Millimeter wave radar human body pulse wave signal extraction method based on signal decomposition and reconstruction

Through the millimeter wave radar technology of signal decomposition and reconstruction, the contactability and noise interference problems of traditional pulse wave detection methods are solved, and non-contact and accurate pulse wave signal extraction is achieved, which is suitable for self-monitoring and telemedicine for patients with chronic diseases.

CN118986296APending Publication Date: 2024-11-22CHONGQING UNIV OF POSTS & TELECOMM
View PDF 9 Cites 0 Cited by

Patent Information

Application Number
CN202410939922.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-12
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

Traditional pulse wave detection methods require contact with the skin, which may cause discomfort and are difficult to achieve accurate long-term monitoring in complex environments, especially for patients with skin sensitivity or burns, and violate privacy.

Method used

Using millimeter wave radar technology based on signal decomposition and reconstruction, the human neck phase signal is collected through frequency modulation continuous wave radar, combined with circular dynamic tracking algorithm to eliminate nonlinear distortion, and used extended differential and cross-multiplied algorithm to solve the phase fuzzy problem, and through The empirical modal decomposition algorithm of the time-varying filter removes noise and reconstructs the pulse wave signal.

Benefits of technology

It realizes non-contact, accurate pulse wave signal extraction, removes environmental noise, retains the detailed characteristics of pulse waves, helps predict heartbeat and blood pressure parameters, and is suitable for self-monitoring and telemedicine in patients with chronic diseases.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118986296A_ABST
    Figure CN118986296A_ABST
Patent Text Reader

Abstract

The invention discloses a millimeter wave radar human body pulse wave signal extraction method based on signal decomposition and reconstruction. The method comprises the following steps: firstly, acquiring human body data by deploying a millimeter wave radar, eliminating nonlinear distortion in a radar circuit by adopting a circular dynamic tracking algorithm, and obtaining a human body neck vibration phase signal, and secondly, solving the problems of phase ambiguity and phase change through an extended differential and cross multiplication algorithm; the phase signal is decomposed into a plurality of modes through an empirical mode decomposition (TVF-EMD) algorithm based on a time varying filter, and then the phase signal is decomposed into a plurality of modes through the empirical mode decomposition (TVF-EMD) algorithm based on the time varying filter; finally, according to a modal signal selection criterion, main waves, second harmonics and third harmonics of pulse waves are selected respectively, environmental noise is removed, and pulse wave signals are reconstructed. According to the method, noise removal of the pulse wave signals is achieved through the TVF-EMD algorithm, detail features of the pulse waves are reserved, and follow-up prediction of parameters such as heartbeat and blood pressure according to the pulse waves is facilitated.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to life signal detection technology, and specifically relates to a method for extracting human pulse wave signals using a millimeter wave radar based on signal decomposition and reconstruction. Background Art

[0002] In modern society, the detection of life signals such as pulse and heartbeat is extremely important. As basic physiological function indicators of the human body, the changes of pulse and heartbeat can provide valuable health information, which is crucial for disease prevention, medical diagnosis, monitoring and nursing, and safety protection. The pulse wave is the propagation of the pressure wave generated by the heart every time it beats in the artery. By detecting the pulse wave, we can understand the working state of the heart and the elasticity of the artery. Abnormal pulse wave morphology may indicate cardiovascular diseases such as hypertension and arteriosclerosis. Secondly, pulse wave velocity (PWV) is closely related to blood pressure. Arterial blood pressure can be estimated through pulse wave detection, which provides an important reference for the screening and management of diseases such as hypertension.

[0003] However, traditional pulse wave detection requires a sensor that contacts the skin, which may cause discomfort to some patients, especially those with sensitive skin, skin damage or burns, and is not suitable for long-term monitoring. In addition, camera equipment often requires bright light and good environmental conditions, which infringes on privacy and may be disturbed in complex backgrounds. Therefore, a non-contact, convenient and accurate method is needed to monitor pulse wave signals. In this regard, millimeter wave radar technology shows great potential and advantages. Millimeter wave radar uses electromagnetic waves in the millimeter wave frequency band for detection and imaging, and has the characteristics of high resolution, strong penetration and non-contact detection. Radar technology can achieve real-time pulse wave detection and is easy to integrate with telemedicine systems. This allows patients with chronic diseases to monitor themselves at home and transmit data to doctors through telemedicine systems, so as to adjust treatment plans in a timely manner.

