A method and device for feature extraction of vital sign signals

Abstract

The application discloses a kind of feature extraction method and device of vital signs signal, comprising: the vital signs signal caused by thoracic movement is modeled;Continuous wave radar system echo signal mixing frequency acquisition beat frequency signal is modeled;The phase gradient of the beat frequency signal obtained is calculated, and second-order delay product is estimated;Second-order delay product is used to estimate second-order cyclic time-varying moment;According to second-order cyclic time-varying moment, enhanced spectrum is constructed, and the strongest breathing harmonic frequency in enhanced spectrum is extracted;The strongest breathing harmonic order is estimated, so as to extract target breathing frequencyAccording to breathing frequency, harmonic suppression enhanced spectrum is constructed, and target heartbeat frequency is extracted;Iterative optimization is estimated target breathing frequency and heartbeat frequency.The application can realize the accurate extraction of non-contact target breathing and heartbeat frequency under strong harmonic interference.

Description

Technical Field

[0001] This invention belongs to the field of signal processing technology, specifically relating to a method and apparatus for feature extraction of vital signs signals. Background Technology

[0002] With the improvement of people's living standards and the continuous enhancement of health awareness, vital sign monitoring technology has significant application value. In recent years, non-contact respiratory rate and heart rate monitoring technology based on millimeter-wave radar has gradually become a research focus in vital sign monitoring. However, the performance of non-contact monitoring based on millimeter-wave radar is easily affected by complex environments such as strong interference and low signal-to-noise ratio.

[0003] Typical vital sign monitoring algorithms include two main categories and their derivatives: arctangent demodulation and complex signal demodulation. Arctangent demodulation restricts the range of phase values, leading to discontinuities in the extracted phase signal. It is also sensitive to DC bias, requiring preprocessing to suppress DC. Complex signal demodulation is insensitive to DC bias, but its spectrum is modulated by the Bessel function. The amplitude of chest wall movement causes changes in the function's argument, resulting in a degree of unpredictability in the demodulated signal spectrum. Specifically, this manifests as a large number of high-order harmonics and cross-terms in the spectrum, affecting the accuracy of vital sign parameter estimation. Therefore, the development of vital sign monitoring technology under strong harmonic interference is of great value.

[0004] Currently, the advantages of cyclostationary signal processing in the field of vital sign monitoring are becoming increasingly apparent. It can enhance the characteristics of cyclostationary signals under strong noise environments, thereby obtaining more accurate signal parameter estimation results. Based on this, this invention addresses the problem of low accuracy in estimating vital sign characteristic parameters under complex environments by proposing a harmonic suppression phase gradient demodulation method based on cyclostationary characteristics for non-contact vital sign characteristic parameter estimation under strong noise and interference conditions. Summary of the Invention

[0005] Purpose of the invention: The present invention provides a method and apparatus for feature extraction of vital signs signals, which can accurately extract the respiratory rate and heart rate of a target in complex environments with low signal-to-noise ratio and strong harmonic interference.

[0006] Technical solution: The feature extraction method for vital sign signals described in this invention specifically includes the following steps:

[0007] (1) Model the chest movement caused by the pre-acquired target vital signs signal to obtain the distance between the target and the radar;

[0008] (2) The continuous wave radar transmits a single-frequency sinusoidal signal and simultaneously receives the radar echo signal. The echo is mixed to obtain the intermediate frequency signal b[t].

[0009] (3) Perform a second-order delay product on the intermediate frequency signal b[t] based on the phase gradient method. The estimate, its second-order delay product depends on time t and delay. Based on the second-order delay product, its second-order cyclic time-varying moment in the α-cyclic frequency domain is further estimated. Correspondingly, the spectral structure of the second-order cyclic time-varying moment in the cyclic frequency domain α depends on its time delay.

