructed respiratory signal in the time domain provided in an embodiment of this application;

[0049] Figure 7 This is a waveform of the reconstructed heartbeat signal in the time domain provided in an embodiment of this application;

[0050] Figure 8 This is a distribution diagram of the respiratory rate estimation results of the scheme provided in one embodiment of this application and the WPD algorithm;

[0051] Figure 9 This is a probability distribution diagram of the cumulative error of the present solution and the WPD algorithm in respiratory rate estimation provided in an embodiment of this application;

[0052] Figure 10 This is a distribution diagram of the heart rate estimation results provided by an embodiment of this application and the WPD algorithm;

[0053] Figure 11 This is a probability distribution diagram of the cumulative error of the scheme provided in this application and the WPD algorithm in heart rate estimation, according to an embodiment of this application.

[0054] Figure 12 This is a graph showing the IBI error results estimated by several methods according to an embodiment of this application;

[0055] Figure 13 This is a diagram showing the SDNN error results estimated by several methods according to an embodiment of this application;

[0056] Figure 14 This is a graph showing the RMSSD error results estimated by several methods according to an embodiment of this application;

[0057] Figure 15 This is a graph showing the PNN50 error results estimated by several methods according to an embodiment of this application;

[0058] Figure 16 This is a structural block diagram of a radar health monitoring device based on physiological signal reconstruction provided in one embodiment of this application;

[0059] Figure 17 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention;

[0060] 1. Computer equipment; 2. External devices; 3. Processing unit; 4. Bus; 5. Network adapter; 6. I / O interface; 7. Display; 8. Memory; 9. Random access memory; 10. Cache memory; 11. Storage system; 12. Program / utility; 13. Program module. Detailed Implementation

[0061] To make the objectives, features, and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0062] The inventors discovered through analysis of existing technologies that current physiological signal reconstruction methods are generally time-domain analysis, frequency-domain analysis, or time-frequency analysis, but none of them can effectively solve the problem of poor physiological signal separation quality caused by the spectral aliasing of respiratory harmonics and heartbeat components.

[0063] Reference Figure 1 This application illustrates a radar health monitoring method based on physiological signal reconstruction according to an embodiment of the present application. The method monitors human health by periodically detecting physiological signals of the human body using radar. The physiological signals include an initial respiratory signal, an initial heartbeat signal, and an additive noise component.

[0064] S110. Receive physiological signals corresponding to the target human body, and perform delay domain processing on the physiological signals to remove the initial heartbeat signal and additive noise components, and generate a reconstructed respiratory signal.

[0065] S120. Determine the residual signal based on the physiological signal and the reconstructed respiratory signal, and perform delay domain processing on the residual signal to remove the residual respiratory signal and the additive noise component, thereby generating the predicted heartbeat signal.

[0066] S130. The estimated heartbeat signal is decomposed and filtered according to a preset frequency band to generate a reconstructed heartbeat signal.

[0067] S140. Generate health monitoring results based on the reconstructed respiratory signal and the reconstructed heartbeat signal.

[0068] In the embodiments of this application, in contrast to the problem of poor physiological signal separation quality caused by the spectral aliasing of respiratory harmonics and heartbeat components in the prior art, this application provides a solution for physiological signal reconstruction using local projection denoising and wavelet packet decomposition. Specifically, it is a radar health monitoring method based on physiological signal reconstruction. The method monitors the health of the human body by periodically detecting physiological signals of the human body using radar. The physiological signals include an initial respiratory signal, an initial heartbeat signal, and an additive noise component. The method includes: receiving physiological signals corresponding to the target human body and performing delay domain processing on the physiological signals to remove the initial heartbeat signal and the additive noise component, generating a reconstructed respiratory signal; determining residual signals based on the physiological signals and the reconstructed respiratory signals, and performing delay domain processing on the residual signals to remove the residual respiratory signals and the additive noise component, generating an estimated heartbeat signal; performing frequency band decomposition and filtering on the estimated heartbeat signal according to a preset frequency band to generate a reconstructed heartbeat signal; and generating a health monitoring result based on the reconstructed respiratory signals and the reconstructed heartbeat signals. By reconstructing physiological signals through time-domain processing and frequency band decomposition and screening, respiratory harmonics can be effectively suppressed, and accurate reconstruction of heartbeat signals can be achieved. The estimation of respiratory rate and heart rate shows high accuracy and good robustness in actual observation datasets in different scenarios, meeting the accuracy requirements for analyzing complex physiological states of the human body in daily scenarios.

[0069] The following will further explain a radar health monitoring method based on physiological signal reconstruction in this exemplary embodiment.

[0070] As described in step S110, physiological signals corresponding to the target human body are received, and the physiological signals are processed in the delay domain to remove the initial heartbeat signal and additive noise components, thereby generating a reconstructed respiratory signal.

[0071] It should be noted that by receiving the radar reflection signal modulated by the chest wall displacement of the target human body, and extracting the phase information of the radar signal, the chest wall displacement signal is derived. This signal is the original physiological signal containing the initial respiratory signal, the initial heartbeat signal, and additive noise components.

[0072] Parameter presets: Based on the time scale difference between respiration and heartbeat (respiration time scale [1.6-10]s, heartbeat time scale [0.5-1.25]s), the high-dimensional phase space reconstruction parameters are preset as follows: the embedding dimension is set to 20 to ensure that the embedding window includes the complete heartbeat cycle and is smaller than the respiratory cycle; the sampling delay is set to 30 to avoid confusion between respiratory and heartbeat signal components; the projection dimension is set to 3 to ensure that only the low-dimensional spatial components corresponding to the respiratory signal are extracted; the neighborhood radius is processed by a high-pass filter with a cutoff frequency of 0.8Hz to process the original physiological signal, and the median of the range of the column vectors of the filtered signal matrix is ​​calculated. 1.5 times this median (empirical coefficient) is taken to ensure that the neighborhood range covers the amplitude of the heartbeat component.

[0073] Constructing a high-dimensional phase space: Based on the Takens embedding principle, and with a preset embedding dimension and sampling delay, the one-dimensional original physiological signal is reconstructed into a high-dimensional phase space. Each column vector in this phase space is a phase point, and the total number of phase points is determined by the length of the original signal and the embedding dimension.

[0074] Identifying the low-dimensional space of respiratory signals: Calculate the total number of phase points in the neighborhood of each phase point with a radius equal to the preset neighborhood radius, determine the centroid of the neighborhood and construct the neighborhood matrix, and then generate the neighborhood covariance matrix; perform eigenvalue decomposition on the covariance matrix to obtain eigenvectors and eigenvalues ​​arranged in descending order, where the eigenvectors corresponding to the first 3 larger eigenvalues ​​constitute the low-dimensional space (projection dimension is 3) where the respiratory signal is located, and the eigenvectors corresponding to the remaining eigenvalues ​​constitute the noise space containing the initial heartbeat signal and additive noise.

[0075] Projection denoising generates reconstructed respiratory signals: Subtract the projection component in the noise space from each phase point in the phase space to remove the initial heartbeat signal and additive noise components; average the time-series signals in all updated phase points to obtain a reconstructed respiratory signal that retains only the effective components of the initial respiratory signal.