[0004] The advantage of millimeter wave radar in measuring breathing and heartbeat signals is that it is non-contact and does not require direct contact with the human body, thus avoiding the discomfort or interference that may be caused by traditional methods. In addition, millimeter waves have a high resolution and can achieve high-precision imaging and detection of tiny movements of the human body. It has strong penetrating power and can pass through clothing and obstacles to achieve signal detection inside the human body. Therefore, as an advanced wireless sensing technology, millimeter wave radar has great potential and advantages in the field of breathing and heartbeat signal detection. It can provide more convenient, accurate and reliable solutions for health monitoring, medical diagnosis and safety protection. The present invention proposes a millimeter wave radar human pulse wave signal extraction method based on signal decomposition and reconstruction. First, by deploying millimeter-wave radar to collect human body data, the circular dynamic tracking algorithm is used to eliminate the nonlinear distortion in the radar circuit and obtain the human neck vibration phase signal. Secondly, the extended differential and cross-multiplication algorithms are used to solve the problems of phase ambiguity and phase variation. The phase signal is decomposed into multiple modes by the empirical mode decomposition (TVF-EMD) algorithm based on the time-varying filter. Finally, according to the modal signal selection criteria, the main wave, second harmonic and third harmonic of the pulse wave are selected respectively to remove the environmental noise and reconstruct the pulse wave signal. The patent of this invention realizes the noise removal of the pulse wave signal through the TVF-EMD algorithm, retains the detailed characteristics of the pulse wave, and is helpful for the subsequent prediction of heart rate, blood pressure and other parameters based on the pulse wave. Summary of the invention

[0005] The purpose of the present invention is to provide an efficient non-contact pulse wave signal extraction method, which obtains the human neck skin vibration phase signal through millimeter wave radar, and when there is environmental noise interference in the phase signal, the pulse wave signal is de-noised through the TVF-EMD algorithm. The method accurately restores the original form and detailed features of the pulse wave, which helps to extract relevant human cardiovascular information.

[0006] The invention discloses a method for extracting a millimeter wave radar human pulse wave signal based on signal decomposition and reconstruction, comprising the following steps:

[0007] Step 1: Use frequency modulated continuous wave (FMCW) radar to collect the phase signal of the human chest position. By performing FFT on each row of the echo signal matrix, we can obtain the distance information of the human neck under the current Chirp. Once the range bin of each Chirp is determined, we can extract the original phase signal of the neck. Then, the circular dynamic tracking algorithm is used to optimize the fitting results of the circle through the least squares method and gradient descent method to eliminate the direct current (DC) bias introduced by radar circuit defects and static object interference. Secondly, in order to prevent phase ambiguity and phase packaging problems, we applied the extended differential and cross-multiplication (DACM) algorithm.

[0008] Step 2: Use the time varying filter based empirical mode decomposition (TVF-EMD) to denoise the phase signal and eliminate the interference of environmental noise on the pulse wave. The TVF-EMD algorithm combines time varying filtering technology through repeated envelope interpolation, mean calculation and detail signal extraction to gradually decompose the signal into several intrinsic mode functions (IMF) and a residual signal. First, the local cutoff frequency of the signal X(t) is estimated using an analytical method. Then adjust use Construct B-spline nodes and obtain the local mean m(t) through B-spline estimation. Subtract the local mean from the original signal to obtain the residual signal r(t) = X(t) - m(t), and determine whether r(t) satisfies the condition that the IMF is a local narrowband. If so, the corresponding IMF is obtained; otherwise, repeat the above steps.

[0009] Step 3: Select the corresponding IMF to reconstruct the pulse wave signal according to the modal selection criteria. Calculate the percentage of the main wave of the heartbeat signal and the corresponding second harmonic and third harmonic in each IMF signal. After obtaining the correlation between each IMF and the corresponding harmonic, select the three IMF signals with the highest correlation with the main wave of the heartbeat and the corresponding second harmonic and third harmonic, and add the three signals to obtain the reconstructed pulse wave signal. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 It is a flow chart of the present invention;

[0011] Figure 2 It is the time domain diagram of the original phase signal;

[0012] Figure 3 This is the time domain diagram of the signal after TVF-EMD denoising.

[0013] Figure 4 This is the frequency domain diagram of the pulse wave signal after TVF-EMD.

