A method for identifying weak human features using life detection radar

Abstract

The present invention discloses a method for identifying weak human features of a life detection radar, the method comprising acquiring a radar echo signal, wherein the radar echo signal comprises a heartbeat signal and a breathing signal; decomposing the radar echo signal to obtain M IMF components; calculating the energy of each IMF component respectively, and analyzing to obtain a first echo energy signal and a second echo energy signal; performing PCA denoising processing on the first echo energy signal and the second echo energy signal respectively to obtain a first denoised echo signal and a second denoised echo signal; performing autocorrelation processing on the first denoised echo signal and the second denoised echo signal respectively to extract a heartbeat signal and a breathing signal. The present invention can separate a heartbeat signal from a body motion signal, and the heart rate value extracted from the signal has a strong correlation with the heart rate value extracted from an electrocardiogram signal.

Description

Technical Field

[0001] The invention belongs to the technical field of radar signal processing, and in particular relates to a method for identifying weak human features by a life detection radar. Background Art

[0002] Bio-radar uses electromagnetic waves as a medium, emits electromagnetic pulses, reflects them from the human body, receives echo signals, processes and analyzes them, and obtains physiological information of living organisms. Bio-radar overcomes the problems of laser and infrared detection being seriously affected by temperature and failing when blocked by objects, and also overcomes the problems of ultrasonic detection having large attenuation in spatial propagation, being affected by reflection interference from environmental debris, and failing when blocked by water, ice, and soil. This technology can be widely used in military medicine, disaster medicine, urban counter-terrorism and other fields, and is recognized by the international scientific and technological community as a very important frontier technology field.

[0003] Compared with traditional contact monitoring, non-contact monitoring methods can reduce the discomfort caused by electrodes, wires, and sensors to the monitored target, overcome the monitoring of patients who are difficult to use traditional contact monitoring methods, and avoid the impact of contact devices that may fall off and cause monitoring problems. It is precisely because of the above characteristics of bio-radar that bio-radar can be used to achieve non-contact detection of human vital signs at a certain distance and through a medium of a certain thickness. Life characteristic signals are composed of multiple parameters such as breathing, heartbeat, blood pressure, and blood oxygen concentration, and these signals are usually extremely weak and easily submerged in a strong clutter background. Respiration and heartbeat signals, breathing will cause chest bullying movement, thereby generating positive and negative Doppler frequency domains, which are modulated with the high-frequency signal emitted by the radar and received by the receiver. The receiver can then perform filtering operations to extract the breathing signal. The same is true for the heartbeat signal. Because of the multi-point reflection of the chest cavity and the uneven breathing frequency, many harmonics with different frequencies distributed near the zero frequency will be generated. Data collection and analysis show that life signals are low signal-to-noise ratio, extremely low frequency, quasi-periodic, and multi-harmonic signals. The clutter signal can be approximated as Gaussian colored noise, so the signal processing of bio-radar can be simplified to extract signals from harmonic models interfered by Gaussian colored noise, suppress clutter, and improve the signal-to-noise ratio. Bio-radar can detect human life signals (breathing, body movement, etc.) without contact, and is widely used in homes, community hospitals, and other occasions. Existing bio-radar non-contact detection uses radar detection through blood pressure and radar detection through respiratory signals. Among them, the radar detection of blood pressure includes: using radar to collect the aortic pulse waveform signal of the subject; extracting feature points of the aortic pulse waveform signal to obtain each main feature point; determining the mapping relationship between the time period of the aortic pulse and each feature point; solving the pulse wave conduction time from the two feature points of the first contraction peak of the aortic pulse and the ejection phase stop point, i.e., the dicrotic wave trough, according to the mapping relationship; solving the corresponding blood pressure value according to the conduction time, and then realizing the detection of human life signals according to the detected blood pressure value; It is very easy to detect the breathing signal by using radar, and the breathing signal can be extracted by sampling a simple digital filter with a cutoff frequency of 0.8Hz, and then realizing the detection of human life signals according to the detected breathing signal.

[0004] However, in practical applications, it is also necessary to detect the human heartbeat signal in order to better reflect the human life status. However, directly using a bandpass filter (the lower cutoff frequency is 0.8Hz and the upper cutoff frequency is 3Hz) to detect the heartbeat signal will also detect the high-order harmonics of the breathing signal and the environmental Gaussian noise, making it difficult to extract the heartbeat signal. Summary of the invention

[0005] In order to solve the above problems existing in the prior art, the present invention provides a method for identifying weak human features by a life detection radar.