[0010] (4) Based on the second-order cyclic time-varying moment, select a smaller time delay range Δτ r ∈(0,τ r The corresponding cyclic spectra are multiplied together to construct the enhanced spectrum. Using the peak lookup method, the breathing harmonic frequencies corresponding to the strong spectral peaks are extracted from the cyclic frequency domain of the enhanced spectrum within the range of [0.1, 0.6] Hz.

[0011] (5) Estimate the harmonic order corresponding to the strongest spectral peak based on the spectral structure of the enhanced spectrum, and determine the target breathing frequency based on the harmonic order and its frequency. Extraction;

[0012] (6) Based on the estimated respiratory rate, select the corresponding time delay in the time delay domain of the second-order cyclic time-varying moment, and multiply the cyclic spectrum of the corresponding respiratory rate and its harmonics to construct the harmonic suppression enhancement spectrum. in Given a positive integer, the target heartbeat frequency is extracted using the peak lookup method within the cyclic frequency domain, selecting a frequency range of [0.9, 2.5] Hz.

[0013] (7) Based on the estimation of the target heart rate, the second-order cyclic time-varying moment is resampled in the time delay domain, and the target respiratory rate parameter is further optimized. The target heart rate is estimated by iterating 1 to 3 times.

[0014] Furthermore, the vital signs signals mentioned in step (1) are respiratory signals and heartbeat signals.

[0015] Furthermore, the implementation process of step (1) is as follows:

[0016] x[t] = R0 + Δx[t]

[0017]

[0018] Where x[t] is the distance between the target and the radar, R0 is the initial distance of the target, Δx[t] is the chest movement model corresponding to the life signal, and R kr and R hThese represent the k-th order respiratory harmonic and the amplitude of the heartbeat waveform, respectively, where k = 1, ..., K is the order of the respiratory harmonic, K is the order of the most likely harmonic, and f r and f h These are the respiratory base frequency and heart rate, kf r denoted by , cos(·) represents the frequency of the k-th order breathing harmonic, and cos(·) represents the cosine function.

[0019] Further, the intermediate frequency signal mentioned in step (2) is:

[0020]

[0021] Where, △x[t]=c / f c For continuous wave radar carrier frequency f c The corresponding wavelength, c is the speed of light, and θ0 is the initial phase;

[0022] Jacobi-Anger expansion of the intermediate frequency signal:

[0023]

[0024]

[0025] in, The harmonic order vector, For harmonic frequency vectors, and Representing integers and real numbers respectively. and Represented as a K-dimensional integer vector and a real vector respectively, with the superscript T indicating transpose; inner product The inner product represents the cross term between respiratory harmonics and heartbeat signals. Determine whether different intersection terms are real numbers; With J n (a h These represent the amplitudes of the respiratory harmonic component and the heartbeat component, respectively. With J n (·) represent m respectively k Bessel functions of the first kind of order n and nth order.

[0026] Furthermore, step (3) includes the following steps:

[0027] (3.1) For the obtained intermediate frequency signal, calculate its phase gradient and complete the second-order delay product. The estimation for time t and delay The second-order delay product of the intermediate frequency signal is:

[0028]

[0029] in, For the selected delay variable, the superscript (*) indicates the optional complex conjugate operation for each product factor. Here, the first product factor is not conjugated, and the second delayed signal is complex conjugate. sin(·) represents the sine function.

[0030] (3.2) The estimated second-order delay product Taking the odd-numbered cross terms, a Fourier transform is performed on the delay dimension to obtain the estimated value of the second-order cyclic time-varying moment:

[0031]

[0032] in, This represents the Fourier transform along time t. The imaginary part operation is used to ensure that the result contains only the intersection terms of odd-numbered respiratory harmonics and heart rate signals. The second-order cyclic time-varying moment contains only α-m. T f r -nf h The cyclic frequency component.