[0076] In one embodiment of the present invention, the specific process of "receiving physiological signals corresponding to the target human body" in step S110 can be further described in conjunction with the following description.

[0077] As described in the following steps, radar phase information is generated based on the radar signal;

[0078] It should be noted that the radar signal received from a monostatic radar and modulated by the chest wall displacement of the target human body is a high-frequency electromagnetic signal with a fixed carrier frequency, containing characteristic parameters such as amplitude, phase, and frequency. By using signal demodulation technology to separate the phase component of the radar reflection signal, eliminating amplitude noise caused by signal attenuation and environmental interference, radar phase information directly related to chest wall displacement is extracted. This phase information exhibits periodic changes with chest wall respiration and heartbeat, and its variation pattern is consistent with the time characteristics of chest wall displacement.

[0079] The chest wall change signal is determined based on the radar phase information as described in the following steps;

[0080] It should be noted that a mapping relationship between phase information and chest wall displacement is established based on the physical correlation between radar phase and range (phase change is proportional to target range change). Since a 2π phase change in the radar signal corresponds to a range change of λ / 2 (λ is the radar signal wavelength), a phase unwrapping algorithm is used to process the radar phase information, eliminating phase ambiguity and converting periodic phase changes into continuous range changes. This range change is the chest wall displacement signal over time (chest wall change signal), which includes low-frequency large-amplitude displacement components caused by respiratory activity and high-frequency small-amplitude displacement components caused by cardiac activity.

[0081] Physiological signals are determined based on the chest wall change signals as described in the following steps.

[0082] It should be noted that the chest wall change signal is preprocessed (e.g., DC component removal and baseline drift elimination) to obtain a signal containing dynamic change components. This signal directly reflects the vibration state of the chest wall, and its core component is the physiological signal. Among them, the low-frequency component corresponds to the initial respiratory signal, the high-frequency component corresponds to the initial heartbeat signal, and the signal is also superimposed with additive noise components formed by radar receiver link noise, environmental electromagnetic interference, etc. The final result is a physiological signal containing the initial respiratory signal, the initial heartbeat signal, and additive noise components.

[0083] In one embodiment of the present invention, the specific process of "performing delay domain processing on the physiological signal, removing the initial heartbeat signal and additive noise components, and generating a reconstructed respiratory signal" in step S110 can be further described in conjunction with the following description.

[0084] As described in the following steps, a high-dimensional phase space of the physiological signal is constructed based on preset parameters, including embedding dimension, sampling delay, projection dimension, and neighborhood radius;

[0085] It should be noted that the embedding dimension is set to 20 to ensure that the embedding window includes the complete heartbeat cycle and is smaller than the respiratory cycle, adapting to the time scale difference between respiration and heartbeat; the sampling delay is set to 30 to avoid confusion between respiratory and heartbeat signal components in phase space; the projection dimension is set to 3 to characterize the dimension of the low-dimensional space in which the respiratory signal is located, ensuring that only the effective respiratory components are extracted; the neighborhood radius is calculated by applying a high-pass filter with a cutoff frequency of 0.8 Hz to the original physiological signal, calculating the range (peak-to-peak value) of the column vectors of the filtered signal matrix, taking the median of all ranges, and using 1.5 times this median as the neighborhood radius, ensuring that the neighborhood range covers the amplitude of the heartbeat component and avoiding mistakenly including the heartbeat signal in the respiratory signal space. Based on the Takens embedding principle, and with a preset embedding dimension and sampling delay, a one-dimensional original physiological signal is reconstructed: starting from the beginning of the physiological signal, signal segments with a length equal to the embedding dimension are extracted sequentially, with adjacent segments spaced apart by a sampling delay of 100 sampling points. Each signal segment serves as a column vector (phase point) in a high-dimensional phase space, and all phase points together constitute a high-dimensional phase space matrix. The total number of phase points is determined by the length of the original physiological signal, the embedding dimension, and the sampling delay.

[0086] As described in the following steps, the low-dimensional space corresponding to the respiratory signal is obtained, and the high-dimensional phase space is projected onto the low-dimensional space to generate a reconstructed respiratory signal.

[0087] It should be noted that, firstly, each phase point in the high-dimensional phase space is traversed, and a neighborhood of each phase point is determined within a preset neighborhood radius. The total number of phase points within the neighborhood is counted, and the centroid of the neighborhood is calculated. A neighborhood matrix is ​​constructed based on all phase points within the neighborhood. The covariance matrix of this neighborhood matrix is ​​solved, and the eigenvalues ​​and corresponding eigenvectors are obtained in descending order through eigenvalue decomposition. Since the respiratory signal is a signal generated by a low-dimensional dynamic system, its corresponding energy concentration in the high-dimensional phase space is a low-dimensional subspace. The eigenvectors corresponding to the top three largest eigenvalues ​​are selected, and the space spanned by these three eigenvectors is the low-dimensional space corresponding to the respiratory signal. This space only carries the effective respiratory components, while the initial heartbeat signal and additive noise components are distributed in the remaining high-dimensional spaces. Each phase point in the high-dimensional phase space is projected onto the aforementioned low-dimensional space, and the high-dimensional space components belonging to the initial heartbeat signal and additive noise components in the phase point are removed. The time-series signals within all phase points after projection are averaged to obtain a signal that retains only the effective respiratory components. This signal is the reconstructed respiratory signal.

[0088] As an example, such as Figure 2 As shown, the radar health monitoring system includes a monostatic radar with combined transmit and receive capabilities, and a radar that detects human targets. The signal emitted by the radar is modulated by the displacement of the human chest cavity and then received by the radar. Considering a one-dimensional scenario with a single antenna, This indicates the initial distance between the human chest wall and the radar. This indicates the displacement changes of the chest wall over time;

[0089] Radar signals received from the human body can be represented as:

[0090] ,

[0091] in, Indicates the transmission of a pulse signal. This indicates the pulse repetition frequency of the radar. This indicates the arrival time delay of the echo signal reflected from the human body. This indicates the signal carrier frequency. After determining the initial position of the human body, the phase information of the radar signal can be extracted.

[0092] ,

[0093] Further, the chest wall displacement signal can be extracted:

[0094] ,

[0095] in, and These represent chest wall displacement caused by respiration and heartbeat, respectively.

[0096] Actual physiological signal sampling data can be represented as:

[0097] ,

[0098] These include respiratory signals. Heartbeat signal and additive noise components .

[0099] As described in step S120, the residual signal is determined based on the physiological signal and the reconstructed respiratory signal, and the residual signal is processed in the delay domain to remove the residual respiratory signal and the additive noise component, thereby generating the estimated heartbeat signal.

[0100] It should be noted that the original physiological signal includes the initial respiratory signal, the initial heartbeat signal, and an additive noise component, while the reconstructed respiratory signal has removed the initial heartbeat signal and the additive noise component, retaining only the effective component of the initial respiratory signal. By calculating the difference between the original physiological signal and the reconstructed respiratory signal, the residual signal can be separated: the core component of this residual signal is the initial heartbeat signal, but it still contains residual respiratory signal components that were not completely removed (such as respiratory higher harmonics) and residual additive noise components, forming a mixed signal of "initial heartbeat signal + residual respiratory signal components + additive noise components".