[0006] An embodiment of the present invention provides a method for identifying weak human features using a life detection radar, comprising:

[0007] Acquire a radar echo signal, wherein the radar echo signal includes a heartbeat signal and a breathing signal;

[0008] Decomposing the radar echo signal to obtain M IMF components;

[0009] Calculate the energy of each IMF component respectively, and analyze to obtain the first echo energy signal and the second echo energy signal;

[0010] Performing PCA noise reduction processing on the first echo energy signal and the second echo energy signal to obtain a first noise-reduced echo signal and a second noise-reduced echo signal respectively;

[0011] Autocorrelation processing is performed on the first denoised echo signal and the second denoised echo signal respectively to extract a heartbeat signal and a breathing signal.

[0012] In one embodiment of the present invention, the acquired radar echo signal is expressed as:

[0013]

[0014] in, β0=2kr0, λ is the emission wavelength, d is the thickness of the wall, ε is the dielectric constant, ω1=2πf1, ω2=2πf2, f1 and f2 represent the frequencies of breathing and heartbeat respectively, Δ1 and Δ2 represent the amplitudes of breathing and heartbeat respectively, and φ2 is the constant phase.

[0015] In one embodiment of the present invention, decomposing the radar echo signal to obtain M IMF components includes:

[0016] The radar echo signal is decomposed by using the CEEMDAN algorithm to obtain the M IMF components.

[0017] In one embodiment of the present invention, respectively calculating the energy of each IMF component and analyzing to obtain the first echo energy signal and the second echo energy signal comprises:

[0018] Calculate the energy of each IMF component to obtain an IMF energy vector;

[0019] Normalizing the IMF energy vector to obtain a normalized weight vector;

[0020] The first echo energy signal and the second echo energy signal are determined from the M IMF components according to the normalized weight vector.

[0021] In one embodiment of the present invention, performing PCA denoising processing on the first echo energy signal and the second echo energy signal to obtain a first denoised echo signal and a second denoised echo signal respectively further includes:

[0022] adding noise to the first echo energy signal and the second echo energy signal respectively to obtain a first noisy echo energy signal and a second noisy echo energy signal;

[0023] The first noisy echo energy signal and the second noisy echo energy signal are respectively subjected to PCA denoising processing to obtain the first denoised echo signal and the second denoised echo signal.

[0024] In one embodiment of the present invention, performing autocorrelation processing on the first noise reduction echo signal and the second noise reduction echo signal to extract the heartbeat signal and the breathing signal respectively includes:

[0025] Multiple autocorrelation processing is performed on the first denoised echo signal and the second denoised echo signal respectively to extract a heartbeat signal and a breathing signal.

[0026] In one embodiment of the present invention, the heartbeat signal frequency range is 0.8 Hz to 3.0 Hz.

[0027] In one embodiment of the present invention, the frequency range of the breathing signal is 0.05 Hz to 0.8 Hz.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] The life detection radar weak human feature recognition method provided by the present invention can separate the heartbeat signal from the body motion signal, and the heart rate value extracted from the signal has a strong correlation with the heart rate value extracted from the electrocardiogram signal.

[0030] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a flow chart of a method for identifying weak human features by a life detection radar provided by an embodiment of the present invention;

[0032] Figure 2 Schematic diagram of PCA principal component analysis provided by an embodiment of the present invention;

[0033] Figure 3 (a)~ Figure 3 (b) is a schematic diagram of a radar echo signal containing noise and a radar echo signal after denoising provided by an embodiment of the present invention;

[0034] Figure 4is a schematic diagram of a time domain analysis result of a radar echo signal provided by an embodiment of the present invention;

[0035] Figure 5 It is a schematic diagram of an FFT image of a radar echo signal under different IMF components provided by an embodiment of the present invention;

[0036] Figure 6 (a)~ Figure 6 (b) is a schematic diagram of simulation results of a respiratory signal and a heartbeat signal at a fixed respiratory frequency and heartbeat frequency provided by an embodiment of the present invention;

[0037] Figure 7 (a)~ Figure 7 (b) is a schematic diagram of simulation results of a respiratory signal and a heartbeat signal under a fixed respiratory frequency and a variable heartbeat frequency provided by an embodiment of the present invention;

[0038] Figure 8 It is a structural schematic diagram of a device for identifying weak human features of a life detection radar provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0039] The present invention is further described in detail below with reference to specific embodiments, but the embodiments of the present invention are not limited thereto.

[0040] Embodiment 1

[0041] See also Figure 1 , Figure 1 This is a flow chart of a method for identifying weak human features using a life detection radar provided by an embodiment of the present invention. This embodiment provides a method for identifying weak human features using a life detection radar, and the method includes the following steps:

[0042] S1. Acquire a radar echo signal, wherein the radar echo signal includes a heartbeat signal and a breathing signal.