[0033] Furthermore, step (4) includes the following steps:

[0034] (4.1) Select the time delay range Δτ r ∈(0,τ r The corresponding second-order time-varying moments are multiplied together to construct the enhanced spectrum. The maximum value within the time delay range is:

[0035]

[0036] Among them, K max j′ is the minimum harmonic order to be suppressed. 1,1 R is the first zero of the derivative of the first-order Bessel function. krmax The maximum possible amplitude of the k-th harmonic is given by arcsin(·), where arcsin(·) is the arcsine function and min(·) is the function that takes the minimum value; enhancement spectrum

[0037]

[0038] Where ∏(·) represents the product factor of a multinomial function;

[0039] (4.2) Using the peak search method, in the enhanced spectrum In the cyclic frequency domain, the respiratory harmonic frequencies corresponding to the strong spectral peaks are extracted from the frequency range of [0.1, 0.6] Hz.

[0040]

[0041] Where, arg max α (·) indicates the value of variable α corresponding to the maximum value.

[0042] Furthermore, step (5) includes the following steps:

[0043] (5.1) Based on the respiratory harmonic frequency Estimate its maximum possible harmonic order K r :

[0044]

[0045] Among them, f rmin =0.1Hz is the minimum possible respiratory rate. For rounding;

[0046] (5.2) The estimated harmonic order corresponding to the strongest spectral peak is based on the enhanced spectral structure and the breathing harmonic frequency. Finish:

[0047]

[0048] (5.3) Extract the target respiratory rate

[0049]

[0050] Furthermore, step (6) includes the following steps:

[0051] (6.1) Based on the estimated respiratory rate The harmonic suppression enhancement spectrum is obtained by multiplying the second-order cyclic time-varying moments corresponding to the time delays of respiration and its harmonics. for:

[0052]

[0053] Among them, K h The preset maximum number of enhancement cycles, with typical values ​​of 2, 3, 4, and 5;

[0054] (6.2) Using the peak search method, in the harmonic suppression enhancement spectrum In the cyclic frequency domain, the target heart rate is extracted from the frequency range of [0.9, 2.5] Hz.

[0055]

[0056] Further, the target respiratory rate in step (7) is:

[0057]

[0058]

[0059] Where ε is the refinement time interval, and arg min α (·) indicates the value of variable α corresponding to the minimum value. Its value is proportional to the minimum time delay interval and is generally 1 to 3 times the minimum time delay interval. Based on this, the respiratory rate is estimated. The iterative step (6) completes the heart rate estimation. Repeat the iteration 1 to 3 times to obtain accurate target respiratory rate and heart rate parameters through iterative optimization.

[0060] Based on the same inventive concept, the present invention provides an apparatus comprising a memory and a processor, wherein:

[0061] Memory is used to store computer programs that can run on a processor;

[0062] A processor is configured to execute the feature extraction method steps for vital sign signals as described above when running the computer program.

[0063] Beneficial effects: Compared with the prior art, the beneficial effects of the present invention are as follows: The harmonic suppression phase gradient demodulation method based on the cyclic stability characteristics proposed in this invention enhances the local signal-to-interference-plus-noise ratio of respiratory and heartbeat signals by utilizing the cyclic stability characteristics of the signal, and can complete the non-contact extraction of target respiratory and heartbeat frequencies under strong harmonic interference and strong noise environments. Attached Figure Description

[0064] Figure 1 A flowchart of a method for feature extraction of vital signs signals;

[0065] Figure 2 The time-domain waveforms of vital signs signals under strong harmonic interference environment and their corresponding time-domain thoracic respiration and heartbeat signals;

[0066] Figure 3 The image shows the enhanced spectrum and respiration rate estimation results under strong harmonic interference environment;

[0067] Figure 4 The graph shows the harmonic suppression enhancement spectrum and heart rate estimation results under strong harmonic interference environment;

[0068] Figure 5 The time-domain waveforms of vital signs signals under strong noise and strong harmonic interference, and the corresponding chest respiration and heartbeat signals at the corresponding times;

[0069] Figure 6 The image shows the enhanced spectrum and respiration rate estimation results under strong noise and strong harmonic interference environments.

