Millimeter wave radar vital sign detection method and system based on ICEEMDAN combined with Hilbert spectrum analysis

Abstract

The invention discloses a millimeter wave radar vital sign detection method and system based on ICEEMDAN combined with Hilbert spectrum analysis, and the method comprises the steps: building a millimeter wave radar experiment system, placing a radar in front of a human body, collecting a signal, obtaining an IF signal, and carrying out the preprocessing of the IF signal; static clutters are suppressed by adopting a moving target display method, and a human body target position is limited through incoherent accumulation and a range gate. The phase of the vital sign signal is recovered by using the linear characteristic of arc tangent demodulation, and after the phase information of the target unit is extracted, the phase wrapping jump is eliminated through a phase unwrapping algorithm; after phase difference and moving average filtering are used for optimizing the phase signals, alternative component signals are obtained through an ICEEMDAN algorithm; and finally, performing spectrum analysis on the selected components through a Hilbert spectrum analysis algorithm to obtain an estimated frequency of the vital sign signal. The algorithm verification is carried out on the actually measured signal, and the result shows that the modal aliasing problem caused by the existing EMD algorithm in the signal modal decomposition process is reduced, and the high detection precision is obtained.

Description

Technical Field

[0001] This invention belongs to the field of biomedical signal processing and radar technology, and particularly relates to a method for detecting human vital signs based on frequency modulated continuous wave (FMCW) radar, involving non-contact detection technology for vital signs such as heart rate and respiratory rate. Background Technology

[0002] Respiration and heart rate, as core physiological parameters of the human body, are of great value for disease early warning, health management, and clinical diagnosis. Traditional contact-based detection relies on sensors or electrodes in direct contact with human skin, and this physical contact requirement limits the expansion of application scenarios. Furthermore, prolonged wear leads to decreased user cooperation, and monitoring is interrupted when the user is not wearing the device or when the battery is depleted. With the development of sensing and signal processing technologies, vital sign detection technology is gradually evolving from contact-based to non-contact-based methods. Common non-contact detection technologies include those based on infrared thermal imaging, visible light image analysis, and acoustic wave detection. However, because the chest displacement caused by vital activities such as respiration is limited and has a low frequency, the detected echo signal is concentrated in the low-frequency band (near DC), making it susceptible to interference from environmental noise and multipath effects, resulting in a reduced signal-to-noise ratio.

[0003] Compared to optical and acoustic methods, radar technology offers advantages such as strong penetration and all-weather operation, avoiding the risks of skin irritation and allergic reactions caused by electrode contact, and is insensitive to changes in ambient light and temperature. Millimeter-wave (mmWave) bands (typically 30-300 GHz) have shorter wavelengths, resulting in higher detection sensitivity for minute displacements. The amplitudes of localized vibrations on the body surface caused by respiration and heartbeat are approximately 1-5 mm and 100-500 µm, respectively. Millimeter-wave wavelengths are suitable for monitoring these small vibrations, thus enabling accurate monitoring of physiological signs. Furthermore, millimeter-wave radar possesses technical characteristics such as non-contact measurement, high precision (millimeter-level), penetration through clothing, and resistance to environmental interference, making it suitable for continuous monitoring scenarios. However, respiratory and heartbeat signals are typically weak and susceptible to interference from environmental noise and subtle body movements. Currently, the estimation of respiratory rate and heart rate relies heavily on signal decomposition algorithms, and existing methods suffer from problems such as mode aliasing and insufficient noise suppression, leading to increased errors in heart rate and respiratory rate estimation. Summary of the Invention

[0004] This invention addresses the numerous shortcomings in current vital sign detection signal processing. The technical problem it aims to solve is to overcome the deficiencies in existing technologies. This invention provides a millimeter-wave radar vital sign detection method and system based on ICEEMDAN combined with Hilbert spectral analysis, which can effectively enhance and extract vital sign signals, and achieve accurate estimation of respiration and heartbeat.

[0005] To address the aforementioned technical problems, this invention provides a method and system for detecting vital signs using millimeter-wave radar based on ICEEMDAN combined with Hilbert spectral analysis. The technical solution adopted is as follows:

[0006] Firstly, this paper presents a millimeter-wave radar vital sign detection method based on ICEEMDAN combined with Hilbert spectral analysis. The method involves acquiring raw vital sign signals using a millimeter-wave radar positioned directly in front of the target human body to obtain the IF signal. The raw signal undergoes preprocessing to obtain surface micro-motion signals. An frequency-modulated continuous wave radar module captures the target's echo signal, and further preprocessing is performed. The captured signal is converted to a range-amplitude spectrum using a 1D-FFT, and a zero-fill FFT is used to reduce phase jumps caused by body micro-motion. The MTI (Moving Target Indication) method is employed to suppress static clutter. After range FFT and static clutter suppression, the system determines the human target position by using incoherent accumulation and range-thresholding to locate the range cell with the highest energy. The phase of the vital sign signal is recovered using the linear characteristics of arctangent demodulation, and phase extraction is performed. After extracting the phase information of the target range cell, a phase unwrapping algorithm is used to eliminate phase wrapping jumps and remove phase ambiguity. Then, the adjacent frame difference method is used to enhance the heartbeat component and suppress the respiratory signal, and the moving average filter is used to remove impulse noise and improve signal quality. The ICEEMDAN algorithm is used to obtain candidate components of the vital signs signal; finally, the Hilbert spectral analysis algorithm is used to perform spectral analysis on the selected components to obtain the estimated frequency of the vital signs signal, realizing the detection of vital signs by millimeter-wave radar.

[0007] Furthermore, raw signals are obtained by collecting vital signs signals using millimeter-wave radar, including:

[0008] The FMCW millimeter-wave radar continuously transmits linear frequency modulated millimeter-wave signals to the area where the subject is located, including the subject and stationary objects; it receives the echo signals reflected from the subject's area, which include phase-modulated vital signs signals reflected from the subject's chest cavity due to breathing and heartbeat, as well as zero-frequency or fixed-phase clutter signals reflected from stationary objects.

[0009] Furthermore, the original signal undergoes signal preprocessing to obtain micro-motion signals on the body surface, including:

[0010] (1) Static target removal

[0011] After signal acquisition, the raw data is first read and reconstructed into a format suitable for subsequent processing. Then, a one-dimensional Fast Fourier Transform (FFT) is used to convert the signal into a range-amplitude spectrum that reflects the relationship between target distance and amplitude. Next, the Moving Target Indication (MTI) method is used to analyze the signal, distinguishing between static and dynamic targets. By extracting micro-motion signals related to human vital signs and suppressing clutter interference from static objects such as walls and furniture, static targets are ultimately filtered out, while dynamic information related to vital signs such as breathing and heartbeat is retained and highlighted.