[0043] Specifically, in this embodiment, it is assumed that the radar transmits a single-frequency continuous sine wave, and its complex signal is expressed as:

[0044]

[0045] Wherein, U0 represents the complex signal amplitude, and ω0 represents the complex signal phase. The distance between the target and the radar in this embodiment can be expressed as:

[0046] r(t)=r0+Δr(t) (2)

[0047] Among them, r0 represents the distance between the life target and the radar (including the thickness of the wall), and Δr(t) represents the distance change caused by the target's breathing and heartbeat. The echo signal received by the radar is expressed as:

[0048]

[0049] Wherein, μ represents the attenuation factor (including the two-way attenuation in free space and in the process of penetrating the wall), A represents the amplitude modulation of the echo by the target, β0 and 2kΔr(t) represent the phase shift, β0 = 2kr0, λ is the emission wavelength.

[0050] Since the propagation speed of electromagnetic waves in different media is inversely proportional to the square root of the dielectric constant of the medium, additional time delay will occur when electromagnetic waves penetrate the wall, which will cause additional phase difference. Assuming the thickness of the wall is d and the dielectric constant is ε, formula (3) can be re-expressed as:

[0051]

[0052] in, Usually breathing and heartbeat can be regarded as simple harmonic vibrations, and the fluctuations in the distance between the radar and the life target caused by them are expressed as:

[0053] Δr(t)=Δ1sin(ω1t)+Δ2sin(ω2t+φ2) (5)

[0054] Wherein, ω1=2πf1, ω2=2πf2, f1 and f2 represent the frequencies of breathing and heartbeat respectively, Δ1 and Δ2 represent the amplitudes of breathing and heartbeat respectively, and φ2 is a constant phase. The Doppler signal is reflected in the phase of the echo. This embodiment uses the orthogonal phase detection method to remove the carrier frequency, and the zero intermediate frequency signal is expressed as:

[0055]

[0056] Formula (4) is normalized by formula (6), and the normalized signal is expressed as:

[0057] S s (t)=exp{-jφ0-j2k[Δ1 sin(ω1t)+Δ2sin(ω2t+φ2)]} (7)

[0058] Since the phase contains a sinusoidal expression, in order to analyze the relationship between the fundamental sinusoidal wave and each harmonic, this embodiment uses the Bessel function to express it as:

[0059]

[0060] Then, formula (8) is used to expand formula (7) to express it as:

[0061]

[0062] Wherein, m1 and m2 are integers. It can be seen from formula (9) that the spectrum of the radar echo signal of this embodiment includes three parts:

[0063] (1) Constant component when m1=m2=0;

[0064] (2) fundamental angular frequencies ω1 and ω2 of breathing and heartbeat;

[0065] (3) The combination of respiratory and heartbeat angular frequency harmonics m1ω1+m2ω2.

[0066] S2. Decompose the radar echo signal to obtain M IMF components.

[0067] Specifically, the embodiment decomposes the radar echo signal to obtain M intrinsic mode functions (IMF) components, including: using the complete integrated empirical mode decomposition algorithm with adaptive noise (Complete EEMD with Adaptive Noise, CEEMDAN for short) to decompose the radar echo signal to obtain M IMF components. Specifically:

[0068] Since both the breathing signal and the heartbeat signal exist in the echo signal, and since the small signal (heartbeat signal) can be easily filtered out as the noise of the large signal (breathing signal) when the signal is denoised, we should first separate the breathing and heartbeat signals. According to the difference in the frequency of the breathing signal and the heartbeat signal, this embodiment uses the CEEMDAN algorithm to decompose the radar echo signal obtained by S1 into multiple IMF components according to the signal peak value and its corresponding frequency point, so as to facilitate the separation of the two effective signals of the breathing signal and the heartbeat signal. Specifically, this embodiment adopts the method of adding a finite number of adaptive white noises. In the CEEMDAN algorithm, the modal IMF components obtained by decomposition are used. Indicates. Operator E j (·) represents the jth modal IMF component sequence of a given signal obtained by the CEEMDAN algorithm, ω i is the Gaussian white noise of N(0,1) added for the i-th time, S s (t) is the radar echo signal sequence obtained by S1, which is expressed as S after adding white noise s (t)=ε0ω i (t), the CEEMDAN algorithm is used to decompose the radar echo signal sequence after adding white noise. The process is as follows:

[0069] First, the CEEMDAN algorithm is used to decompose the first modal component Then calculate the first residual signal The signal R1(t)+ε1E1ω is analyzed by CEEMDAN algorithm.i (t) Repeat the decomposition N times to obtain the second modal component For k = 2, 3, ..., M, calculate the kth residual Repeatedly use the radar echo signal sequence decomposed by the CEEMDAN algorithm to calculate the k+1th modal component expressed as:

[0070]

[0071] The decomposition is completed after multiple operations until the residual is no longer suitable for decomposition. The final residual is expressed as:

[0072]

[0073] Finally, the radar echo signal of this embodiment is decomposed into M modal components IMF expressed as:

[0074]

[0075] S3. Calculate the energy of each IMF component respectively, and analyze to obtain a first echo energy signal and a second echo energy signal.