[0101] The residual signal is processed in the delay domain to remove the residual respiratory signal and the additive noise component. First, the parameters for the residual signal delay domain processing are preset: the embedding dimension is set to 10, which is less than the time scale of the heartbeat signal [0.5-1.25]s, to ensure that the embedding window adapts to the periodic characteristics of the heartbeat signal; the sampling delay is set to 10 to avoid signal component confusion; the projection dimension is determined adaptively by performing eigenvalue decomposition on the phase space neighborhood covariance matrix constructed by the residual signal, and selecting the point with the most drastic change in adjacent eigenvalues ​​as the projection dimension to adapt to the unknown dynamic system dimension of the heartbeat signal; the neighborhood radius is obtained by processing the residual signal with a high-pass filter with a cutoff frequency of 2.5 Hz, estimating the high-frequency noise amplitude, and taking 1.5 times this amplitude as the neighborhood radius. Based on the aforementioned preset parameters, a high-dimensional phase space for the residual signal is constructed according to the Takens embedding principle: residual signal segments are truncated using the embedding dimension and sampling delay to form a phase point matrix in the phase space; each phase point is traversed, its neighborhood range is determined, and the neighborhood centroid is calculated to construct a neighborhood matrix and a covariance matrix; eigenvalue decomposition yields eigenvalues ​​and eigenvectors arranged in descending order. The eigenvectors corresponding to larger eigenvalues ​​span the low-dimensional space corresponding to the heartbeat signal, while the eigenvectors corresponding to smaller eigenvalues ​​constitute the noise space containing the residual respiratory signal components and additive noise components; the projection component of each phase point in the phase space is subtracted from its projection component in the noise space to eliminate the residual respiratory signal components and additive noise components.

[0102] The time-series signals within all phase points after delay domain processing are averaged to obtain a signal that retains only the effective components of the initial heartbeat signal. This signal is the predicted heartbeat signal.

[0103] In one embodiment of the present invention, the specific process of step S120, which involves "determining the residual signal based on the physiological signal and the reconstructed respiratory signal, performing delay domain processing on the residual signal, removing the residual respiratory signal and the additive noise component, and generating an estimated heartbeat signal", can be further described in conjunction with the following description.

[0104] As described in the following steps, a residual signal is generated based on the physiological signal and the reconstructed respiratory signal;

[0105] It should be noted that the original physiological signal is known to consist of the initial respiratory signal, the initial heartbeat signal, and additive noise components. The reconstructed respiratory signal, after delay domain processing, removes the initial heartbeat signal and additive noise components, retaining only the effective components of the initial respiratory signal. By performing point-by-point subtraction between the original physiological signal and the reconstructed respiratory signal, the residual signal can be obtained. The main component of this residual signal is the initial heartbeat signal, but it still contains residual respiratory signal components that were not completely removed (such as respiratory harmonics) and residual additive noise components. That is, residual signal = initial heartbeat signal + residual respiratory signal components + residual additive noise components.

[0106] As described in the following steps, the residual signal is subjected to local projection noise reduction processing to obtain the estimated heartbeat signal.

[0107] It should be noted that the preset parameters for local projection noise reduction are as follows: the embedding dimension is set to 10; the sampling delay is set to 10; the projection dimension is determined adaptively—eigenvalue decomposition is performed on the phase space neighborhood covariance matrix constructed from the residual signal, and the point with the most drastic change in adjacent eigenvalues ​​is selected as the projection dimension; the neighborhood radius is processed by a high-pass filter with a cutoff frequency of 2.5 Hz to process the residual signal, and the high-frequency noise amplitude is estimated and taken as 1.5 times the amplitude.

[0108] Constructing a high-dimensional phase space for the residual signal: Based on the Takens embedding principle, the one-dimensional residual signal is reconstructed using a preset embedding dimension and sampling delay. Signal segments with a length equal to the embedding dimension are extracted sequentially, with adjacent segments spaced apart by a sampling delay of 1000 sampling points. Each signal segment serves as a column vector in the high-dimensional phase space, and all phase points together constitute the high-dimensional phase space matrix of the residual signal.

[0109] Identifying the low-dimensional space corresponding to the heartbeat signal: Traverse each phase point in the high-dimensional phase space, determine the neighborhood of each phase point with a preset neighborhood radius, count the total number of phase points in the neighborhood and calculate the centroid of the neighborhood, construct a neighborhood matrix based on all phase points in the neighborhood, and then solve the neighborhood covariance matrix; perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors in descending order, where the space spanned by the eigenvectors corresponding to the larger eigenvalues ​​is the low-dimensional space corresponding to the heartbeat signal, and the space formed by the eigenvectors corresponding to the smaller eigenvalues ​​is the noise space where the residual respiratory signal components and residual additive noise components are located.

[0110] Projection denoising and signal reconstruction: Subtract the projection component in the noise space from each phase point in the phase space to completely eliminate the residual respiratory signal components and residual additive noise components; average the time-series signals in all phase points after projection processing to obtain a signal that retains only the effective components of the initial heartbeat signal, which is the predicted heartbeat signal.

[0111] As an example, when dealing with one-dimensional raw physiological signals Respiratory signals during high-dimensional phase space delay reconstruction and heartbeat signals Due to differences in intensity and the dimensionality of the associated dynamic system (lungs, heart), different distribution characteristics will be exhibited in high-dimensional phase space. Therefore, this invention employs the Local Projection Noise Reduction (LPNR) method to... and Initial separation is carried out.

[0112] respiratory signals In the separation, assuming the heartbeat component is a noise component, then This can be considered an observation signal generated by the respiratory dynamics system but interfered with by high-dimensional noise (heartbeat signal and additive noise). The Takens embedding principle states that when reconstructing the phase space... Embedding dimension ( When the dimension of the dynamical system is (i.e., the dimension of the dynamical system), It has the same topological characteristics as the original power system. In delay The following is the reconstruction formula:

[0113] ,

[0114] Each column vector To represent a phase point, the total One phase point.

[0115] exist In the middle, respiratory signals are in a low dimension (assuming it is). (step) space Within this space, due to the presence of noise components, components also exist in higher-dimensional spaces. Therefore, the core idea of ​​local projection denoising is to identify the low-dimensional space where the respiratory signal resides. The original signal is then projected into this space to achieve noise reduction.

[0116] Define phase point The neighborhood is , express The radius is The total number of phase points in the neighborhood, calculate the neighborhood. center of mass And construct a neighborhood matrix Thus, the neighborhood covariance matrix is ​​constructed. and to Performing eigenvalue decomposition yields eigenvectors sorted in descending order. and eigenvalues Assuming respiratory signals exist... In a 3D space, the smallest The eigenvectors corresponding to each eigenvalue To create a noise space, the phase point Subtracting the projected components in this space yields the estimated original signal phase points as follows:

[0117] ,

[0118] After updating each phase point, the reconstructed respiratory signal is obtained by averaging the time-series signals within each phase point. .

[0119] The performance of local projection denoising algorithms largely depends on the selection of parameters, including the embedding dimension. Sampling delay Neighborhood radius and projection dimension .