[0012] (2) Distance door locking

[0013] After obtaining the distance-amplitude spectrum after removing static clutter, a zero-fill FFT technique is employed to smooth the spectrum by increasing frequency domain sampling points, effectively reducing phase jumps caused by subtle body movements and improving spectrum stability. The system uses incoherent accumulation to superimpose and average the spectral amplitude information from multiple consecutive time points, enhancing the stability of the human target signal while further suppressing random noise and residual interference. An adaptive energy detection threshold is set in the distance dimension. By comparing the energy intensity of each distance unit, the system accurately locates the distance unit with the most concentrated and significant energy, determining the precise distance position of the human target and clearly identifying it from the complex environmental background.

[0014] (3) Phase extraction

[0015] After locking onto the distance cell containing the human target, the phase value is calculated from the signal using the linearity of the arctangent demodulation method. This process directly converts the minute periodic displacement changes caused by respiration and heartbeat into corresponding, continuously fluctuating phase change curves over time, thus extracting the phase. Further, after performing adjacent frame differential and impulse noise removal processing on the micro-motion signal, the improved ICEEMDAN algorithm with adaptive noise is used for signal separation and extraction, yielding candidate components of the vital signs signal, including:

[0016] After extracting the phase information of the target range cell, the original phase curve is processed using a phase unwrapping algorithm to eliminate phase ambiguity and obtain the true displacement trajectory. Subsequently, a method of adjacent frame difference is used to reduce low-frequency trends such as respiratory signals and relatively enhance high-frequency fluctuations corresponding to heartbeats. A moving average filter is used to smooth the signal and filter out residual sudden impulse noise and high-frequency interference.

[0017] The improved ICEEMDAN algorithm with adaptive noise is used to perform mode decomposition on the signal to obtain the intrinsic mode components. According to the target's inherent modal components Signal separation and extraction are performed to obtain candidate components of vital sign signals.

[0018] Furthermore, the improved ICEEMDAN algorithm with adaptive noise is used to perform mode decomposition on the signal to obtain the intrinsic mode components. According to the target's inherent modal components Signal separation and extraction are performed to obtain candidate components of vital sign signals, including:

[0019] Step S1, through empirical mode decomposition The algorithm processes the preset Gaussian white noise signal and decomposes it into a series of intrinsic mode components arranged from high frequency to low frequency. These components are labeled as white noise modes at different stages in sequence.

[0020] Step S2: First-stage white noise mode components are added to the preprocessed vital sign phase signals to form the first mixed signal sequence. Empirical mode decomposition (EMD) is performed on this mixed sequence, and the first-order residual signal is calculated. The original phase signal is subtracted from this first-order residual signal; the result is the extracted first-order target intrinsic mode component.

[0021] Step S3: Based on the obtained first-order residual signal, a second-stage white noise mode component is added to construct a second mixed signal sequence. Empirical mode decomposition is then performed on this sequence again to obtain the second-order residual signal. The first-order residual signal is subtracted from the second-order residual signal to obtain the second-order target intrinsic mode component.

[0022] Step S4: Repeat the above iterative process: In the latest obtained residual signal, add the white noise mode components of the next stage in sequence to construct a new sequence, decompose to obtain a new residual, and obtain the new target intrinsic mode components by calculating the difference between two adjacent residuals, until the characteristics of the current residual signal meet the algorithm's preset stopping condition, at which point the iteration terminates.

[0023] Step S5: Collect all target intrinsic mode components obtained during the iteration process. Based on the typical frequency ranges corresponding to the respiratory and heartbeat signals, select the set of low-frequency components representing respiratory fluctuations and the set of high-frequency components representing heartbeat fluctuations from these components. By superimposing the components within each set, the candidate components of the vital signs signal are finally obtained.

[0024] Furthermore, the selected components are subjected to spectral analysis using the Hilbert spectral analysis algorithm to obtain the estimated frequencies of the vital signs signals, including:

[0025] Step S1 involves performing a Hilbert transform on the candidate components of the vital signs signal, converting the original real signal into a complex analytic signal while fully preserving the amplitude and phase information of the original signal.

[0026] Step S2: Extract the instantaneous amplitude sequence representing the change of signal intensity over time and the instantaneous phase sequence representing the change of signal angle over time from the generated analytical signal, and calculate the instantaneous frequency.

[0027] Step S3 integrates the information from the three dimensions of time, instantaneous frequency, and corresponding instantaneous amplitude, and accumulates and maps them on the time-frequency plane to form a Hilbert spectrum that can intuitively display the time-frequency distribution of signal energy.

[0028] Step S4: Integrate the Hilbert time spectrum along the time axis to calculate the Hilbert marginal spectrum of the cumulative energy distribution of each frequency component of the signal throughout the entire observation time.

[0029] Step S5: Peak values ​​are then extracted within the corresponding frequency band of the vital signs signal to calculate the estimated results of the vital signs parameters.

[0030] Secondly, a millimeter-wave radar vital sign detection system based on ICEEMDAN combined with Hilbert spectral analysis is provided, including:

[0031] The signal acquisition module is used to obtain raw signals by monitoring vital signs through FMCW millimeter-wave radar. The raw signals include phase-modulated vital sign signals reflected by the chest cavity of the subject due to breathing and heartbeat micro-movements, as well as zero-frequency or fixed-phase clutter signals reflected by stationary objects.

[0032] The signal preprocessing module is used to perform clutter processing and phase extraction on the raw signal.

[0033] The signal separation and reconstruction module is used to perform adjacent frame difference and impulse noise removal processing on the micro-motion signal of the body surface, and then use the improved ICEEMDAN algorithm with adaptive noise to separate and extract the signal to obtain the candidate components of the vital signs signal.

[0034] The spectrum analysis module uses the Hilbert spectrum analysis algorithm to perform spectrum analysis on the selected components and obtain the estimated frequency of the vital signs signal.

[0035] Furthermore, the signal acquisition module is specifically used to place the millimeter-wave radar directly in front of the human body being tested. It continuously transmits linearly frequency-modulated millimeter-wave signals to the area where the subject is located via the FMCW millimeter-wave radar. This area includes the subject and stationary objects. It receives the echo signals reflected from the subject's area. The echo signals include phase-modulated vital signs signals reflected from the subject's chest cavity due to breathing and heartbeat, as well as zero-frequency or fixed-phase clutter signals reflected from stationary objects.