[0070] Figure 7The graph shows the harmonic suppression enhancement spectrum and heart rate estimation results under strong noise and strong harmonic interference environments. Detailed Implementation

[0071] The present invention will now be described in further detail with reference to the accompanying drawings:

[0072] like Figure 1 As shown, this invention proposes a feature extraction method for vital sign signals, used for vital sign monitoring. The vital signs in this invention are respiratory and heartbeat signals, and the monitoring focuses on the frequency of these signals. Specifically, the method includes the following steps:

[0073] Step 1: Model the chest movement caused by the target's life signals to obtain the distance between the target and the radar:

[0074] x[t]=R0+△x[t],

[0075] Where R0 is the initial distance to the target, and △x[t] is the chest movement model corresponding to the vital signal.

[0076] The model of thoracic cavity movement caused by vital signs is as follows:

[0077]

[0078] Among them, R kr and R h These represent the k-th order respiratory harmonic and the amplitude of the heartbeat waveform, respectively (k = 1, ..., K are the order of the respiratory harmonic, and K is the order of the most likely harmonic). r and f h These are the respiratory base frequency and heart rate, kf r denoted by , cos(·) represents the frequency of the k-th order breathing harmonic, and cos(·) represents the cosine function.

[0079] The maximum harmonic order is set to K=3, the respiratory rate to 0.51Hz (30.6 breaths / min), the heart rate to 1.65Hz (99 heartbeats / min), the fundamental respiratory amplitude to 0.008m, the second harmonic amplitude to 0.012m, the third harmonic amplitude to 0.008m, and the heartbeat waveform amplitude to 0.00035m, which is 0.4 times the respiratory amplitude. The respiratory and heartbeat signals are shown below. Figure 2 The respiratory displacement signal and heart rate displacement signal are shown.

[0080] Step 2: The continuous wave radar transmits a single-frequency sine wave signal and simultaneously receives the radar echo signal. The echo is mixed to obtain the intermediate frequency signal b[t].

[0081] The intermediate frequency signal model is:

[0082]

[0083] Where, △x[t]=c / f c For continuous wave radar carrier frequency f c The corresponding wavelengths are given, where c is the speed of light, θ0 is the initial phase, and n[t] is additive white Gaussian noise. In this embodiment, the white noise is generated according to a signal-to-noise ratio (SNR) of -5dB, and the carrier frequency is set to f. c =24GHz. According to f s A sampling rate of 100Hz yields N = 4000 sampling points for the intermediate frequency (IF) signal. The time-domain image of the IF signal is shown below. Figure 2 As shown.

[0084] The Jacobi-Anger expansion of the noise-free portion of the intermediate frequency signal can be expressed as follows:

[0085]

[0086]

[0087] in, The harmonic order vector, For harmonic frequency vectors, and Representing integers and real numbers respectively. and Let these be represented as a K-dimensional integer vector and a real vector, respectively, with the superscript T indicating transpose. Inner product. The inner product represents the cross term between respiratory harmonics and heartbeat signals. Determine whether different intersection terms are real numbers. With J n (a h These represent the amplitudes of the respiratory harmonic component and the heartbeat component, respectively. With J n (·) represent m respectively k Bessel functions of the first kind of order n and nth order.

[0088] Step 3: Perform the second-order delay product based on the phase gradient method. The estimation, for the intermediate frequency signal b[t], shows that its second-order delay product depends on time t and time delay. Based on the second-order delay product estimation, its second-order cyclic time-varying moment in the α-cyclic frequency domain is further estimated. Correspondingly, the spectral structure of the second-order cyclic time-varying moment in the cyclic frequency domain α depends on its time delay. The specific steps are as follows:

[0089] (3.1) For the obtained intermediate frequency signal, calculate its phase gradient and complete the second-order delay product. The estimation for time t and delay The second-order delay product of the intermediate frequency signal is,

[0090]

[0091] in, The selected delay variable is indicated by the superscript (*), which represents the complex conjugate operation that can be performed for each product factor. Here, the first product factor is not conjugated, and the second delayed signal is conjugated. sin(·) represents the sine function.