[0120] respiratory signals During the extraction, since the time scales of the respiratory and cardiac fundamental frequency components are different, namely [1.610] s and [0.5125] s respectively, in order to ensure that an embedding window (phase point) basically contains a complete heartbeat cycle and is shorter than the respiratory cycle, a selection was made. , To ensure the extraction of the respiratory fundamental frequency component while avoiding extraction errors... It contains a heartbeat component, and is selected Neighborhood radius It should be ensured that the amplitude is greater than that of the heartbeat component, therefore... The result is obtained through a high-pass filter with a cutoff frequency of 0.8 Hz. and according to the same parameters , Constructing phase space ,use The peak value of the estimated heartbeat component amplitude at each phase point is as follows:

[0121]

[0122] in This indicates that the range of each column vector of the matrix is ​​calculated. This represents the median of the vector. We choose 1.5 times the median of the amplitude of the heartbeat component estimated by each embedding window as the neighborhood radius (1.5 is an empirical value).

[0123] Extracting respiratory signals Then, a noisy heartbeat signal was obtained. The noise components include residual breathing harmonics and additive noise, therefore... The initial estimate of the heartbeat signal is obtained by performing projection denoising again. The principle is the same as above, and the parameters are selected as follows. (Smaller than the time scale of a heartbeat). However, since the system dimension of the heartbeat is unknown, an adaptive projection dimension estimation is used. The projection dimension is selected to be the point that causes the most drastic change in adjacent feature values; the neighborhood radius is... The selection of is the same as in the formula above, the only difference being that The amplitude of high-frequency noise is estimated using a high-pass filter with a cutoff frequency of 2.5 Hz.

[0124] As described in step S130, the estimated heartbeat signal is decomposed and filtered according to the preset frequency band to generate a reconstructed heartbeat signal.

[0125] It should be noted that the number of wavelet packet decomposition layers is preset first, and the frequency band division is determined based on the number of decomposition layers to ensure that the precision of the frequency band division can cover the heartbeat fundamental frequency range.

[0126] Define the preset heartbeat fundamental frequency band range, traverse all frequency band nodes after wavelet packet decomposition, filter out target frequency band nodes that contain the heartbeat fundamental frequency band, remove redundant frequency band nodes that do not contain the heartbeat fundamental frequency, and determine the effective frequency range of the heartbeat signal.

[0127] For all selected target frequency band nodes, calculate the energy of the signal corresponding to each node and sum them to obtain the total energy of the target frequency band nodes. Set the energy threshold as the total energy divided by the number of target frequency band nodes. This threshold is used to filter out the signal components with effective energy in the target nodes and eliminate residual noise in the frequency band.

[0128] Each target frequency band node is checked to see if its signal energy is not less than the set energy threshold. The signals of nodes with qualified energy are retained, and nodes with energy below the threshold are removed. All retained valid node signals are superimposed and summed to obtain a signal with clean frequency and thorough noise removal. This signal is the reconstructed heartbeat signal.

[0129] In one embodiment of the present invention, the specific process of step S130, "performing frequency band decomposition and filtering of the estimated heartbeat signal according to a preset frequency band to generate a reconstructed heartbeat signal," can be further explained in conjunction with the following description.

[0130] As described in the following steps, the estimated heartbeat signal is divided into several equal bandwidth frequency bands according to the preset wavelet packet decomposition level;

[0131] The target frequency band node is determined based on the heartbeat fundamental frequency band as described in the following steps;

[0132] The energy threshold is determined based on the total energy of the target node, as described in the following steps.

[0133] As described in the following steps, a summation signal is determined based on the energy threshold, and the summation signal is summed to obtain a reconstructed heartbeat signal.

[0134] In a specific implementation, the wavelet packet decomposition process is as follows: Figure 3 As shown, the preset wavelet packet decomposition layer is 3 layers, which can be flexibly adjusted according to the signal sampling frequency. The core requirement is to ensure that the frequency band division accuracy can cover the heartbeat fundamental frequency range.

[0135] The horizontal axis represents frequency. This indicates the signal sampling frequency, and finally the frequency band is divided into equal parts. A total of 8 frequency bands (compared to the output of Discrete Wavelet Transform (DWT)) , , , (This yields a more refined spectral division), with the vertical axis representing the decomposition order, from top to bottom as layers 0, 1, 2, and 3, with layer 0 being the highest. Represents the original signal .

[0136] The WPD algorithm can be mathematically represented as:

[0137]

[0138] in Indicates the first Layer The decomposed signal corresponding to each node. and These represent the coefficients of the approximation and detail filters, respectively. Indicates positional parameters.

[0139] For a given heartbeat signal conduct Layer wavelet packet decomposition, then the first The first layer The frequency band range of each node is bandwidth is . The specific settings depend on the sampling frequency. The bandwidth requirement is to meet the requirements of each node. To achieve sufficiently accurate spectral resolution, after wavelet packet decomposition, only nodes within the heartbeat fundamental frequency band are retained as follows:

[0140] ,

[0141] in The smallest interval containing the heart rate fundamental frequency band [0.8, 2] Hz is selected, and the frequency band is filtered according to the signal energy:

[0142] Calculate the total energy of the heartbeat fundamental frequency band .

[0143] Threshold set to .

[0144] Remove energy The node.

[0145] The reconstructed heartbeat signal is obtained by summing the remaining node signals. .

[0146] In another specific implementation, local projection denoising and wavelet packet decomposition for physiological signal reconstruction are achieved through the following algorithm 1:

[0147] Input: raw physiological signal x(n), frequency band of the heartbeat fundamental frequency signal [f hb ,f hc [: Embedding dimensions K1, K2; Time delay τ1, τ2; Projection dimensions ρ1, ρ2; Neighborhood radii r1, r2; Number of wavelet packet decomposition layers L;]

[0148] Output: Reconstructed respiratory signal b r (n) and heartbeat signal h r (n);

[0149] h r (n)=0; / / Initialize and reconstruct the heartbeat signal

[0150] b r (n)=LPNR(x(n),K1,τ1,r1,ρ1); / / Reconstruct respiratory signal

[0151] h r2 (n)=LPNR(x(n)-b r (n),K2,τ2,r2,ρ2); / / Preliminary extraction of heartbeat signal

[0152] w L，m (n)=WPD(h r2 ,L); / / Wavelet packet decomposition signal

[0153] m1=[f hb ×2 L+1 / f s ],m1=[f hb ×2 L+1 / f s / / Determine the heart rate band

[0154] / / Reconstruct the heartbeat signal according to the frequency domain energy distribution

[0155] ;

[0156] ;

[0157] For m=m1:m2do

[0158] if then

[0159] h r (n)=h r (n)+w L，m (n);

[0160] end if

[0161] end for

[0162] Return b r (n), h r (n)

[0163] Algorithm 1

[0164] As described in step S140, a health monitoring result is generated based on the reconstructed respiratory signal and the reconstructed heartbeat signal.

[0165] It should be noted that spectral analysis, such as Fast Fourier Transform, is performed on the reconstructed respiratory signal to identify the fundamental frequency band of the respiratory signal and extract the peak frequency within that band, which is the respiratory frequency. At the same time, the amplitude variation range of the respiratory signal is statistically analyzed to help determine the stability of breathing.