[0036] Furthermore, the signal preprocessing module specifically converts the captured signal into a range-amplitude spectrum using a 1D-FFT, employs MTI (Moving Target Indication) to suppress static clutter, and distinguishes between static and dynamic targets. After completing the range FFT and static clutter suppression, a zero-fill FFT is used to reduce phase jumps caused by subtle body movements. The system determines the human target's location by using incoherent accumulation and range-thresholding to locate the range cell with the highest energy. The phase of the vital signs signal is recovered using the linear characteristics of arctangent demodulation, and phase extraction is performed.

[0037] Furthermore, the signal separation and reconstruction module includes:

[0038] In the signal optimization section, after extracting the phase information of the target range cell, the original phase curve is processed using a phase unwrapping algorithm to eliminate phase ambiguity and obtain the true displacement trajectory. Subsequently, a method of adjacent frame difference is used to reduce low-frequency trends such as respiratory signals and relatively enhance high-frequency fluctuations corresponding to heartbeats. A moving average filter is used to smooth the signal, filtering out residual sudden impulse noise and high-frequency interference.

[0039] In the signal reconstruction section, the improved ICEEMDAN algorithm with adaptive noise is used to perform mode decomposition on the signal to obtain the intrinsic mode components (IMFs). Based on the target IMFs, the signal is separated and extracted to obtain the candidate components of the vital signs signal.

[0040] Furthermore, the signal reconstruction section is specifically used to perform the following steps:

[0041] Step S1: The preset Gaussian white noise signal is processed by the Empirical Mode Decomposition (EMD) algorithm and decomposed into a series of intrinsic mode components arranged from high frequency to low frequency. These components are labeled as white noise modes at different stages in sequence.

[0042] Step S2: First-stage white noise mode components are added to the preprocessed vital sign phase signals to form the first mixed signal sequence. Empirical mode decomposition (EMD) is performed on this mixed sequence, and the first-order residual signal is calculated. The original phase signal is subtracted from this first-order residual signal; the result is the extracted first-order target intrinsic mode component.

[0043] Step S3: Based on the obtained first-order residual signal, a second-stage white noise mode component is added to construct a second mixed signal sequence. Empirical mode decomposition is then performed on this sequence again to obtain the second-order residual signal. The first-order residual signal is subtracted from the second-order residual signal to obtain the second-order target intrinsic mode component.

[0044] Step S4: Repeat the above iterative process: In the latest obtained residual signal, add the white noise mode components of the next stage in sequence to construct a new sequence, decompose to obtain a new residual, and obtain the new target intrinsic mode components by calculating the difference between two adjacent residuals, until the characteristics of the current residual signal meet the algorithm's preset stopping condition, at which point the iteration terminates.

[0045] Step S5: Collect all target intrinsic mode components obtained during the iteration process. Based on the typical frequency ranges corresponding to the respiratory and heartbeat signals, select the set of low-frequency components representing respiratory fluctuations and the set of high-frequency components representing heartbeat fluctuations from these components. By superimposing the components within each set, the candidate components of the vital signs signal are finally obtained.

[0046] Furthermore, the spectrum analysis module is used to perform spectrum analysis on the selected components using the Hilbert spectral analysis algorithm to obtain the estimated frequencies of the vital signs signal. Specifically, it performs the following steps:

[0047] Step S1 involves performing a Hilbert transform on the candidate components of the vital signs signal, converting the original real signal into a complex analytic signal while fully preserving the amplitude and phase information of the original signal.

[0048] Step S2: Extract the instantaneous amplitude sequence representing the change of signal intensity over time and the instantaneous phase sequence representing the change of signal angle over time from the generated analytical signal, and calculate the instantaneous frequency.

[0049] Step S3 integrates the information from the three dimensions of time, instantaneous frequency, and corresponding instantaneous amplitude, and accumulates and maps them on the time-frequency plane to form a Hilbert spectrum that can intuitively display the time-frequency distribution of signal energy.

[0050] Step S4: Integrate the Hilbert time spectrum along the time axis to calculate the Hilbert marginal spectrum of the cumulative energy distribution of each frequency component of the signal throughout the entire observation time.

[0051] Step S5: Peak values ​​are then extracted within the corresponding frequency band of the vital signs signal to calculate the estimated results of the vital signs parameters.

[0052] In recent years, the application of millimeter-wave radar technology in the field of vital sign detection has received increasing attention. Compared with traditional contact monitoring technologies, it has advantages such as non-contact measurement, real-time monitoring, high penetration, and strong anti-interference capabilities. Millimeter-wave radar-based vital sign detection technology has been widely used in many application scenarios, such as medical care, monitoring, and rescue. Extracting vital sign signals from radar echo signals is a key aspect of this technology. The Hilbert-Huang Transform (HHT) is an algorithm designed for time-frequency analysis, based on the Mode Decomposition (EMD) method. It is often used for vital sign signal extraction. The Hilbert marginal spectrum provides the energy distribution of signal frequency components in local segments, helping to identify the dominant frequency components in the signal. Frequency estimation can be performed quickly using the Hilbert marginal spectrum. However, due to some inherent problems of EMD, especially under low signal-to-noise ratio conditions, the EMD algorithm performs poorly. Signals of different frequencies are easily mixed in a single IMF, affecting the quality of signal decomposition and the final analysis results of HHT, thus leading to mode aliasing problems in the extraction results. To address this problem, this invention enhances the heartbeat signal through adjacent frame differential analysis, employs an improved fully empirical mode decomposition (EMD) with adaptive noise for vital sign signal separation, and combines this with Hilbert spectral analysis to obtain estimated frequencies of the vital sign signals. This overcomes the problems of the HHT algorithm being susceptible to noise and the insufficient frequency resolution of traditional spectral estimation methods. Compared with existing technologies, the millimeter-wave radar vital sign detection method and system proposed in this invention have the following advantages:

[0053] 1. With the high resolution and measurement accuracy of the FMCW radar, the system can sensitively capture weak physiological micro-motion signals such as breathing and heartbeat.

[0054] 2. This invention utilizes the long-range detection advantage of FMCW radar, and the entire process of vital sign monitoring does not require any electrodes or sensors to come into contact with the skin, which not only eliminates wearing discomfort, but also completely avoids the risk of cross-infection, achieving both comfort and safety.

[0055] 3. The algorithm of this invention performs zero-filled FFT and adjacent frame difference operations on the signal sequentially, which significantly improves the availability of phase features and recognition accuracy.