[0092] (3.2) The estimated second-order delay product Taking the odd-numbered cross terms, a Fourier transform is performed on the delay dimension to obtain the estimated value of the second-order cyclic time-varying moment.

[0093]

[0094] in, This represents the Fourier transform along time t. The imaginary part operation is used to ensure that the result contains only the intersection terms of odd-numbered respiratory harmonics and heart rate signals. The second-order cyclic time-varying moment contains only α-m. T f r -nf h The cyclic frequency component. In this embodiment, the Fourier transform is replaced by an 8000-point FFT.

[0095] Step 4: Based on the second-order cyclic time-varying moment, select a smaller time delay range Δτ. r ∈(0,τ r The corresponding cyclic spectra are multiplied together to construct the enhanced spectrum. Using the peak lookup method, the breathing harmonic frequencies corresponding to the strong spectral peaks are extracted from the cyclic frequency domain of the enhanced spectrum within the range of [0.1, 0.6] Hz. The specific steps are as follows:

[0096] (4.1) Select the time delay range Δτ r ∈(0,τ r The corresponding second-order time-varying moments are multiplied together to construct the enhanced spectrum. The maximum value within the aforementioned delay range is determined by the following formula:

[0097]

[0098] Among them, K max Let j1′,1 be the minimum harmonic order to be suppressed, j1′,1 be the first zero of the derivative of the first-order Bessel function, and R be the minimum harmonic order to be suppressed. krmax Let be the maximum possible amplitude of the k-th harmonic, arcsin(·) be the arcsine function, and min(·) be the function for finding the minimum value.

[0099] Construct Enhancement Spectrum The cumulative product of the second-order time-varying moment moduli within the time delay range enhances the spectrum. As shown below:

[0100]

[0101] Where ∏(·) represents the product factor of a multinomial function. In this embodiment, the cyclic frequency domain image of the enhanced spectrum is as follows: Figure 3 As shown, note that the stronger peaks do not appear at the breathing fundamental frequency, which is also the main problem to be solved in subsequent harmonic order estimation.

[0102] (4.2) Using the peak search method, in the enhanced spectrum In the cyclic frequency domain, the respiratory harmonic frequencies corresponding to the strong spectral peaks are extracted from the frequency range of [0.1, 0.6] Hz.

[0103]

[0104] Where, arg max α (·) indicates the value of variable α corresponding to the maximum value. The estimated respiratory harmonic frequency corresponding to the strong spectral peak is... It is a relatively strong second harmonic of breathing.

[0105] Step 5: Estimate the harmonic order corresponding to the strongest spectral peak based on the spectral structure of the enhanced spectrum, and determine the target breathing frequency based on the harmonic order and its frequency. The extraction process includes the following steps:

[0106] (5.1) Based on the respiratory harmonic frequency Estimate its maximum possible harmonic order K r As shown below:

[0107]

[0108] Among them, f rmin =0.1Hz is the minimum possible respiratory rate. This is for rounding up.

[0109] (5.2) The estimated harmonic order corresponding to the strongest spectral peak is based on the enhanced spectral structure and the breathing harmonic frequency. Completed, as shown below:

[0110]

[0111] The harmonic order estimation result here is

[0112] (5.3) Extract the target respiratory rate As shown below:

[0113]

[0114] like Figure 3 As shown, the target respiratory rate estimate is It matches the set value.