[0166] The maximum points of the reconstructed heartbeat signal are extracted. These maximum points correspond to the R-peaks of the heartbeat waveform and form a candidate set of R-peaks. Based on the dynamically updated average RR interval and minimum RR interval sampling points, the candidate set of R-peaks is filtered to remove redundant candidate peaks with a spacing smaller than the minimum RR interval sampling point. The filtered set of R-peaks is then supplemented to eliminate any missed detections. In the fourth step, the total number of R-peaks within one minute is counted, and this total number is the heart rate data.

[0167] It integrates respiratory rate and heart rate data over a continuous period to generate a time series trend curve; it presets a health reference range and compares the real-time extracted physiological parameters with the reference range to determine if there are any abnormalities; if parameters exceed the reference range or the trend curve changes abruptly, it is marked as an abnormal warning message.

[0168] By integrating the extracted physiological parameters such as respiratory rate and heart rate, along with trend analysis results and abnormal warning information, structured health monitoring results are generated.

[0169] In one embodiment of the present invention, the specific process of "generating health monitoring results based on the reconstructed respiratory signal and the reconstructed heartbeat signal" in step S140 can be further explained in conjunction with the following description.

[0170] As described in the following steps, the maximum points of the reconstructed heartbeat signal are extracted to generate a candidate set of R peaks;

[0171] It should be noted that the reconstructed heartbeat signal is preprocessed to reduce the interference of residual noise on peak identification. A peak detection algorithm is used to traverse the reconstructed heartbeat signal after processing. A peak identification threshold is set, which is 30%-50% of the maximum signal amplitude. Maximum points with instantaneous signal values ​​greater than the threshold and greater than the signal values ​​of two adjacent sampling points are selected. All maximum points that meet the conditions correspond to candidate R-peaks of the heartbeat waveform. The position information of these candidate points is organized to form a candidate R-peak set.

[0172] As described in the following steps, the R-peaks in the candidate R-peak set are filtered and completed according to preset filtering conditions;

[0173] The initial value of the average RR interval is the reciprocal of the estimated heart rate fundamental frequency. For example, if the estimated heart rate fundamental frequency is 1.2 Hz, the initial RR interval is approximately 0.83 seconds. The minimum RR interval sampling point is the rounded result of 0.6 multiplied by the average RR interval and then by the signal sampling frequency. For example, if the sampling frequency is 100 Hz and the RR interval is 0.83 seconds, the minimum RR interval sampling point is approximately 50. The candidate R-peak set is sorted in chronological order, and the candidate set is traversed: if the distance between two adjacent candidate R-peaks is less than the minimum RR interval sampling point, the candidate peak with smaller amplitude is removed, and the effective R-peak with larger amplitude is retained. When the number of R-peak points after screening is greater than 5, the average RR interval and the minimum RR interval sampling point are updated in real time to dynamically optimize the screening conditions. After screening, the set of R-peaks is checked: if the distance between two adjacent effective R-peaks is greater than 1.5 times the updated average RR interval, it is determined that there is a missed R-peak. Within this distance, signal points that meet the peak value conditions are searched again to complete the missed R-peaks and form a complete set of effective R-peaks.

[0174] As described in the following steps, heart rate data is determined based on the number of R peaks in a preset time period;

[0175] A preset statistical period is defined, which is then converted into a corresponding range of sampling points based on the signal sampling frequency. The total number of R-peaks within this range of sampling points in the valid R-peak set is counted. Since each R-peak corresponds to one heartbeat, this total number represents the number of heartbeats within the preset period and is directly used as the heart rate data. If the statistical period is not one minute, the heart rate per minute needs to be calculated proportionally.

[0176] The respiratory rate data of the reconstructed respiratory signal is extracted as described in the following steps;

[0177] A fast Fourier transform is performed on the reconstructed respiratory signal to convert the time-domain signal into a frequency-domain signal. Based on the respiratory fundamental frequency band, the spectral peak within the band is extracted, and the frequency corresponding to the peak is the respiratory fundamental frequency. The respiratory fundamental frequency is multiplied by 60 to obtain the number of breaths per minute, which is the respiratory rate data. At the same time, the amplitude fluctuation range of the respiratory signal is statistically analyzed as an auxiliary evaluation index of respiratory stability.

[0178] As described in the following steps, based on the heart rate data and the respiratory rate data, a health monitoring result containing physiological parameter values ​​and trend analysis is generated.

[0179] It should be noted that the system integrates real-time extracted heart rate data, respiratory rate data, and respiratory amplitude fluctuation range to form a physiological parameter table, clearly labeling the specific values, units, and collection time of each parameter. If continuous monitoring data is available, time-series trend curves of heart rate and respiratory rate are plotted to analyze parameter trends, such as whether heart rate continuously increases / decreases or respiratory rate fluctuates drastically. Preset health reference ranges, such as heart rate 60-100 beats / minute and respiratory rate 12-20 breaths / minute, are used to compare real-time parameters with these ranges to determine health status. If parameters exceed the reference range or the trend curve shows abrupt changes, such as a sudden increase in heart rate exceeding 20 beats / minute, an abnormal warning is generated, indicating the type of abnormality and potential health risks. Finally, the physiological parameter values, trend analysis charts, health status assessment, and abnormal warning information are integrated to generate a structured health monitoring report, which represents the final health monitoring result.

[0180] In one embodiment of the present invention, the specific process of "screening and completing the R-peaks in the candidate set of R-peaks according to preset screening conditions" can be further explained in conjunction with the following description.

[0181] As described in the following steps, the candidate set of R peaks is screened and missed R peaks are filled in based on the average RR interval and the minimum RR interval sampling point; wherein, the initial value of the average RR interval is the reciprocal of the estimated heart rate fundamental frequency, and the minimum RR interval sampling point is 0.6 multiplied by the average RR interval and then multiplied by the signal sampling frequency; when the number of R peak points is greater than 5, the average RR interval and the minimum RR interval sampling point are updated in real time.

[0182] It should be noted that the initial value of the mean RR interval is determined by the reciprocal of the estimated heart rate. For example, if the estimated heart rate is 1.2 Hz, then the initial mean RR interval = 1 / 1.2 ≈ 0.833 seconds. This value is the initial estimate of the heart rate cycle.

[0183] Minimum RR interval sampling points: The result is calculated by the formula "0.6 × average RR interval × signal sampling frequency" and rounded. For example, if the signal sampling frequency is 100 Hz, substituting it into the average RR interval above will give 0.6 × 0.833 × 100 ≈ 50 sampling points. This value is used to define the minimum spacing of effective R peaks and avoid misjudging redundant peaks.

[0184] Sort the generated candidate R-peaks in chronological order and iterate through all adjacent candidate R-peaks:

[0185] Calculate the distance between two adjacent candidate R peaks;

[0186] If the interval is less than the minimum RR interval sampling point, it is determined that there are redundant candidate peaks. The candidate peaks with smaller amplitudes are removed, and the candidate peaks with larger amplitudes are retained as valid R peaks.

[0187] If the interval is not less than the minimum RR interval sampling point, two candidate peaks are directly retained as valid R peaks, and finally a set of valid R peaks after preliminary screening is formed.