[0056] 4. This invention acquires raw vital sign signals using millimeter-wave radar, preprocesses these signals to obtain surface micro-motion signals, performs adjacent-frame differential analysis and impulse noise removal on these signals, and then applies an improved adaptive noise-complete ensemble empirical mode decomposition (ICEEMDAN) algorithm for mode decomposition. This reconstructs candidate components of the vital sign signals, and the selected components are then analyzed using the Hilbert spectral analysis algorithm to obtain estimated frequencies of the vital sign signals. By enhancing the heartbeat signal through adjacent-frame differential analysis and reconstructing the heartbeat and respiratory micro-motion signals using the ICEEMDAN algorithm, the mode aliasing problem caused by existing EMD algorithms is reduced. Furthermore, after denoising and signal separation using the ICEEMDAN algorithm, the Hilbert spectral analysis algorithm is used for spectral analysis to suppress interference and noise, improve the accuracy of the final spectral analysis, and effectively extract and estimate the frequencies of the vital sign signals. Attached Figure Description

[0057] Figure 1 This is a flowchart of the vital sign detection method based on ICEEMDAN combined with Hilbert spectral analysis according to the present invention;

[0058] Figure 2 This is a structural diagram of the vital signs detection system based on ICEEMDAN combined with Hilbert spectral analysis according to the present invention;

[0059] Figure 3 A slow-time matrix diagram for static clutter removal;

[0060] Figure 4 This is a target range gate information map in the echo signal;

[0061] Figure 5 To extract the phase information map of the target range cell;

[0062] Figure 6 This is a phase information diagram after impulse noise cancellation;

[0063] Figure 7 The isolated respiratory signal diagram;

[0064] Figure 8 The image shows the isolated heartbeat signal. Detailed Implementation Plan

[0065] To enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0066] Example 1: Principle of Millimeter-Wave Radar Vital Sign Signal Detection

[0067] 1.1 Human vital signs signals

[0068] Human vital signs manifest as subtle vibrations on the surface of the chest cavity caused by heartbeat and respiration. Under resting conditions, a healthy adult's heart beats approximately 60–120 times per minute, and their respiration is approximately 10–20 breaths per minute; as shown in Table 1, these signals are characterized by both weak amplitude and slow speed.

[0069] Table 1 Vital Signs

[0070] vital signs amplitude frequency breathe 1 - 12 mm 0.2 - 0.6 Hz Heartbeat 0.1 - 0.5 mm 0.8 - 2 Hz

[0071] Vital signs exhibit periodic changes; respiration and heartbeat can be approximated as sinusoidal components of varying frequencies. According to Table 1, the echo signals received by the FMCW radar can be modeled as periodic waves corresponding to this sinusoidal characteristic. For example, in formula (1):

[0072] (1)

[0073] in, The amplitude of the human respiratory signal vibration. The amplitude of the vibration signal of the human heartbeat. The respiratory signal frequency, The frequency of the heartbeat signal. This is a noise signal.

[0074] 1.2 Millimeter Wave Radar Working Principle

[0075] Frequency modulated continuous wave (FMCW) radar can continuously transmit frequency modulated signals, enabling it to measure the distance, velocity, and angle of objects. The transmitted waveform of FMCW radar is represented by formula (2):

[0076] (2)

[0077] In the formula, The center frequency of the radar. The slope of the sawtooth wave. This is the initial phase. The instantaneous distance from the antenna to the surface of the human chest cavity is given by formula (3):

[0078] (3)

[0079] In the formula, This is the distance between the vibration center of the human chest cavity and the antenna. The distance caused by the body's vital movements, as shown in formula (4):

[0080] (4)

[0081] In the formula, The amplitude of breathing. The frequency of breathing. This represents the heart rate amplitude. The frequency of heartbeats, This is due to other distance changes caused by vibrations. When radar waves encounter obstacles, they produce echoes. The signal and echo signal are:

[0082] (5)

[0083] (6)

[0084] In the formula, At the speed of light, For amplitude, This is the echo delay.

[0085] The two signals are mixed to obtain the IF intermediate frequency signal. .

[0086] (7)

[0087] In the formula, For the period of the sawtooth wave, For intermediate frequency signal frequency, phase .

[0088] Because respiration and heartbeat cause very small changes in the chest cavity, resulting in very small frequency variations, it is difficult to calculate the distance the chest cavity has moved. However, the distance can be calculated based on the phase of the intermediate frequency signal. Therefore, measuring small-scale vibrations is essentially monitoring the phase change of the frequency-modulated continuous wave signal within the target distance unit over time.

[0089] Example 2: A Millimeter-Wave Radar Vital Sign Detection Method Based on ICEEMDAN Combined with Hilbert Spectral Analysis

[0090] like Figure 1 As shown, this embodiment of the invention provides a millimeter-wave radar vital sign detection method based on ICEEMDAN combined with Hilbert spectral analysis, including:

[0091] Raw signals are obtained by collecting vital signs signals using millimeter-wave radar.

[0092] In the practical application of this embodiment, the key component of the FMCW millimeter-wave radar is the Texas Instruments AWR2243 millimeter-wave radar sensor, which, together with the DCA1000 high-speed data acquisition card, forms the hardware core. The AWR2243 integrates two transmit channels and four receive channels, and can output continuous frequency modulated waves in the 77–81 GHz frequency band; its radio frequency front end radiates electromagnetic waves directionally to the human chest cavity through a microstrip antenna, and receives the weak echoes generated by breathing fluctuations and heartbeat micro-movements.

[0093] Signal preprocessing is performed on the original signal to obtain micro-motion signals on the body surface;

[0094] Since the echo signal contains other useless clutter signals in addition to vital signs, signal preprocessing is required, including:

[0095] (1) Static target removal: The original data is read and recombined. The signal is converted into a distance-amplitude spectrum that reflects the relationship between the target distance and amplitude through a one-dimensional fast Fourier transform of the distance dimension. Then, the signal is analyzed using the moving target display method MTI to distinguish between static and dynamic targets. Micro-motion signals related to human vital signs are extracted, while suppressing clutter interference from static objects such as walls and furniture.

[0096] (2) Range gate locking: After obtaining the range-amplitude spectrum after removing static clutter, zero-fill FFT technology is used to reduce phase jumps caused by body micro-movements and improve the stability of the spectrum. The system superimposes and averages the spectral amplitude information of multiple consecutive moments through incoherent accumulation. An adaptive energy detection threshold is set in the range dimension. By comparing the energy intensity of each range unit, the range unit with the most concentrated and significant energy is accurately located, and the precise range position of the human target is determined, clearly identifying it from the complex environmental background.

[0097] (3) Phase extraction: After locking the distance cell where the human target is located, the phase value is calculated from the signal by using the linear characteristics of the arctangent demodulation method to complete the phase extraction.