[0115] Step 6: Based on the estimated respiratory rate, select the corresponding time delay in the time delay domain of the second-order cyclic time-varying moment, and multiply the cyclic spectra of the corresponding respiration and its harmonics to construct the harmonic suppression enhancement spectrum. in Given a positive integer, the target heartbeat frequency is extracted using the peak lookup method within the cyclic frequency domain, selecting a frequency range of [0.9, 2.5] Hz. Specifically, the following steps are included:

[0116] (6.1) Based on the estimated respiratory rate The harmonic suppression enhancement spectrum is obtained by multiplying the second-order cyclic time-varying moments corresponding to the time delays of respiration and its harmonics. As shown below,

[0117]

[0118] Among them, K h This is the preset maximum number of enhancement cycles, typically 2, 3, 4, or 5. In this embodiment, the maximum number of enhancement cycles is set to K. h =5, other values ​​such as 2, 3, 4 can all obtain the same accuracy estimation results. Higher cycle numbers have stronger estimation robustness, but higher computational complexity, while lower values ​​can achieve rapid heart rate estimation.

[0119] (6.2) Using the peak search method, in the harmonic suppression enhancement spectrum In the cyclic frequency domain, the target heart rate is extracted from the frequency range of [0.9, 2.5] Hz.

[0120]

[0121] Constructing a harmonic suppression enhancement spectrum as follows Figure 4 As shown, there is only a single distinct peak in the spectrum, which corresponds to the heart rate peak. The estimated heart rate is [value missing]. It matches the set value.

[0122] Step 7: Based on the estimated target heart rate, resample the second-order cyclic time-varying moment in the time delay domain to further optimize the estimated target respiratory rate parameters. Iterate 1-3 times to estimate the target heart rate.

[0123]

[0124]

[0125] Where ε is the refinement time interval, and arg min α (·) indicates the value of variable α corresponding to the minimum value. Its value is proportional to the minimum time delay interval and is generally taken as 1 to 3 times the minimum time delay interval.

[0126] Based on this, the respiratory rate is estimated, and the heart rate is estimated in iterative step (6). This iteration is repeated 1 to 3 times to optimize and obtain accurate target respiratory and heart rate parameters. In this embodiment, a total of 3 iterations are performed to obtain a higher accuracy estimate, enhancement spectrum, and harmonic suppression enhancement spectrum, such as... Figure 3 , Figure 4 As shown, this invention can obtain relatively accurate estimates of respiratory rate and heart rate even under low signal-to-noise ratio and strong respiratory harmonic interference, and can be used to monitor human respiratory rate and heart rate in real-world complex interference scenarios.

[0127] Based on the same inventive concept, the present invention also provides an apparatus comprising a memory and a processor, wherein: the memory is used to store a computer program capable of running on the processor; and the processor is used to execute the feature extraction method steps of the vital signs signal as described above when running the computer program.

[0128] To further test the noise immunity of this invention under strong interference scenarios, the noise power was increased, and the signal-to-noise ratio (SNR) was adjusted to -10 dB. The fundamental respiratory waveform amplitude was set to 0.008 m, the second harmonic respiratory waveform amplitude to 0.004 m, the third harmonic respiratory waveform amplitude to 0.0024 m, and the heartbeat waveform amplitude to 0.00035 m. The respiratory rate and heart rate settings remained unchanged, and their corresponding intermediate frequency signals and respiratory-heartbeat displacement signals are as follows: Figure 5 As shown, the obtained enhancement spectrum and respiratory rate estimate are as follows: Figure 6 As shown, the harmonic suppression enhancement spectrum and heart rate estimate are as follows: Figure 7 As shown. By Figure 6 and Figure 7 It can be seen that when estimating respiratory rate, the enhanced spectrum shows several strong respiratory harmonic peaks, while when estimating heart rate, the harmonic suppression enhanced spectrum shows a single heart rate peak. This method can effectively achieve accurate respiratory rate estimation and effectively suppress the influence of respiratory harmonics when estimating heart rate, thus obtaining accurate heart rate estimates.