[0188] Real-time statistics of the number of peaks in the effective R-peak set after initial screening:

[0189] When the number of peaks is greater than 5, it indicates that enough effective heartbeat cycle data have been acquired. At this time, the parameters are updated based on the current effective R peak set: the new average RR interval = the average of the distances between all adjacent effective R peaks, and the new minimum RR interval sampling point = 0.6 × the updated average RR interval × the signal sampling frequency.

[0190] If the number of peak points is ≤5, do not update the parameters for now, and continue to use the initial parameters for filtering until the number of peak points meets the update conditions, so as to ensure that the parameters always match the actual heartbeat cycle characteristics.

[0191] Based on the updated average RR interval, a second traversal is performed on the filtered set of valid R peaks:

[0192] Calculate the actual distance between two adjacent valid R peaks. If the actual distance is greater than 1.5 × the updated average RR interval, it is determined that there is a missed R peak in this interval.

[0193] Within the signal segment corresponding to this interval, peak detection is re-executed to search for signal points that satisfy the condition that the amplitude is not less than 60% of the average amplitude of the effective R peak and the interval between the effective R peaks before and after is not less than the minimum RR interval sampling point, and these points are filled in as the missed R peaks.

[0194] After completion, the spacing between all adjacent R peaks is checked again to ensure that there are no omissions or redundancies, and finally a complete set of valid R peaks is formed.

[0195] In one specific implementation, the time-domain R-peak localization of the heartbeat signal is achieved through the following algorithm 2:

[0196] Input: Reconstructed heartbeat signal h r (n), signal sampling frequency f s : Candidate set of R peak positions C R ;

[0197] Output: Set of R peak locations S R ;

[0198] m n =0.8; / / Initial estimate of mean RR interval

[0199] δ n=Round(0.4×m) n ×f s / / Minimum RR interval sampling point, Round is rounded.

[0200] S R =C R [i]; / / C R Sort in ascending order, [i] represents the i-th value in the set.

[0201] fori=1:|C R |-1 do

[0202] C R =S R [end]; / / Current R peak, [end] represents the last value of the set.

[0203] n R =C R [i+1]; / / Next candidate value for R peak

[0204] Ifn R -C R <δ n then

[0205] Ifh r (n) R >h r (C) R then

[0206] S R [end]=n R ; / / S R Replace the last element with n R

[0207] else

[0208] Continue;

[0209] end if

[0210] else

[0211] S R =S R ∪{n R}; / / will n R Add to S R

[0212] end if

[0213] If |SR|>5 then

[0214] m n =Mean(Diff(SR )); / / Update m n δ n

[0215] δ n =Round(0.4×m n ×f s );

[0216] end if

[0217] end for

[0218] if S R [1]-C R [1]>δ n

[0219] N R ={c R |c R R [1],c R ∈C R}; / / Define a new candidate set N R

[0220] for i=|N R |-1 downto 1 do

[0221] if S R [1]-N R [i]>δ n then

[0222] S R ={N R [i]}∪S R ; / / N R [i] Add to S R

[0223] else

[0224] Continue;

[0225] end if

[0226] end for

[0227] end if

[0228] Return to S R

[0229] Algorithm 2

[0230] High-precision physiological parameter estimation effect

[0231] ​This invention can be applied to typical radar signal waveforms, including but not limited to ultra-wideband signals, frequency-modulated continuous wave signals, and continuous wave signals. Here, an ultra-wideband signal module is used for effect verification. The experiment used a base station with ultra-wideband signal transceiver capabilities, an electrocardiogram (ECG) module as a reference signal, and a static target. The target's posture included lying down and sitting, and the distances included 0.6m and 1m. More than 2400 sets of 1-minute physiological data were collected from a total of 10 subjects. The data are shown in Table 1.

[0232] Table 1

[0233]

[0234] To demonstrate the high precision and robustness of this invention in health monitoring, two widely used physiological signal reconstruction methods, Variational Mode Decomposition (VMD) and Discrete Wavelet Transform (DWT), are introduced for comparison. The performance of the three methods on the dataset is statistically analyzed, with the estimation accuracy and error of physiological vital signs parameters (respiratory rate (RR) and heart rate (HR)) as the evaluation metrics.

[0235] Tables 2 and 3 show the performance of the three methods in estimating physiological parameters in the collected dataset. Table 2 summarizes the accuracy and error of the respiratory rate estimated by the three methods in four scenarios. It can be seen that the accuracy of the respiratory rate estimation of the present invention is no less than 98.299% and the error is no more than 0.258 bpm in the four scenarios. Compared with the VMD and DWT methods, the accuracy is improved by at least 5.173% and 8.71%, respectively, while the error is reduced by at least 0.719 bpm and 1.221 bpm. Table 3 summarizes the accuracy and error of the heart rate estimation of the three methods in four scenarios. It can be seen that the accuracy of the heart rate estimation of the present invention is no less than 94.097% and the error is no more than 4.297 bpm in the four scenarios. Compared with the VMD and DWT methods, the accuracy is improved by at least 4.310% and 6.887%, respectively, while the error is reduced by at least 3.03 bpm and 4.657 bpm, respectively. In a 1m lying scenario, the accuracy of the present invention in heart rate estimation is as high as 96.664%, which is far higher than the accuracy of VMD (84.613%) and DWT (89.777%), demonstrating the superior performance of the present invention in the accuracy of respiratory rate and heart rate estimation.

[0236] Table 2

[0237]

[0238] Table 3

[0239]

[0240] Respiratory Harmonic Suppression Effect: To demonstrate the advantages of the physiological signal reconstruction algorithm based on local projection denoising and wavelet packet decomposition in harmonic suppression as described in the above embodiments, the physiological signal reconstruction algorithm based on local projection denoising and wavelet packet decomposition is compared with the aforementioned physiological signal reconstruction algorithm based solely on wavelet packet decomposition. The harmonic suppression effect of the present invention is shown below in the frequency domain, time domain, and statistical domain.

[0241] Figure 4 The distribution of respiratory signals reconstructed in the frequency domain by the physiological signal reconstruction algorithms of local projection denoising and wavelet packet decomposition proposed in this application is shown. It can be seen that because the respiratory signal has a high intensity and is easy to distinguish, the respiratory fundamental frequency component can be reconstructed well by using Algorithm 1 or WPD method. Figure 5 The distribution of the heartbeat signal reconstructed by the physiological signal reconstruction algorithm based on local projection denoising and wavelet packet decomposition in the frequency domain is presented. It can be seen that the frequency domain-based physiological signal reconstruction method (WPD) cannot solve the influence of respiratory harmonic peaks in the [0.8-1.2] Hz frequency band. The intensity of respiratory harmonics is even higher than the fundamental frequency component of the heartbeat (around 1.4 Hz). However, the heartbeat signal reconstructed by Algorithm 1 effectively suppresses respiratory harmonics, significantly reduces the intensity of respiratory harmonics, and highlights the fundamental frequency of the heartbeat. At the same time, the WPD algorithm successfully filters out the weak noise interference after the fundamental frequency of the heartbeat (around 1.4 Hz).