[0098] After performing adjacent frame difference and impulse noise removal processing on the micro-motion signal on the body surface, the improved ICEEMDAN algorithm with adaptive noise is used for signal separation and extraction to obtain candidate components of the vital signs signal.

[0099] After extracting the phase information of the target range cell, the original phase curve is processed using a phase unwrapping algorithm. Subsequently, a method of adjacent frame difference is employed for further processing. A moving average filter is then used to smooth the signal, filtering out residual burst noise and high-frequency interference.

[0100] The improved ICEEMDAN algorithm with adaptive noise was used to perform mode decomposition on the signal to obtain candidate components of the vital signs signal.

[0101] Mode decomposition of the signal is performed using the ICEEMDAN algorithm. This indicates the input of vital sign phase signals; Indicates the first This is the first implementation of standard Gaussian white noise. ; This indicates the result obtained after decomposing the signal within the parentheses using EMD. Step ; For the first Noise intensity coefficient, Satisfy the following formula: , , This represents the current standard deviation of the residuals; For the first The residual after the next iteration ; Indicates the first The inherent modal components of the "target" (i.e., the final IMF).

[0102] The ICEEMDAN algorithm performs mode decomposition on signals as follows:

[0103] Step S1: The preset Gaussian white noise signal is processed using the Empirical Mode Decomposition (EMD) algorithm, decomposing it into a series of intrinsic mode components arranged from high frequency to low frequency. These components are sequentially labeled as white noise modes at different stages. For each... By performing EMD, we obtain a series of IMFs: ( , Large enough); put Called "the "White noise mode of the phase", for reference.

[0104] Step S2: Add the white noise mode component from the first stage to the preprocessed vital sign phase signal to form the first mixed signal sequence. Construction Mixed sequences, For each Do EMD, take the first one The overall average yields the first-order residual. Subtracting the original phase signal from this first-order residual signal yields the extracted first-order target intrinsic mode components. The first-order target components are then calculated. (Right now ).

[0105] Step S3: Based on the obtained first-order residual signal, a second-stage white noise mode component is added to construct a second mixed signal sequence. The sequence is then subjected to Empirical Mode Decomposition (EMD) again, and the first value is taken after EMD. Obtain the second-order residual signal Subtracting the first-order residual signal from the second-order residual signal yields the second-order target intrinsic mode components. (Right now ).

[0106] Step S4, repeat the above iterative process: in the latest obtained residual signal: The white noise modal components of the next stage are added sequentially to construct a new sequence: ( The first IMF of EMD is decomposed to obtain new residuals: The new target intrinsic mode components are obtained by calculating the difference between two adjacent residuals. (Right now The iteration continues until the characteristics of the current residual signal meet the algorithm's preset stopping condition, at which point the iteration terminates. Stop when any of the following conditions is met: Number of extreme points ; It is a monotonic function; (Preset maximum order).

[0107] Step S5: Collect all target intrinsic mode components obtained during the iteration process. Based on the typical frequency ranges corresponding to the respiratory and heartbeat signals, select the set of low-frequency components representing respiratory fluctuations and the set of high-frequency components representing heartbeat fluctuations from these components. By superimposing the components within each set, the candidate components of the vital signs signal are finally obtained.

[0108] The selected components are subjected to spectral analysis using the Hilbert spectral analysis algorithm to obtain the estimated frequency of the vital signs signal.

[0109] The principle of the Hilbert spectral analysis algorithm is to perform Hilbert transform on each intrinsic mode function (IMF) to construct an analytical signal, extract instantaneous amplitude and instantaneous frequency, and finally perform statistical analysis on the time-frequency plane with energy as the weight to obtain the marginal spectrum.

[0110] The execution flow of the Hilbert spectral analysis algorithm is as follows:

[0111] Step S1 involves performing a Hilbert transform on the candidate components of the vital signs signal, converting the original real signal into a complex analytic signal while fully preserving the amplitude and phase information of the original signal. For any real signal... Its Hilbert transform Defined as convolution, as in formula (8):

[0112] (8)

[0113] Step S2: Extract the instantaneous amplitude sequence representing the change of signal intensity over time and the instantaneous phase sequence representing the change of signal angle over time from the generated analytical signal, and calculate the instantaneous frequency. Combining it with its Hilbert transform, we obtain a complex analytic signal, as shown in formula (9):

[0114] (9)

[0115] Wherein: instantaneous amplitude Instantaneous phase

[0116] instantaneous frequency Defined as the derivative of the phase:

[0117] The instantaneous frequency of a discrete signal is calculated as shown in formula (10):

[0118] (10)

[0119] in, The sampling rate.

[0120] Step S3 integrates the information from the three dimensions of time, instantaneous frequency, and corresponding instantaneous amplitude, and accumulates and maps them on the time-frequency plane to form a Hilbert spectrum that can intuitively display the time-frequency distribution of signal energy.

[0121] For each IMF component Calculate its analytic signal The instantaneous amplitude is obtained. and instantaneous frequency The Hilbert spectrum is defined as the energy distribution of all IMFs in the time-frequency plane, as shown in equation (11):

[0122] (11)

[0123] Step S4: Integrate the Hilbert time spectrum along the time axis to calculate the Hilbert marginal spectrum of the cumulative energy distribution of each frequency component of the signal throughout the observation time. The marginal spectrum is the cumulative energy distribution along the frequency axis obtained by integrating the Hilbert spectrum over time, as shown in formula (12):

[0124] (12)

[0125] Step S5: Subsequently, peak values ​​are extracted within the corresponding frequency bands of the vital signs signals to calculate the estimated results of the vital signs parameters.

[0126] The principle of the embodiments of the present invention is as follows:

[0127] A millimeter-wave radar is positioned directly in front of the human body being detected to acquire vital signs signals, resulting in an IF signal. A frequency-modulated continuous-wave radar module captures the target's echo signal, undergoing signal preprocessing. The acquired signal is converted to a range-amplitude spectrum using a 1D-FFT, and static clutter is suppressed using the MTI (Moving Target Indication) method. After range FFT and static clutter suppression, zero-fill FFT is used to reduce phase jumps caused by subtle body movements. The system determines the human target's position by using incoherent accumulation and range-thresholding to locate the range cell with the highest energy. The phase of the vital signs signal is recovered using the linear characteristics of arctangent demodulation, and phase extraction is performed. After extracting the phase information from the target range cell, a phase unwrapping algorithm is used to eliminate phase ambiguity. Then, an adjacent-frame difference method is used to enhance the heartbeat component and suppress respiratory signals, and a moving average filter is used to remove impulse noise, improving signal quality. The ICEEMDAN algorithm is used to obtain candidate components of the vital signs signal; finally, the Hilbert spectral analysis algorithm is used to perform spectral analysis on the selected components to obtain the estimated frequency of the vital signs signal, thus realizing the detection of vital signs by millimeter-wave radar.