[0076] Specifically, the target signal of this embodiment obtains M IMF components through the CEEMDAN algorithm, and each IMF component spectrum has a corresponding center frequency, which represents the target frequency or noise frequency. Since the target information is unknown, it is difficult to extract the target signal characteristics from the perspective of signal spectrum analysis. The energy spectrum of each IMF component of the signal can better show the energy changes of different target IMF components. Therefore, this embodiment proposes a feature extraction method based on the energy ratio of IMF components, and specifically S3 includes S3.1, S3.2, and S3.3:

[0077] S3.1. Calculate the energy of each IMF component to obtain an IMF energy vector.

[0078] Specifically, this embodiment calculates the energy E of each IMF component i It is expressed as:

[0079]

[0080] Thus, the IMF energy vector V corresponding to the M IMF components is obtained E =[E1,E2,…,E M ].

[0081] S3.2. Normalize the IMF energy vector to obtain a normalized weight vector.

[0082] Specifically, in this embodiment, the IMF energy vector V EAfter normalization, the normalized weight vector is expressed as:

[0083] V′ E =[p1,p2,…,p M ] (14)

[0084] in, represents the normalized weight of each IMF component, represents the sum of the energies of M IMF components.

[0085] S3.3. Determine a first echo energy signal and a second echo energy signal from the M IMF components according to the normalized weight vector.

[0086] Specifically, the calculation method according to formula (14) in this embodiment can make the IMF component with a high energy proportion obtain a larger weight, while the IMF component with a small proportion can only obtain a smaller weight, which is more conducive to highlighting the main components of the signal and reducing the bandwidth impact of IMF energy analysis. The energy difference characteristics of the high and low frequency bands of the signal analyze the differences in the vibration mechanisms of different acoustic targets, so that the energy distribution characteristics of each IMF component of the target radiation noise have obvious differences. However, due to the existence of environmental noise or measurement system noise, the measurement test signal of the target radiation noise will be affected to a certain extent. At this time, only the IMF energy vector V E As a target feature parameter, it cannot complete target recognition. Therefore, according to the characteristics of the target, the high-frequency and low-frequency energy difference of the target signal is defined as the signal feature parameter, and the frequency band energy difference characteristics of the target signal and the IMF energy vector V E Analysis is combined to achieve target identification and classification.

[0087] Assume that the radar echo signal is decomposed into M IMF components, and the i-th IMF component has L sampling points, where the instantaneous frequency of the p-th sampling point is f i,p , the instantaneous amplitude is A i,p , then the instantaneous intensity at point p is According to the frequency distribution characteristics of typical radar echo signals, the frequency value of 50-800Hz is defined as the low frequency band, and 800-3000Hz is defined as the high frequency band. The instantaneous intensity of the time-frequency sampling point in the low frequency band is denoted as B l1 ,B l2 ,…,B ln , n is the number of time-frequency sampling points in the low-frequency band, then the total low-frequency energy of the radar echo signal P l It is expressed as:

[0088]

[0089] Similarly, the instantaneous intensity of the time-frequency sampling point in the high frequency band is recorded as B l1 ,B l2 ,…,Blm , m is the number of time-frequency sampling points in the high frequency band, then the total high-frequency energy of the radar echo signal P h It is expressed as:

[0090]

[0091] Then, the high-frequency and low-frequency energy difference of the radar echo signal is defined as:

[0092] ΔP=P h -P l (17)

[0093] After the above energy feature analysis and denoising, two signals can be obtained, namely the first echo energy signal f1(t) signal of 0.05Hz to 0.8Hz and the second echo energy signal f2(t) signal of 0.8Hz to 3Hz. The main components in the signals at both ends correspond to the breathing signal and the heartbeat signal respectively.

[0094] S4. Perform PCA noise reduction processing on the first echo energy signal and the second echo energy signal to obtain a first noise-reduced echo signal and a second noise-reduced echo signal respectively.