[0242] Figures 6-7 The effect of the present invention on respiratory harmonic suppression in the time domain is given, and it can be seen that the present invention can achieve reliable reconstruction of physiological signals. Figure 6 and Figure 7 The reconstructed respiratory and heart rate signals in the time domain are presented respectively. It can be seen that the reconstructed respiratory signal is quite ideal, with its peak value closely matching the ECG reference signal, achieving effective reconstruction of the respiratory signal. Figure 7 As can be seen, Algorithm 1, by effectively suppressing respiratory harmonics, successfully recovered the heartbeat peak at positions (37s, 43s) that the traditional frequency domain method (WPD) might miss, resulting in a more accurate heartbeat signal reconstruction compared to the WPD method.

[0243] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment.

[0244] Reference Figure 16This application illustrates a radar health monitoring device based on physiological signal reconstruction, according to an embodiment of the present application. The device monitors human health by periodically detecting physiological signals from the human body using radar. These physiological signals include an initial respiratory signal, an initial heartbeat signal, and an additive noise component. Specifically, the device includes the following modules:

[0245] The respiratory signal reconstruction module 210 is used to receive physiological signals corresponding to the target human body, perform delay domain processing on the physiological signals, remove the initial heartbeat signal and additive noise components, and generate a reconstructed respiratory signal.

[0246] The heartbeat prediction signal module 220 is used to determine the residual signal based on the physiological signal and the reconstructed respiratory signal, and to perform delay domain processing on the residual signal to remove the residual respiratory signal and the additive noise component, thereby generating the heartbeat prediction signal.

[0247] The reconstructed heartbeat signal module 230 is used to perform frequency band decomposition and filtering on the estimated heartbeat signal according to a preset frequency band to generate a reconstructed heartbeat signal.

[0248] The result generation module 240 is used to generate health monitoring results based on the reconstructed respiratory signal and the reconstructed heartbeat signal.

[0249] In one embodiment of the present invention, the respiratory signal reconstruction module 210 includes:

[0250] A radar phase information submodule is used to generate radar phase information based on the radar signal;

[0251] A chest wall change signal submodule is used to determine chest wall change signals based on the radar phase information.

[0252] The physiological signal submodule is used to determine physiological signals based on the chest wall change signals.

[0253] In one embodiment of the present invention, the respiratory signal reconstruction module 210 includes:

[0254] The high-dimensional phase space submodule is used to construct the high-dimensional phase space of the physiological signal according to preset parameters, including embedding dimension, sampling delay, projection dimension and neighborhood radius;

[0255] The respiratory signal submodule is used to acquire the low-dimensional space corresponding to the respiratory signal, project the high-dimensional phase space onto the low-dimensional space, and generate a reconstructed respiratory signal.

[0256] In one embodiment of the present invention, the heartbeat prediction signal module 220 includes:

[0257] The residual signal submodule is used to generate a residual signal based on the physiological signal and the reconstructed respiratory signal;

[0258] The heartbeat signal submodule is used to perform local projection noise reduction processing on the residual signal to obtain the estimated heartbeat signal.

[0259] In one embodiment of the present invention, the reconstructed heartbeat signal module 230 includes:

[0260] The wavelet packet decomposition submodule is used to divide the estimated heartbeat signal into several equal bandwidth frequency bands according to a preset wavelet packet decomposition layer.

[0261] The target frequency band node submodule is used to determine the target frequency band node based on the heartbeat base frequency band;

[0262] The energy threshold submodule is used to determine the energy threshold based on the total energy of the target node.

[0263] The summation submodule is used to determine the summation signal based on the energy threshold, and to sum the summation signal to obtain the reconstructed heartbeat signal.

[0264] In one embodiment of the present invention, the result generation module 240 includes:

[0265] The candidate set submodule is used to extract the maximum points of the reconstructed heartbeat signal and generate an R-peak candidate set;

[0266] The filtering submodule is used to filter and complete the R-peaks in the candidate set of R-peaks according to preset filtering conditions;

[0267] The heart rate submodule is used to determine heart rate data based on the number of R peaks in a preset time period;

[0268] The respiratory rate submodule is used to extract the respiratory rate data of the reconstructed respiratory signal;

[0269] The monitoring results submodule is used to generate health monitoring results that include physiological parameter values ​​and trend analysis based on the heart rate data and respiratory rate data.

[0270] In one embodiment of the present invention, the filtering submodule includes:

[0271] The sampling point unit is used to filter the candidate set of R peaks and fill in the missed R peaks based on the average RR interval and the minimum RR interval sampling point; wherein, the initial value of the average RR interval is the reciprocal of the heartbeat fundamental frequency estimate, and the minimum RR interval sampling point is 0.6 multiplied by the average RR interval and then multiplied by the signal sampling frequency; when the number of R peak points is greater than 5, the average RR interval and the minimum RR interval sampling point are updated in real time.

[0272] Reference Figure 17The diagram illustrates a computer device for implementing a radar health monitoring method based on physiological signal reconstruction according to the present invention, which may specifically include the following:

[0273] The aforementioned computer device 1 is in the form of a general-purpose computing device. The components of the computer device 1 may include, but are not limited to: one or more processors or processing units 3, memory 8, and a bus 4 connecting different system components (including memory 8 and processing unit 3).

[0274] Bus 4 represents one or more of several bus architectures, including memory buses or memory controllers, peripheral buses, graphics acceleration ports, processors, or local buses using any of the various bus architectures. For example, these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Audio / Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0275] Computer device 1 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by computer device 1, including volatile and non-volatile media, removable and non-removable media.

[0276] Memory 8 may include computer system readable media in the form of volatile memory, such as random access memory 9 and / or cache memory 10. Computer device 1 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 11 may be used to read and write non-removable, non-volatile magnetic media (commonly referred to as a "hard disk drive"). Although Figure 17 As not shown, a disk drive for reading and writing to a removable non-volatile disk (such as a "floppy disk") and an optical disk drive for reading and writing to a removable non-volatile optical disk (such as a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 4 via one or more data media interfaces. The memory may include at least one program product having a set (e.g., at least one) of program modules 13 configured to perform the functions of the embodiments of this application.

[0277] A program / utility 12 having a set (at least one) of program modules 13 may be stored, for example, in memory. Such program modules 13 include—but are not limited to—an operating system, one or more application programs, other program modules 13, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 13 typically perform the functions and / or methods described in the embodiments of this application.

[0278] Computer device 1 can also communicate with one or more external devices 2 (e.g., keyboard, pointing device, monitor 7, camera, etc.), and with one or more devices that enable an operator to interact with computer device 1, and / or with any device that enables computer device 1 to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed through I / O interface 6. Furthermore, computer device 1 can also communicate with one or more networks (e.g., local area network (LAN)), wide area network (WAN), and / or public networks (e.g., the Internet) through network adapter 5. Figure 17 As shown, network adapter 5 communicates with other modules of computer device 1 via bus 4. It should be understood that, although... Figure 17 Not shown, it can be combined with computer device 1 to use other hardware and / or software modules, including but not limited to: microcode, device drivers, redundant processing unit 3, external disk drive array, RAID system, tape drive and data backup storage system 11, etc.