[0129] On the other hand, it is noted that the selected respiratory and heart rates are both relatively high values, demonstrating that the present invention can accurately estimate respiratory and heart rates over a high range under conditions of strong noise and strong respiratory harmonic interference. Therefore, the above embodiments demonstrate that the method can achieve excellent monitoring performance of vital signs signals under conditions of strong noise, strong respiratory harmonic interference, and extreme heart and respiratory rates.

Claims

1. A method for feature extraction of vital signs signals, characterized in that, Includes the following steps: (1) Model the chest movement caused by the pre-acquired target vital signs signal to obtain the distance between the target and the radar; (2) The continuous wave radar transmits a single-frequency sinusoidal signal and simultaneously receives the radar echo signal. The echo is mixed to obtain the intermediate frequency signal b[t]. (3) Perform the second-order delay product for the intermediate frequency signal b[t] based on the phase gradient method. The estimation of its second-order delay product depends on time t and delay. Based on the second-order delay product, its second-order cyclic time-varying moment in the α-cyclic frequency domain is further estimated. Correspondingly, the spectral structure of the second-order cyclic time-varying moment in the cyclic frequency domain α depends on its time delay. (4) Based on the second-order cyclic time-varying moment, select a smaller time delay range Δτ r ∈(0,τ r The corresponding cyclic spectra are multiplied together to construct the enhanced spectrum. Using the peak lookup method, the breathing harmonic frequencies corresponding to the strong spectral peaks are extracted from the cyclic frequency domain of the enhanced spectrum within the range of [0.1, 0.6] Hz. (5) Estimate the harmonic order corresponding to the strongest spectral peak based on the spectral structure of the enhanced spectrum, and determine the target breathing frequency based on the harmonic order and its frequency. Extraction; (6) Based on the estimated respiratory rate, select the corresponding time delay in the time delay domain of the second-order cyclic time-varying moment, and multiply the cyclic spectrum of the corresponding respiratory rate and its harmonics to construct the harmonic suppression enhancement spectrum. in Given a positive integer, the target heartbeat frequency is extracted using the peak lookup method within the cyclic frequency domain, selecting a frequency range of [0.9, 2.5] Hz. (7) Based on the estimation of the target heart rate, the second-order cyclic time-varying moment is resampled in the time delay domain, and the target respiratory rate parameter is further optimized. The target heart rate is estimated by iterating 1 to 3 times.

2. The method for feature extraction of vital sign signals according to claim 1, characterized in that, The vital signs signals mentioned in step (1) are respiratory signals and heartbeat signals.

3. The method for feature extraction of vital sign signals according to claim 1, characterized in that, The implementation process of step (1) is as follows: x[t] = R0 + Δx[t] Where x[t] is the distance between the target and the radar, R0 is the initial distance of the target, Δx[t] is the chest movement model corresponding to the life signal, and R kr and R h These represent the k-th order respiratory harmonic and the amplitude of the heartbeat waveform, respectively, where k = 1, ..., K is the order of the respiratory harmonic, K is the order of the most likely harmonic, and f r and f h These are the respiratory base frequency and heart rate, kf r denoted by , cos(·) represents the frequency of the k-th order breathing harmonic, and cos(·) represents the cosine function.

4. The method for feature extraction of vital sign signals according to claim 1, characterized in that, The intermediate frequency signal mentioned in step (2) is: Where, △x[t]=c / f c For continuous wave radar carrier frequency f c The corresponding wavelength, c is the speed of light, and θ0 is the initial phase; Jacobi-Anger expansion of the intermediate frequency signal: in, The harmonic order vector, For harmonic frequency vectors, and Representing integers and real numbers respectively. and Represented as a K-dimensional integer vector and a real vector respectively, with the superscript T indicating transpose; inner product The inner product represents the cross term between respiratory harmonics and heartbeat signals. Determine whether different intersection terms are real numbers; With J n (a h These represent the amplitudes of the respiratory harmonic component and the heartbeat component, respectively. With J n (·) represent m respectively k Bessel functions of the first kind of order n and nth order.