[0095] Specifically, since PCA principal component analysis screens the principal components in the signal, when the signal-to-noise ratio is low (noise drowns the useful signal), PCA is likely to filter out the useful signal as noise, and the noise signal is amplified again because it contains a large amount of energy, making the signal-to-noise ratio worse. Therefore, in this embodiment, before PCA performs noise reduction processing on the first echo energy signal and the second echo energy signal, Gaussian white noise is added to obtain the first noisy echo energy signal and the second noisy echo energy signal, and the signal-to-noise ratio is appropriately improved, and then PCA is used to perform PCA noise reduction processing on the first noisy echo energy signal and the second noisy echo energy signal to obtain the first noise-reduced echo signal and the second noise-reduced echo signal, specifically:

[0096] PCA is essentially a lossy feature compression process that retains the most original information. To achieve this goal, the data points after dimensionality reduction (projection) are as dispersed as possible. Figure 2 , Figure 2 This is a schematic diagram of PCA principal component analysis provided by an embodiment of the present invention. Compared with the long arrow, if the projection is performed on the short arrow, there will be more overlapping points, which means more information loss. Therefore, this embodiment selects the long arrow direction for projection. The variance is used to represent this degree of dispersion: Let the feature after PCA dimensionality reduction be A, is the variance, a i represents the value in feature A, m is the total number of samples, and because the feature will be zero-meaned before PCA dimension reduction, it is recorded as: Similarly, in order to reduce the redundant information of features, the features after PCA dimensionality reduction should be uncorrelated with each other, and the covariance is used to measure the correlation. Specifically, the two features after PCA dimensionality reduction are A and B, then is 0.

[0097] The first noise reduction echo signal and the second noise reduction echo signal are expressed as Among them, f 1 / 2 (t) The subscript 1 / 2 indicates that it includes two signals, 1 is the breathing signal and 2 is the heartbeat signal. 1 / 2 (t) Construct the covariance matrix and multiply it by the coefficient Define the covariance matrix R xx It is expressed as:

[0098]

[0099] It can be seen from formula (18) that the diagonal elements of formula (18) are the variances of each feature, and the elements at other positions are the covariances between the features. Next, the eigenvectors and eigenvalues ​​of the covariance matrix are solved. There are two traditional implementation methods: eigenvalue decomposition and SVD algorithm decomposition. This embodiment uses the eigenvalue decomposition method. Assume that the covariance matrix R xx The eigenvalue of is λ, and the eigenvector is x, then:

[0100] R xx x=λx (19)

[0101] Further deduction can be obtained:

[0102] (λE-R xx )x=0 (20) The eigenvalue λ and eigenvector x are obtained from formula (20), then:

[0103] R xx =x T Σx (21)

[0104] Among them, R xxRepresents the covariance matrix, x represents the eigenvector of the covariance matrix, and Σ represents the matrix composed of eigenvalues. At this time, the eigenvalues ​​are sorted from large to small in the matrix Σ, and the eigenvectors corresponding to the eigenvalues ​​are also transformed accordingly. Each element in the characteristic matrix Σ finally obtained is the proportion of each IMF component in the principal component after the radar echo signal is decomposed. By setting a fixed threshold or using an adaptive threshold determination method, several components with a large proportion are selected as the principal component. The row of eigenvectors corresponding to the eigenvalue λ and the first noisy echo energy signal matrix are used to calculate the components of each first noisy echo energy signal component after PCA denoising. Finally, the selected components are added to obtain the first denoised echo signal after PCA denoising, which is expressed as:

[0105] y 1 / 2 (t) = xX T (twenty two)

[0106] Similarly, the second noisy echo energy signal can be obtained by the above processing to obtain the second noise-reduced echo signal.

[0107] S5. Perform autocorrelation processing on the first noise reduction echo signal and the second noise reduction echo signal respectively to extract a heartbeat signal and a breathing signal.

[0108] Specifically, after PCA principal component analysis, the heartbeat signal and the breathing signal are basically screened out. However, the noise is still very large. The traditional autocorrelation detection method is to perform autocorrelation operation on the input signal and the input signal after delay τ, and use the uncorrelated characteristics between signal and noise, and between noise and noise to achieve the purpose of improving the signal-to-noise ratio. For example, formula (22) can be expressed as:

[0109]

[0110] The autocorrelation function corresponding to formula (23) is expressed as:

[0111] R Y (τ)=E[y 1 / 2 (t)·y 1 / 2 (t+τ)]=Rs(τ)+E[s(t)·n(t+τ)]+E[s(t+τ)·n(t)]+R N (τ) (24)

[0112] The time autocorrelation function of the sample function is used to replace the autocorrelation function of the random process. At this time, the autocorrelation function in formula (24) is re-expressed as:

[0113]

[0114] In actual measurement, considering that the actual observation time T is limited, formula (25) is re-expressed as:

[0115]

[0116] Then the autocorrelation function of the signal in this embodiment is expressed as:

[0117]

[0118] If the noise is standard Gaussian white noise, then E[n(t)] and E[n(t+τ)] are both 0, and thus E[s(t)n(t+τ)] and E[s(t+τ)n(t)] are also 0. However, in actual measurement, due to the limited observation time and the degree of noise whitening may not be ideal, E[n(t)] and E[n(t+τ)] are not necessarily zero. Therefore, according to formula (26), the correlation function between the signal and the noise is expressed as:

[0119]

[0120]

[0121] Although in theory, all values ​​of Gaussian white noise are 0 except τ = 0, in actual measurement, the noise cannot reach the value envisioned in theory. N (τ)(τ≠0) always exists and is a function of τ. However, its amplitude is significantly reduced compared to the original noise and can be regarded as a new noise. N (0) is a relatively large number. It can be ignored in actual measurement or simulation and replaced by 0. Then formula (24) can be re-expressed as:

[0122]

[0123] Changing the integration time from T-τ to T, formula (30) is simplified to:

[0124]

[0125] Then the formula (24) of this embodiment is re-expressed as:

[0126]

[0127] in, is the superposition of Rs(τ) and E[s(t+τ)·n(t)]; n′(t) is the superposition of E[s(t)·n(t+τ)] and R N The superposition of (τ).

[0128] Comparing formula (23) and (32), although the amplitude and phase of the two signals are different, the frequency does not change. It increases the signal-to-noise ratio through correlation operation, but the degree of improvement is limited, thus limiting the ability to detect weak signals. This embodiment selects multiple autocorrelations and converts y′ 1 / 2 (t) as y 1 / 2 (t), repeat the above steps many times. The more times of autocorrelation, the more the signal-to-noise ratio is improved, so weak signals submerged in noise can be detected. The normalized autocorrelation function value of random noise is the largest at zero, and the other points decay to zero immediately; while the normalized autocorrelation function value of general signal is the largest at zero, but the other points do not decay to zero immediately, but go through a slow decline process. Since the randomness of the values ​​of random noise at each time point is weakly correlated, the values ​​of general signal at each time point have a certain correlation, which is strong.

[0129] In this embodiment, the first denoised echo signal and the second denoised echo signal are respectively subjected to the above-mentioned multiple autocorrelation processing to extract the heartbeat signal and the breathing signal. Although the single correlation operation can improve the signal-to-noise ratio of the source signal, the ability to detect and extract weak signals is limited due to the limited degree of improvement in the signal-to-noise ratio. In this embodiment, multiple autocorrelations are used to greatly improve the signal-to-noise ratio by repeatedly performing the steps of the single correlation operation, so that the noise reduction effect is obvious, and it is more conducive to analyzing and extracting the heartbeat signal and the breathing signal from the first denoised echo signal and the second denoised echo signal respectively.

[0130] It should be noted that the autocorrelation signal must be a real signal. During the calculation, the real and imaginary parts of the complex signal are subjected to multiple autocorrelation processing respectively. After noise reduction, the correlation between the real and imaginary parts is used, and the cross-correlation is used for noise reduction and then added.

[0131] In order to verify the effectiveness of the method for identifying weak human features by life detection radar proposed in this embodiment, the following simulation experiment is conducted to further prove it.

[0132] First, the breathing and heartbeat signals are analyzed and the radar echo signal is modeled on the MATLAB platform. A 2GHz radar is selected to transmit a single-frequency continuous sine wave, which is modulated with the breathing and heartbeat signals, and Gaussian noise is added to obtain the final radar echo signal containing noise. Since the amplitude of the breathing signal is more than ten times that of the heartbeat signal, the signal-to-noise ratio is set to 10dB, and the signal-to-noise ratio of the heartbeat signal can reach between -10dB and -15dB. Next, we start to extract the heartbeat signal from the noisy radar echo signal: first, use the CEEMDAN modal decomposition algorithm to decompose the radar echo signal into several IMF components with different frequency components, calculate the IMF components corresponding to 0.05Hz~0.8Hz, 0.8Hz~3.0Hz, and the energy of other IMF components, and filter out the corresponding breathing and heartbeat signals through adaptive thresholds. At this time, the signal-to-noise ratio of the heartbeat signal is improved to a certain extent, but it is still weaker than the noise. Further use PCA principal component analysis to re-screen the selected effective heartbeat signals to further increase the signal-to-noise ratio. Finally, the final heartbeat signal is obtained through multiple autocorrelations.

[0133] See Figure Figure 3 (a)~ Figure 3 (b) Figure 4 , Figure 5 , Figure 3 (a)~ Figure 3 (b) is a schematic diagram of a radar echo signal containing noise and a radar echo signal after denoising provided by an embodiment of the present invention, Figure 4 is a schematic diagram of the time domain analysis result of the radar echo signal provided by an embodiment of the present invention, Figure 5 It is a schematic diagram of an FFT image of a radar echo signal provided by an embodiment of the present invention. Figure 3 (a) is the radar echo signal with noise in the frequency domain. Figure 3 (b) is the radar echo signal after passing through a 0-4 Hz low-pass digital filter. ω1=0.23438 is the breathing signal, and the almost submerged ω2=0.99609 is the heartbeat signal. Both signals have been added with 10 dB of noise. Figure 4 right Figure 3 (b) Time domain analysis of radar echo signals after passing through a low-pass digital filter; Figure 5 This is the FFT image of the radar echo signal after energy calculation and analysis. It can be seen that in the low-frequency part (IMF5~IMF7), the IMF component has almost only one prominent peak, while in the higher-frequency part, there are still several frequency points with relatively large energy. In view of this situation, this embodiment uses the energy characteristic analysis method to achieve noise reduction.