[0279] The processing unit 3 executes various functional applications and data processing by running programs stored in memory 8, such as implementing a radar health monitoring method based on physiological signal reconstruction provided in the embodiments of this application.

[0280] That is, when the above-mentioned processing unit 3 executes the above-mentioned program, it realizes: receiving physiological signals corresponding to the target human body, performing delay domain processing on the physiological signals, removing the initial heartbeat signal and additive noise components, and generating a reconstructed respiratory signal;

[0281] The residual signal is determined based on the physiological signal and the reconstructed respiratory signal, and the residual signal is processed in the delay domain to remove the residual respiratory signal and the additive noise component, thereby generating the predicted heartbeat signal.

[0282] The estimated heartbeat signal is decomposed and filtered according to a preset frequency band to generate a reconstructed heartbeat signal;

[0283] Health monitoring results are generated based on the reconstructed respiratory signal and the reconstructed heartbeat signal.

[0284] In this application embodiment, the application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a radar health monitoring method based on physiological signal reconstruction as provided in all embodiments of the application.

[0285] That is, when the program is executed by the processor, it implements the following: receiving physiological signals corresponding to the target human body, performing delay domain processing on the physiological signals, removing the initial heartbeat signal and additive noise components, and generating a reconstructed respiratory signal;

[0286] The residual signal is determined based on the physiological signal and the reconstructed respiratory signal, and the residual signal is processed in the delay domain to remove the residual respiratory signal and the additive noise component, thereby generating the predicted heartbeat signal.

[0287] The estimated heartbeat signal is decomposed and filtered according to a preset frequency band to generate a reconstructed heartbeat signal;

[0288] Health monitoring results are generated based on the reconstructed respiratory signal and the reconstructed heartbeat signal.

[0289] Any combination of one or more computer-readable media may be used. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device.

[0290] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including—but not limited to—electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of transmitting, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0291] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof. These programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the operator's computer, partially on the operator's computer, as a standalone software package, partially on the operator's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the operator's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider). The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably.

[0292] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.

[0293] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0294] The above provides a detailed description of a radar health monitoring method based on physiological signal reconstruction provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and its core ideas. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for radar health monitoring based on physiological signal reconstruction, wherein the method monitors human health by periodically detecting physiological signals of the human body using radar, wherein... The physiological signals include an initial respiratory signal, an initial heartbeat signal, and an additive noise component; characterized in that they include: The system receives physiological signals corresponding to the target human body and performs delay domain processing on the physiological signals to remove the initial heartbeat signal and additive noise components, generating a reconstructed respiratory signal. A high-dimensional phase space of the physiological signal is constructed based on preset parameters, including embedding dimension, sampling delay, projection dimension, and neighborhood radius. A low-dimensional space corresponding to the respiratory signal is obtained, and the high-dimensional phase space is projected onto this low-dimensional space to generate the reconstructed respiratory signal. The residual signal is determined based on the physiological signal and the reconstructed respiratory signal, and the residual signal is processed in the delay domain to remove the residual respiratory signal and the additive noise component, thereby generating the estimated heartbeat signal; the residual signal is generated based on the physiological signal and the reconstructed respiratory signal; the residual signal is processed by local projection noise reduction to obtain the estimated heartbeat signal; The estimated heartbeat signal is decomposed and filtered according to a preset frequency band to generate a reconstructed heartbeat signal; the estimated heartbeat signal is divided into several equal-bandwidth frequency bands according to a preset wavelet packet decomposition level; target frequency band nodes are determined according to the heartbeat fundamental frequency band; an energy threshold is determined according to the total energy of the target frequency band nodes; valid signals are filtered according to the energy threshold, and a reconstructed heartbeat signal is generated from the valid signals. Health monitoring results are generated based on the reconstructed respiratory signal and the reconstructed heartbeat signal.

2. The method according to claim 1, characterized in that, The step of receiving physiological signals corresponding to the target human body includes: Radar phase information is generated based on radar signals; The chest wall change signal is determined based on the radar phase information; Physiological signals are determined based on the chest wall change signals.

3. The method according to claim 1, characterized in that, The step of generating health monitoring results based on the reconstructed respiratory signal and the reconstructed heartbeat signal includes: Extract the maximum points of the reconstructed heartbeat signal to generate a candidate set of R peaks; The R-peaks in the candidate R-peak set are filtered and supplemented according to preset filtering conditions; Heart rate data are determined based on the number of R peaks in a preset time period; Extract the respiratory rate data from the reconstructed respiratory signal; Based on the heart rate data and respiratory rate data, a health monitoring result containing physiological parameter values ​​and trend analysis is generated.

4. The method according to claim 3, characterized in that, The step of filtering and completing the R-peaks in the candidate R-peak set according to preset filtering conditions includes: Based on the average RR interval and the minimum RR interval sampling points, the candidate set of R peaks is screened and the missed R peaks are filled in; wherein, the initial value of the average RR interval is the reciprocal of the estimated heart rate fundamental frequency, and the minimum RR interval sampling point is 0.6 multiplied by the average RR interval and then multiplied by the signal sampling frequency; when the number of R peak points is greater than 5, the average RR interval and the minimum RR interval sampling points are updated in real time.

5. A radar health monitoring device based on physiological signal reconstruction, wherein the device monitors human health by periodically detecting physiological signals of the human body using radar, wherein... The physiological signals include an initial respiratory signal, an initial heartbeat signal, and an additive noise component; characterized in that they include: A respiratory signal reconstruction module is used to receive physiological signals corresponding to the target human body, perform delay domain processing on the physiological signals, remove the initial heartbeat signal and additive noise components, and generate a reconstructed respiratory signal; construct a high-dimensional phase space of the physiological signal according to preset parameters, including embedding dimension, sampling delay, projection dimension and neighborhood radius; obtain the low-dimensional space corresponding to the respiratory signal, and project the high-dimensional phase space onto the low-dimensional space to generate the reconstructed respiratory signal; The heartbeat prediction signal module is used to determine the residual signal based on the physiological signal and the reconstructed respiratory signal, and to perform delay domain processing on the residual signal to remove the residual respiratory signal and the additive noise component, thereby generating the heartbeat prediction signal; to generate the residual signal based on the physiological signal and the reconstructed respiratory signal; and to perform local projection noise reduction processing on the residual signal to obtain the heartbeat prediction signal. The reconstructed heartbeat signal module is used to perform frequency band decomposition and filtering on the estimated heartbeat signal according to a preset frequency band to generate a reconstructed heartbeat signal; divide the estimated heartbeat signal into several equal bandwidth frequency bands according to a preset wavelet packet decomposition level; determine target frequency band nodes according to the heartbeat fundamental frequency band; determine an energy threshold according to the total energy of the target frequency band nodes; filter valid signals according to the energy threshold, and generate a reconstructed heartbeat signal from the valid signals. The result generation module is used to generate health monitoring results based on the reconstructed respiratory signal and the reconstructed heartbeat signal.

6. A radar health monitoring device based on physiological signal reconstruction, characterized in that, The system includes a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program, when executed by the processor, implements the radar health monitoring method based on physiological signal reconstruction as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, which, when executed by a processor, implements the radar health monitoring method based on physiological signal reconstruction as described in any one of claims 1 to 4.

Images