[0014] Figure 5 This is the frequency domain diagram of the pulse wave signal after TVF-EMD. Specific implementation plan

[0015] The present invention will be further described below in conjunction with the accompanying drawings:

[0016] like Figure 1 The method for extracting human pulse wave signals from a millimeter wave radar based on signal decomposition and reconstruction specifically comprises the following steps:

[0017] Step 1: Use frequency modulated continuous wave (FMCW) radar to collect the phase signal of the human neck position. First, perform fast Fourier transform (FFT) on the intermediate frequency signal of each frame to obtain the position of the human body. Then extract the phase signal of the human body position in each frame of data, such as Figure 1 As shown. Before extracting the target phase information, it is necessary to eliminate any nonlinear distortion and interference such as false targets. Radar circuit defects and the influence of static object interference will introduce DC offset, which will distort the signal. A high-quality, distortion-free pulse wave waveform is a prerequisite for predicting blood pressure, so the DC component needs to be removed before the next step of processing. This paper uses a circular dynamic tracking algorithm to optimize the circle fitting results through the least squares method and gradient descent method.

[0018]

[0019] By finding the center and radius of the circle, the offset between the IQ signals can be eliminated.

[0020] There is a linear relationship between the skin displacement caused by the pulse and the phase change, so the pulse signal can be obtained by solving the signal phase. In order to prevent the problem of phase ambiguity and phase change, this paper uses the DACM algorithm to solve the phase of the signal. The DACM algorithm implements the signal inverse tangent differentiation and then integration in discrete form:

[0021]

[0022] The phase ambiguity problem is solved by the DCAM algorithm, and the method is simple in form, low in operation complexity, and can achieve high-speed calculation.

[0023] The discrete form of the phase signal can be represented by a continuous signal. However, it is a challenging task to directly use the phase signal to calculate blood pressure. The phase signal often loses the characteristic details of the original pulse wave signal due to the introduction of noise in the environment, and a complete and clear pulse wave is the prerequisite for accurate blood pressure prediction. Therefore, the signal needs to be further processed to restore the original pulse waveform from the phase signal covered by noise.

[0024] Step 2: Use the time varying filter based empirical mode decomposition (TVF-EMD) to denoise the phase signal and eliminate the interference of environmental noise on the pulse wave. The TVF-EMD algorithm not only has strong noise robustness, but also always maintains time-varying characteristics, and has unique advantages in processing non-stationary vibration signals. The algorithm mainly includes identifying the local cutoff frequency of the signal and using time-varying filtering. Among them, the purpose of separating the local cutoff frequency in the frequency domain is to establish a time-varying filter. The TVF-EMD algorithm includes the following 8 steps:

[0025] Step 1: Use Hilbert to calculate the instantaneous amplitude A(t) and instantaneous frequency of the signal X(t)

[0026]

[0027]

[0028] in, is the Hilbert transform of X(t)

[0029] Step 2: Determine the local maximum of A(t) max}) and the local minimum A({t min})

[0030] Step 3: A({t max}) and A({t min}) Perform local interpolation to obtain β1(t) and β2(t) respectively, and then calculate the instantaneous mean α1(t) and instantaneous envelope α2(t):

[0031]

[0032]

[0033] Step 4: Calculation and The interpolation results η1(t) and η2(t) are then used to calculate the instantaneous frequency component and

[0034]

[0035]

[0036] Step 5: Calculate the local cutoff frequency:

[0037]

[0038] Step 6: Given the previous step May be affected by intermittent noise, so local cutoff frequency needs to be adjusted and recalibrated

[0039] Step 7: Pass The reconstructed signal h(t) can be obtained:

[0040]

[0041] The B-spline interpolation is constructed using the extreme points of h(t), and the local mean function m(t) is obtained by time-varying filtering of X(t).

[0042] Step 8: Calculate the signal stopping criterion θ(t). When θ(t) is an instantaneous narrowband signal, the stopping criterion is met:

[0043]

[0044] Among them, B Loughlin (t) and are the Loughlin instantaneous bandwidth of the two-component signal and the weighted average of the instantaneous frequency of a single component. They are defined as:

[0045]

[0046]

[0047] When θ(t) is less than the threshold ζ, it is considered that X(t) can be regarded as an IMF and meets the termination condition, otherwise X(t) = X(t)-m(t) and continue to perform steps 1-7. Finally, the phase signal of the pulse wave is decomposed into multiple IMF components.