[0128] Example 3: Millimeter-wave radar vital sign detection system based on ICEEMDAN combined with Hilbert spectral analysis

[0129] The above embodiments describe in detail the implementation process and principle of the millimeter-wave radar vital sign detection method based on ICEEMDAN combined with Hilbert spectral analysis. The following embodiments describe a millimeter-wave radar vital sign detection system based on ICEEMDAN combined with Hilbert spectral analysis, such as... Figure 2 As shown, it includes:

[0130] The signal acquisition module is used to acquire raw signals by collecting vital signs signals using millimeter-wave radar.

[0131] The signal preprocessing module is used to perform clutter processing and phase extraction on the raw signal.

[0132] The signal separation and reconstruction module is used to perform adjacent frame difference and impulse noise removal processing on the micro-motion signal of the body surface, and then use the improved ICEEMDAN algorithm with adaptive noise to separate and extract the signal to obtain the candidate components of the vital signs signal.

[0133] The spectrum analysis module uses the Hilbert spectrum analysis algorithm to perform spectrum analysis on the selected components and obtain the estimated frequency of the vital signs signal.

[0134] The principle of the embodiments of the present invention is as follows:

[0135] The signal acquisition module uses a millimeter-wave radar positioned directly in front of the human body being detected to acquire vital signs signals, resulting in an IF signal. The signal preprocessing module uses an frequency-modulated continuous wave radar module to capture the target's echo signal and performs preprocessing steps. This involves converting the captured signal to a range-amplitude spectrum using a 1D-FFT, employing MTI (Moving Target Indication) to suppress static clutter, and then using zero-fill FFT to reduce phase jumps caused by subtle body movements after range FFT and static clutter suppression. The system determines the human target's position by using incoherent accumulation and range-thresholding to locate the range cell with the highest energy. The phase of the vital signs signal is recovered using the linear characteristics of arctangent demodulation, and phase extraction is performed. The signal separation and reconstruction module extracts the phase information from the target range cell and eliminates phase wrapping jumps using a phase unwrapping algorithm to remove phase ambiguity. Then, adjacent frame difference is used to enhance the heartbeat component and suppress respiratory signals, and moving average filtering removes impulse noise, improving signal quality. The ICEEMDAN algorithm is used to obtain candidate components of the vital signs signal; finally, the spectrum analysis module performs spectrum analysis on the selected components using the Hilbert spectrum analysis algorithm to obtain the estimated frequency of the vital signs signal, thus realizing the detection of vital signs by millimeter-wave radar.

[0136] Based on the above Figure 2 The embodiment shown, wherein,

[0137] The signal acquisition module is specifically used to obtain raw signals by monitoring vital signs through FMCW millimeter-wave radar. The raw signals include phase-modulated vital sign signals reflected by the chest cavity of the subject due to breathing and heartbeat micro-movements, as well as zero-frequency or fixed-phase clutter signals reflected by stationary objects.

[0138] Based on the above Figure 2 The embodiment shown, wherein,

[0139] The signal preprocessing module specifically converts the captured signal into a range-amplitude spectrum using a 1D-FFT. It employs the MTI (Moving Target Indication) method to suppress static clutter and distinguish between static and dynamic targets. After range FFT and static clutter suppression, zero-fill FFT is used to reduce phase jumps caused by subtle body movements. The system determines the human target's location by using incoherent accumulation and range-thresholding to locate the range cell with the highest energy. Phase extraction is then performed to recover the phase of the vital signs signal using the linear characteristics of arctangent demodulation.

[0140] Based on the above Figure 2 The embodiment shown, wherein,

[0141] The signal separation and reconstruction module includes:

[0142] In the signal optimization section, after extracting the phase information of the target range cell, the original phase curve is processed using a phase unwrapping algorithm. Adjacent frame difference is then used to reduce low-frequency trends such as respiratory signals and relatively enhance high-frequency fluctuations corresponding to heartbeats. A moving average filter is then used to smooth the signal, filtering out residual burst noise and high-frequency interference.

[0143] In the signal reconstruction section, the improved ICEEMDAN algorithm with adaptive noise fully integrated empirical mode decomposition is used to perform mode decomposition on the signal, obtaining candidate components of the vital sign signal. Specifically, this is used to perform the following steps:

[0144] Step S1: The preset Gaussian white noise signal is processed by the Empirical Mode Decomposition (EMD) algorithm and decomposed into a series of intrinsic mode components arranged from high frequency to low frequency. These components are labeled as white noise modes at different stages in sequence.

[0145] Step S2: First-stage white noise mode components are added to the preprocessed vital sign phase signals to form the first mixed signal sequence. Empirical mode decomposition (EMD) is performed on this mixed sequence, and the first-order residual signal is calculated. The original phase signal is subtracted from this first-order residual signal; the result is the extracted first-order target intrinsic mode component.

[0146] Step S3: Based on the obtained first-order residual signal, a second-stage white noise mode component is added to construct a second mixed signal sequence. Empirical mode decomposition is then performed on this sequence again to obtain the second-order residual signal. The first-order residual signal is subtracted from the second-order residual signal to obtain the second-order target intrinsic mode component.

[0147] Step S4: Repeat the above iterative process: In the latest obtained residual signal, add the white noise mode components of the next stage in sequence to construct a new sequence, decompose to obtain a new residual, and obtain the new target intrinsic mode components by calculating the difference between two adjacent residuals, until the characteristics of the current residual signal meet the algorithm's preset stopping condition, at which point the iteration terminates.

[0148] Step S5: Collect all target intrinsic mode components obtained during the iteration process. Based on the typical frequency ranges corresponding to the respiratory and heartbeat signals, select the set of low-frequency components representing respiratory fluctuations and the set of high-frequency components representing heartbeat fluctuations from these components. By superimposing the components within each set, the candidate components of the vital signs signal are finally obtained.

[0149] Based on the above Figure 2 The embodiment shown, wherein,

[0150] The spectrum analysis module uses the Hilbert spectral analysis algorithm to perform spectrum analysis on the selected components, obtaining the estimated frequencies of vital sign signals. Specifically, it is used to execute the following steps:

[0151] Step S1 involves performing a Hilbert transform on the candidate components of the vital signs signal, converting the original real signal into a complex analytic signal while fully preserving the amplitude and phase information of the original signal.