5. The method for feature extraction of vital sign signals according to claim 1, characterized in that, Step (3) includes the following steps: (3.1) For the obtained intermediate frequency signal, calculate its phase gradient and complete the second-order delay product. The estimation for time t and delay The second-order delay product of the intermediate frequency signal is: in, For the selected delay variable, the superscript (*) indicates the optional complex conjugate operation for each product factor. Here, the first product factor is not conjugated, and the second delayed signal is complex conjugate. sin(·) represents the sine function. (3.2) The estimated second-order delay product Taking the odd-numbered cross terms, a Fourier transform is performed on the delay dimension to obtain the estimated value of the second-order cyclic time-varying moment: in, This represents the Fourier transform along time t. The imaginary part operation is used to ensure that the result contains only the intersection terms of odd-numbered respiratory harmonics and heart rate signals. The second-order cyclic time-varying moment contains only α-m. T f r -nf h The cyclic frequency component.

6. The method for feature extraction of vital sign signals according to claim 1, characterized in that, Step (4) includes the following steps: (4.1) Select the time delay range Δτ r ∈(0,τ r The corresponding second-order time-varying moments are multiplied together to construct the enhanced spectrum. The maximum value within the time delay range is: Among them, K max j1′ is the minimum harmonic order to be suppressed. ,1 R is the first zero of the derivative of the first-order Bessel function. krmax Let arcsin(·) be the maximum possible amplitude of the k-th harmonic, and min(·) be the arcsine function; enhancement spectrum. Where ∏(·) represents the product factor of a multinomial function; (4.2) Using the peak search method, in the enhanced spectrum In the cyclic frequency domain, the respiratory harmonic frequencies corresponding to the strong spectral peaks are extracted from the frequency range of [0.1, 0.6] Hz. Where, arg max α (·) indicates the value of variable α corresponding to the maximum value.

7. The method for feature extraction of vital sign signals according to claim 1, characterized in that, Step (5) includes the following steps: (5.1) Based on the respiratory harmonic frequency Estimate its maximum possible harmonic order K r : Among them, f rmin =0.1Hz is the minimum possible respiratory rate. For rounding; (5.2) The estimated harmonic order corresponding to the strongest spectral peak is based on the enhanced spectral structure and the breathing harmonic frequency. Finish: (5.3) Extract the target respiratory rate 8. The method for feature extraction of vital sign signals according to claim 1, characterized in that, Step (6) includes the following steps: (6.1) Based on the estimated respiratory rate The harmonic suppression enhancement spectrum is obtained by multiplying the second-order cyclic time-varying moments corresponding to the time delays of respiration and its harmonics. for: Among them, K h This is the preset maximum number of enhancement cycles, with typical values ​​of 2, 3, 4, and 5. (6.2) Using the peak search method, in the harmonic suppression enhancement spectrum In the cyclic frequency domain, the target heart rate is extracted from the frequency range of [0.9, 2.5] Hz.

9. The method for feature extraction of vital sign signals according to claim 1, characterized in that, The target respiratory rate in step (7) is: Where ε is the refinement time interval, and arg min α (·) indicates the value of variable α corresponding to the minimum value. Its value is proportional to the minimum time delay interval, and it is generally taken as 1 to 3 times the minimum time delay interval. Based on this, the respiratory rate is estimated. The iterative step (6) completes the heart rate estimation. Repeat the iteration 1 to 3 times to obtain accurate target respiratory rate and heart rate parameters through iterative optimization.

10. A device, characterized in that, Includes memory and processor, wherein: Memory is used to store computer programs that can run on a processor; A processor, configured to, while running the computer program, perform the feature extraction method steps of the vital signs signal as described in any one of claims 1-9.

Images