[0048] Step 4: Since the frequency of the pulse wave is consistent with the heart rate, the main wave frequency f1 of the pulse wave is distributed in the range of 0.8-2.0 Hz. The frequencies of the second harmonic and third harmonic of the pulse wave are 2f1 and 3f1 respectively. Therefore, firstly calculate the percentage of the three modal components in each IMF signal, and then get the correlation degree between each IMF signal and the pulse wave component signal.

[0049]

[0050]

[0051]

[0052] Among them, α j ,β j ,γ j Respectively represent the correlation between the jth IMF signal and the main pulse wave and its second harmonic and third harmonic. E(j) is the energy of the jth IMF signal. E1(j), E2(j), and E3(j) are the energies of the three pulse wave component frequency bands in the jth IMF signal. After obtaining the correlation between each IMF and the pulse wave component, select the three IMF signals with the highest correlation with the pulse wave component. Add the three IMF signals to reconstruct the denoised pulse wave signal. The final time domain and frequency domain signals of the pulse wave are shown in the figure. Figure 3 and Figure 5 shown.

Claims

1. A method for extracting human pulse wave signals from millimeter wave radar based on signal decomposition and reconstruction, characterized in that The following steps are involved: Step 1: Use the time varying filter based empirical mode decomposition (TVF-EMD) to denoise the phase signal and eliminate the interference of environmental noise on the pulse wave. The TVF-EMD algorithm not only has strong noise robustness, but also always maintains time-varying characteristics, and has unique advantages in processing non-stationary vibration signals. The algorithm mainly includes identifying the local cutoff frequency of the signal and using time-varying filtering. Among them, the purpose of separating the local cutoff frequency in the frequency domain is to establish a time-varying filter. The TVF-EMD algorithm includes the following 8 steps: Step 1: Use Hilbert to calculate the instantaneous amplitude A(t) and instantaneous frequency of the signal X(t) in, is the Hilbert transform of X(t) Step 2: Determine the local maximum of A(t) max }) and the local minimum A({t min }) Step 3: A({t max }) and A({t min }) Perform local interpolation to obtain β1(t) and β2(t) respectively, and then calculate the instantaneous mean α1(t) and instantaneous envelope α2(t): Step 4: Calculation and The interpolation results η1(t) and η2(t) are then used to calculate the instantaneous frequency component and Step 5: Calculate the local cutoff frequency: Step 6: Given the previous step May be affected by intermittent noise, so local cutoff frequency needs to be adjusted and recalibrated Step 7: Pass The reconstructed signal h(t) can be obtained: The B-spline interpolation is constructed using the extreme points of h(t), and the local mean function m(t) is obtained by time-varying filtering of X(t). Step 8: Calculate the signal stopping criterion θ(t). When θ(t) is an instantaneous narrowband signal, the stopping criterion is met: Among them, B Loughlin (t) and are the Loughlin instantaneous bandwidth of the two-component signal and the weighted average of the instantaneous frequency of a single component. They are defined as: When θ(t) is less than the threshold ζ, it is considered that X(t) can be regarded as an IMF and meets the termination condition, otherwise X(t) = X(t)-m(t) and continue to perform steps 1-7. Finally, the phase signal of the pulse wave is decomposed into multiple IMF components. Step 2: Since the frequency of the pulse wave is consistent with the heart rate, the main wave frequency f1 of the pulse wave is distributed in the range of 0.8-2.0 Hz. The frequencies of the second harmonic and third harmonic of the pulse wave are 2f1 and 3f1 respectively. Therefore, first calculate the percentage of the three modal components in each IMF signal, and then you can get the correlation between each IMF signal and the pulse wave component signal. Among them, α j ,β j ,γ j They represent the correlation between the jth IMF signal and the main pulse wave and its second harmonic and third harmonic respectively. E(j) is the energy of the jth IMF signal. E1(j), E2(j), and E3(j) are the energies of the three pulse wave component frequency bands in the jth IMF signal. After obtaining the correlation between each IMF and the pulse wave component, the three IMF signals with the highest correlation with the pulse wave component are selected. The three IMF signals are added together to reconstruct the denoised pulse wave signal.

Citation Information

Patent Citations

  • Iterative least square method-based MIMO (multiple input multiple output) radar DOA (direction-of-arrival) estimation method

    CN103744061A

  • Audio signal processing device, audio signal processing method, and recording medium

    CN107708780A

  • Anti-main lobe constant-modulus interference time domain cancellation method and system

    CN111505588A

  • Non-contact heartbeat detection method based on millimeter waves

    CN112754441A

  • In-vehicle life body millimeter wave radar detection method based on machine learning

    CN113640792A