[0134] See also Figure 6 (a)~ Figure 6 (b) Figure 7 (a)~ Figure 7 (b) Figure 6 (a)~ Figure 6 (b) is a schematic diagram of simulation results of a respiratory signal and a heartbeat signal under a fixed respiratory frequency and heartbeat frequency provided by an embodiment of the present invention, wherein: Figure 6 (a) is a schematic diagram of simulation results of a respiratory signal under fixed respiratory frequency and heart rate provided by an embodiment of the present invention, Figure 6 (b) is a schematic diagram of the simulation results of the heartbeat signal under fixed breathing frequency and heartbeat frequency provided by an embodiment of the present invention, Figure 7 (a)~ Figure 7 (b) is a schematic diagram of simulation results of a respiratory signal and a heartbeat signal under a fixed respiratory frequency and a variable heartbeat frequency provided by an embodiment of the present invention, wherein: Figure 7 (a) is a schematic diagram of simulation results of a respiratory signal under a fixed respiratory frequency and a variable heart rate provided by an embodiment of the present invention, Figure 7 (b) is a schematic diagram of the simulation results of the heartbeat signal under a fixed breathing frequency and a variable heartbeat frequency provided by an embodiment of the present invention. Figure 6 (a)~ Figure 6 (b) The breathing rate is fixed at 0.23 Hz and the heart rate is fixed at 1.50 Hz. Figure 7 (a)~ Figure 7 (b) The fixed breathing frequency is 0.23 Hz and the variable heart rate signal heart rate frequency is increased from 1 Hz to 2 Hz in steps of 0.01 Hz. Figure 6 (a)~ Figure 6 (b) Figure 7 (a)~ Figure 7 (b) It can be seen that after being processed by the weak human feature recognition method proposed in this embodiment, the noise is basically eliminated, and the measurement results are basically consistent with the set frequency, that is, the respectively extracted breathing frequency and heart rate measurement results are basically consistent with the set frequency. The method proposed in this embodiment can separate the heart rate signal from the body motion signal.

[0135] In summary, the life detection radar weak human feature recognition method proposed in this embodiment can separate the heartbeat signal from the body motion signal, and the heart rate value extracted from the signal has a strong correlation with the heart rate value extracted from the electrocardiogram signal.

[0136] Embodiment 2

[0137] Based on the above embodiment 1, see Figure 8 , Figure 8 1 is a schematic diagram of the structure of a life detection radar weak human feature recognition device provided by an embodiment of the present invention. This embodiment proposes a life detection radar weak human feature recognition device, and the life detection radar weak human feature recognition device includes:

[0138] The data acquisition module 801 is used to acquire radar echo signals, wherein the radar echo signals include heartbeat signals and breathing signals.

[0139] Specifically, the radar echo signal acquired in the data acquisition module 801 of this embodiment is expressed as:

[0140]

[0141] in, β0=2kr0, λ is the emission wavelength, d is the thickness of the wall, ε is the dielectric constant, ω1=2πf1, ω2=2πf2, f1 and f2 represent the frequencies of breathing and heartbeat respectively, Δ1 and Δ2 represent the amplitudes of breathing and heartbeat respectively, and φ2 is the constant phase.

[0142] The first data processing module 802 is used to decompose the radar echo signal to obtain M IMF components.

[0143] Specifically, in this embodiment, the first data processing module 802 decomposes the radar echo signal to obtain M IMF components including:

[0144] The radar echo signal is decomposed using the CEEMDAN algorithm to obtain M IMF components.

[0145] The data calculation and analysis module 803 is used to calculate the energy of each IMF component respectively, and analyze to obtain a first echo energy signal and a second echo energy signal.

[0146] Specifically, in the data calculation and analysis module 803 of this embodiment, the energy of each IMF component is calculated respectively, and the first echo energy signal and the second echo energy signal are obtained by analysis, including:

[0147] Calculate the energy of each IMF component to obtain an IMF energy vector;

[0148] Normalize the IMF energy vector to obtain a normalized weight vector;

[0149] A first echo energy signal and a second echo energy signal are determined from the M IMF components according to the normalized weight vector.