[0152] Step S2: Extract the instantaneous amplitude sequence representing the change of signal intensity over time and the instantaneous phase sequence representing the change of signal angle over time from the generated analytical signal, and calculate the instantaneous frequency.

[0153] Step S3 integrates the information from the three dimensions of time, instantaneous frequency, and corresponding instantaneous amplitude, and accumulates and maps them on the time-frequency plane to form a Hilbert spectrum that can intuitively display the time-frequency distribution of signal energy.

[0154] Step S4: Integrate the Hilbert time spectrum along the time axis to calculate the Hilbert marginal spectrum of the cumulative energy distribution of each frequency component of the signal throughout the entire observation time.

[0155] Step S5: Peak values ​​are then extracted within the corresponding frequency band of the vital signs signal to calculate the estimated results of the vital signs parameters.

[0156] Example 4: Experimental Results and Analysis

[0157] The accuracy of the algorithm in extracting vital signs was verified by measuring real signals. The entire experimental system used the Texas Instruments AWR2243 millimeter-wave radar sensor as the front end, with the DCA1000 high-speed data acquisition card forming the hardware core. The AWR2243 integrates two transmit channels and four receive channels, and can output continuous frequency modulated waves in the 77–81 GHz frequency band. Its RF front end radiates electromagnetic waves directionally to the human chest cavity through a microstrip antenna and receives weak echoes generated by respiratory fluctuations and heartbeat micro-movements. mmWave Studio, along with a Windows graphical interface program, was used to configure and control the TI radar sensor and receive the data stream from the analog-to-digital converter. The radar was placed stably about 1.5m in front of the subject's chest cavity to collect multiple sets of data. The test subject was an adult female in a sitting position, maintaining normal breathing. Data was collected for 51.2 seconds from any position within the platform, and the DCA1000 high-speed data acquisition card was used to capture the intermediate frequency signal of the AWR2243 in real time. After the system is powered on, the radio frequency signal is reflected back through the chest cavity, mixed with the transmit branch, filtered, and then output as an intermediate frequency (IF) signal. The IF signal is then converted into an ADC data stream through an analog-to-digital converter. The ADC data stream is transmitted to the DCA1000 via a wire. The DCA1000 is connected to a computer via an Ethernet interface, and the collected ADC data stream is transmitted to the computer for further analysis via the Ethernet interface. The collected ADC data stream is processed using the algorithm proposed in this chapter, and respiratory and heart rate data are separated to obtain respiratory and heart rate frequencies. AWR2243 parameters are shown in Table 2.

[0158] Table 2 AWR2243 Parameters

[0159] name parameter signal bandwidth 4GHz Starting frequency 77GHz FM slope 70MHz / μs Chirp cycle 50μs Frame period 50ms Sampling rate 4MHz Frames 1024

[0160] The actual measured target signal was analyzed using an algorithm based on ICEEMDAN combined with Hilbert spectral analysis. The radar's transmitted signal reflects off the target to obtain the echo signal. The MTI (Moving Target Indication) method was employed to suppress static clutter, such as... Figure 3 As shown. After performing range FFT and static clutter suppression, zero-fill FFT is used to reduce phase jumps under body micro-movement conditions. Incoherent accumulation of the FFT results in each direction is performed to reduce the impact of noise. The range-thresholded range cell with the highest energy is used to determine the human target position, as shown. Figure 4 As shown; extract the phase information of the target range cell, such as Figure 5 As shown, the phase of the vital signs signal is recovered using the linear characteristics of arctangent demodulation, and phase wrapping transitions are eliminated through a phase unwrapping algorithm to remove phase ambiguity. Further, adjacent frame difference is used to enhance the heartbeat component and suppress the respiratory signal. Moving average filtering removes impulse noise to obtain optimized phase information. The signal after impulse noise removal is shown in the figure. Figure 6 As shown. The candidate components of the vital signs signal are obtained using the ICEEMDAN algorithm, and then subjected to spectral analysis using the Hilbert spectral analysis algorithm. The separated respiratory signal is shown below. Figure 7 As shown, the separated heartbeat signal is as follows Figure 8 As shown in Tables 3 and 4, comparisons between the respiratory and heart rates obtained from multiple experimental groups and the actual values ​​are presented. The respiratory and heart rate estimation results obtained using this invention show good performance, thus demonstrating the beneficial effects of this invention.

[0161] Table 3 Comparison of respiratory rates

[0162] Group 1 Group 2 Group 3 Group 4 Group 5 Group 6 Group 7 True value 23 22 24 23 23 21 25 Calculated value 20 21 24 23 22 22 26

[0163] Table 4 Comparison of Heart Rates

[0164] Group 1 Group 2 Group 3 Group 4 Group 5 Group 6 Group 7 True value 90 93 87 91 91 92 85 Calculated value 87 91 91 88 91 90 81

Claims

1. A method for vital sign detection based on ICEEMDAN combined with Hilbert spectrum analysis of millimeter wave radar, characterized in that, First, raw vital sign signals are acquired using millimeter-wave radar, positioned directly in front of the human subject. This radar is used to collect vital sign signals, resulting in an IF signal. The raw signal undergoes preprocessing to obtain surface micro-motion signals. An frequency-modulated continuous wave radar module captures the target's echo signal, and further preprocessing steps are performed. The captured signal is converted to a range-amplitude spectrum using a 1D-FFT, and a zero-fill FFT is used to reduce phase jumps caused by body movement. Moving target indicator (MTI) is employed to suppress static clutter. After range FFT and static clutter suppression, the system determines the human target's location by using incoherent accumulation and range-thresholding to locate the range cell with the highest energy. The phase of the vital sign signal is recovered using the linear characteristics of arctangent demodulation, and phase extraction is performed. After extracting the phase information from the target range cell, a phase unwrapping algorithm is used to eliminate phase wrapping jumps and phase ambiguity. Then, an adjacent frame difference method is used to enhance the heartbeat component and suppress respiratory signals, and a moving average filter is used to remove impulse noise, improving signal quality. The ICEEMDAN algorithm is used to obtain candidate components of the vital signs signal; finally, the Hilbert spectral analysis algorithm is used to perform spectral analysis on the selected components to obtain the estimated frequency of the vital signs signal, thus realizing the detection of vital signs by millimeter-wave radar.