[0150] The second data processing module 804 is used to perform PCA noise reduction processing on the first echo energy signal and the second echo energy signal to obtain a first noise reduction echo signal and a second noise reduction echo signal respectively.

[0151] Specifically, in the second data processing module 804 of this embodiment, performing PCA noise reduction processing on the first echo energy signal and the second echo energy signal to obtain the first noise reduction echo signal and the second noise reduction echo signal respectively further includes:

[0152] Adding noise to the first echo energy signal and the second echo energy signal respectively to obtain a first noisy echo energy signal and a second noisy echo energy signal;

[0153] The first noisy echo energy signal and the second noisy echo energy signal are respectively subjected to PCA denoising processing to obtain a first denoised echo signal and a second denoised echo signal.

[0154] The data extraction module 805 is used to perform autocorrelation processing on the first noise reduction echo signal and the second noise reduction echo signal to extract a heartbeat signal and a breathing signal.

[0155] Specifically, in the data extraction module 805 of this embodiment, the first noise reduction echo signal and the second noise reduction echo signal are respectively subjected to autocorrelation processing to extract the heartbeat signal and the breathing signal, including:

[0156] The first denoised echo signal and the second denoised echo signal are respectively subjected to multiple autocorrelation processing to extract a heartbeat signal and a breathing signal.

[0157] Furthermore, the heartbeat signal frequency range is 0.8 Hz to 3.0 Hz.

[0158] Furthermore, the frequency range of the breathing signal is 0.05 Hz to 0.8 Hz.

[0159] The device for identifying weak human features by life detection radar proposed in this embodiment can execute the method embodiment for identifying weak human features by life detection radar described in the first embodiment above. Its implementation principle and technical effects are similar and will not be described in detail here.

[0160] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the protection scope of the present invention.

Claims

1. A method for identifying weak human features using a life detection radar, characterized in that: include: Acquire a radar echo signal, wherein the radar echo signal includes a heartbeat signal and a breathing signal; Decomposing the radar echo signal to obtain M IMF components; The energy of each IMF component is calculated respectively, and the first echo energy signal and the second echo energy signal are obtained by analysis; Performing PCA noise reduction processing on the first echo energy signal and the second echo energy signal to obtain a first noise-reduced echo signal and a second noise-reduced echo signal respectively; Performing multiple autocorrelation processing on the first denoised echo signal and the second denoised echo signal respectively to extract a heartbeat signal and a breathing signal; The energy of each IMF component is calculated respectively, and the first echo energy signal and the second echo energy signal are obtained by analysis, including: Calculate the energy of each IMF component to obtain an IMF energy vector; Normalizing the IMF energy vector to obtain a normalized weight vector; Determine the first echo energy signal and the second echo energy signal from the M IMF components according to the normalized weight vector; wherein the first echo energy signal and the second echo energy signal correspond to a breathing signal and a heartbeat signal, respectively; The method for identifying weak human features also includes: using the energy difference characteristics of the high and low frequency bands of the signal to analyze the differences in the vibration mechanisms of different acoustic targets. Specifically, the high and low frequency energy difference of the target signal is defined as the characteristic parameter of the target signal, and the target is identified and classified by combining the frequency band energy difference characteristics of the target signal with the IMF energy vector analysis.

2. The method for identifying weak human features using life detection radar according to claim 1, characterized in that: The acquired radar echo signal is expressed as: in, λ is the emission wavelength, d is the thickness of the wall, ε is the dielectric constant, ω1=2πf1, ω2=2πf2, f1 and f2 represent the frequencies of breathing and heartbeat respectively, Δ1 and Δ2 represent the amplitudes of breathing and heartbeat respectively, is a constant phase.

3. The method for identifying weak human features using life detection radar according to claim 1, characterized in that: Decomposing the radar echo signal to obtain M IMF components includes: The radar echo signal is decomposed by using the CEEMDAN algorithm to obtain the M IMF components.

4. The method for identifying weak human features using life detection radar according to claim 1, characterized in that: Performing PCA noise reduction processing on the first echo energy signal and the second echo energy signal to obtain a first noise-reduced echo signal and a second noise-reduced echo signal respectively further includes: adding noise to the first echo energy signal and the second echo energy signal respectively to obtain a first noisy echo energy signal and a second noisy echo energy signal; The first noisy echo energy signal and the second noisy echo energy signal are respectively subjected to PCA denoising processing to obtain the first denoised echo signal and the second denoised echo signal.

5. The method for identifying weak human features using life detection radar according to claim 1, characterized in that: The heartbeat signal frequency range is 0.8 Hz to 3.0 Hz.

6. The method for identifying weak human features using life detection radar according to claim 1, characterized in that: The frequency range of the breathing signal is 0.05 Hz to 0.8 Hz.

Images