2. The millimeter wave radar vital sign detection method of claim 1, wherein, The original signal is preprocessed to obtain the micro-motion signal on the body surface. The steps are as follows: (1) Static target removal After signal acquisition, the raw data is first read and reconstructed into a format suitable for subsequent processing. Then, a one-dimensional fast Fourier transform in the distance dimension is used to convert the signal into a distance-amplitude spectrum that reflects the correspondence between target distance and amplitude. Afterward, the moving target indication (MTI) method is used to analyze the signal, distinguishing between static and dynamic targets. By extracting micro-motion signals related to human vital signs and suppressing clutter interference from static objects such as walls and furniture, static targets are ultimately filtered out, while dynamic information related to vital signs such as breathing and heartbeat is retained and highlighted. (2) Distance door locking After obtaining the distance-amplitude spectrum after removing static clutter, a zero-fill FFT technique is employed to smooth the spectrum by increasing frequency domain sampling points, effectively reducing phase jumps caused by subtle body movements and improving spectrum stability. The system uses incoherent accumulation to superimpose and average the spectral amplitude information from multiple consecutive moments, enhancing the stability of the human target signal while further suppressing random noise and residual interference. An adaptive energy detection threshold is set in the distance dimension; by comparing the energy intensity of each distance unit, the system accurately locates the distance unit with the most concentrated and significant energy, determining the precise distance location of the human target and clearly identifying it from the complex environmental background. (3) Phase extraction After locking onto the distance cell containing the human target, the phase value is calculated from the signal using the linearity of the arctangent demodulation method. This process directly converts the minute periodic displacement changes in the signal caused by breathing and heartbeat into corresponding phase change curves that fluctuate continuously over time, thus completing phase extraction.

3. The millimeter wave radar vital sign detection method of claim 1, wherein, The improved ICEEMDAN algorithm with adaptive noise is used to perform mode decomposition on the signal to obtain intrinsic mode components (IMFs). Based on the target IMFs, signal separation and extraction are performed to obtain candidate components of the vital sign signal. The steps are as follows: Step S1: The preset Gaussian white noise signal is processed using the Empirical Mode Decomposition (EMD) algorithm, decomposing it into a series of intrinsic mode components arranged from high frequency to low frequency. These components are sequentially labeled as white noise modes at different stages. For each... Perform EMD to obtain its series : ( , Large enough); put Called "the "White noise mode of the phase", for reference; Step S2: Add the white noise mode component from the first stage to the preprocessed vital sign phase signal to form the first mixed signal sequence. Construction Mixed sequences, For each Do EMD, take the first one The overall average yields the first-order residual. Subtracting the original phase signal from this first-order residual signal yields the extracted first-order target intrinsic mode components. The first-order target components are then calculated. (Right now ); Step S3: Based on the obtained first-order residual signal, a second-stage white noise mode component is added to construct a second mixed signal sequence. The sequence is then subjected to Empirical Mode Decomposition (EMD) again, and the first value is taken after EMD. Obtain the second-order residual signal Subtracting the first-order residual signal from the second-order residual signal yields the second-order target intrinsic mode components. (Right now ) Step S4, repeat the above iterative process: in the latest obtained residual signal: The white noise modal components of the next stage are added sequentially to construct a new sequence: ( The first IMF of EMD is decomposed to obtain new residuals: The new target intrinsic mode components are obtained by calculating the difference between two adjacent residuals. (Right now The iteration continues until the characteristics of the current residual signal meet the algorithm's preset stopping condition, at which point the iteration terminates. Stop when any of the following conditions is met: Number of extreme points ; It is a monotonic function; (Preset maximum order); Step S5: Collect all target intrinsic mode components obtained during the iteration process. Based on the typical frequency ranges corresponding to the respiratory and heartbeat signals, select the low-frequency component set representing respiratory fluctuations and the high-frequency component set representing heartbeat fluctuations from these components. By superimposing the components within each set, the candidate components of the vital signs signal are finally obtained.

4. The millimeter-wave radar vital sign detection method as described in claim 1, characterized in that, The selected components are analyzed using the Hilbert spectral analysis algorithm to obtain the estimated frequencies of vital signs signals. The steps are as follows: Step S1 involves performing a Hilbert transform on the candidate components of the vital signs signal, converting the original real signal into a complex analytic signal while fully preserving the amplitude and phase information of the original signal. For any real signal... Its Hilbert transform Defined as convolution, as in formula (1). （1）； Step S2: Extract the instantaneous amplitude sequence representing the change of signal intensity over time and the instantaneous phase sequence representing the change of signal angle over time from the generated analytical signal, and calculate the instantaneous frequency. Combining it with its Hilbert transform, we obtain a complex analytic signal, as shown in formula (2): （2） Wherein: instantaneous amplitude Instantaneous phase instantaneous frequency Defined as the derivative of the phase: The instantaneous frequency of a discrete signal is calculated as shown in formula (3): （3） in, Sampling rate; Step S3 integrates the information from the three dimensions of time, instantaneous frequency, and corresponding instantaneous amplitude, accumulating and mapping them on the time-frequency plane to form a Hilbert spectrum that can intuitively display the time-frequency distribution of signal energy. For each IMF component Calculate its analytic signal The instantaneous amplitude is obtained. and instantaneous frequency The Hilbert spectrum is defined as the energy distribution of all IMFs in the time-frequency plane, as shown in equation (4): （4）； Step S4: Integrate the Hilbert time spectrum along the time axis to calculate the Hilbert marginal spectrum of the cumulative energy distribution of each frequency component of the signal throughout the observation time. The marginal spectrum is the cumulative energy distribution along the frequency by integrating the Hilbert spectrum over time, as shown in formula (5): （5）； Step S5: Subsequently, peak values ​​are extracted within the corresponding frequency bands of the vital signs signals to calculate the estimated results of the vital signs parameters. .

5. A millimeter-wave radar vital sign detection system based on ICEEMDAN combined with Hilbert spectral analysis, characterized in that, include: The signal acquisition module is used to obtain raw signals by monitoring vital signs through FMCW millimeter-wave radar. The raw signals include phase-modulated vital sign signals reflected by the chest cavity of the subject due to breathing and heartbeat micro-movements, as well as zero-frequency or fixed-phase clutter signals reflected by stationary objects. The signal preprocessing module is used to perform clutter processing and phase extraction on the original signal; The signal separation and reconstruction module is used to perform adjacent frame difference and impulse noise removal processing on the micro-motion signal of the body surface, and then use the improved ICEEMDAN algorithm with adaptive noise to separate and extract the signal to obtain the candidate components of the vital sign signal. The spectrum analysis module uses the Hilbert spectrum analysis algorithm to perform spectrum analysis on the selected components and obtain the estimated frequency of the vital signs signal.

Images