Non-contact vital sign monitoring method and system based on FMCW radar

Abstract

The invention discloses a self-adaptive vital sign monitoring method and system based on an FMCW radar. Firstly, radar echo signals are received, a three-dimensional data matrix is generated, and amplitude signals and phase signals are extracted; then respectively determining an optimal breathing distance unit and an optimal heartbeat distance unit by adopting a self-adaptive distance unit locking technology; extracting a phase signal from the screened optimal breathing distance unit, and extracting a breathing signal by using a band-pass filter; phase signals are extracted from the selected optimal heartbeat distance unit and differentiated, and then combined denoising is carried out through a dynamic double-filter to obtain heartbeat signals. And finally, carrying out error correction on physiological parameters obtained by spectral analysis through sliding window statistical constraint. The method solves the problems of mutual interference of respiration and heartbeat signals, difficulty in extraction of weak heartbeat signals, insufficient environmental noise suppression and the like in existing radar monitoring, and can be widely applied to medical monitoring, smart home and health management scenes.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of radar signal processing and biomedical monitoring, and particularly relates to a non-contact vital sign monitoring method and system based on frequency modulated continuous wave (FMCW) radar, which is particularly suitable for dynamic tracking of respiratory rate and heart rate. BACKGROUND

[0002] In the prior art, vital sign monitoring mainly relies on contact technology, such as pulse oximeter, electrocardiogram (ECG) and phonocardiogram (PCG) and the like. Although these technologies can provide high-precision monitoring results, due to the need to directly contact the patient's body, these methods have some inherent disadvantages, such as patient discomfort, inconvenience during long-term monitoring, and secondary infection risk caused by contact.

[0003] In recent years, non-contact vital sign monitoring technology has been widely studied, among which radar-based monitoring methods have attracted much attention due to their advantages of no need for contact, strong penetration and privacy protection. For example, ultra-wideband (UWB) radar can achieve high-precision distance measurement by emitting short pulse signals, but its wideband characteristics result in high hardware cost and large system complexity; infrared imaging technology relies on body surface temperature changes to detect physiological activities, but it is easily disturbed by environmental temperature and cannot penetrate clothing; computer vision technology needs to rely on lighting conditions and fails in dark or obstructed scenes, and there is a risk of privacy leakage. Compared with the above, frequency modulated continuous wave (FMCW) radar transmits a linear frequency modulated signal and analyzes the frequency difference of the return signal to achieve target detection, which has high precision, low power consumption and anti-interference ability, and becomes an ideal choice for non-contact monitoring.

[0004] However, although FMCW radar technology has significant advantages in vital sign monitoring, it still faces some challenges. First, the distance unit (range bin) of the radar system will change in dynamic situations, especially when monitoring the micro-movement of the target, how to accurately select and lock the distance unit containing effective vital sign information is still a technical problem to be solved. Second, because the strength of the respiratory signal is usually greater than that of the heartbeat signal, the heartbeat signal is easily disturbed by the respiratory signal and noise, and it is still a complex task to extract accurate heartbeat signals. Therefore, the present application proposes a vital sign monitoring method based on adaptive range bin locking of FMCW radar, aiming to solve the above challenges and improve the vital sign monitoring accuracy and robustness under different environmental conditions and postures. SUMMARY

[0005] In order to solve the problems of dynamic drift of distance unit, coupling interference of breathing and heartbeat signals and noise sensitivity in existing FMCW radar in vital sign monitoring, the application provides an adaptive vital sign monitoring method and system based on FMCW radar. Through adaptive distance unit locking optimal breathing and heartbeat signals, dynamic double filter denoising and error correction technology, the monitoring accuracy and robustness of respiratory rate and heart rate are significantly improved.

[0006] In a first aspect, the embodiments of the present application provide a non-contact vital sign monitoring method based on FMCW radar, comprising the following steps:

[0007] S1, receiving radar echo signals and generating a three-dimensional data matrix, completing preprocessing through static clutter suppression and distance dimension fast Fourier transform, and extracting amplitude signals and phase signals.

[0008] S2, adopting an adaptive distance unit locking technology to determine an optimal breathing distance unit and an optimal heartbeat distance unit, wherein:

[0009] S2.1, determining an initial target distance unit based on the standard deviation of the amplitude signal, and selecting a candidate distance unit set centered on the initial target distance unit.

[0010] S2.2, in the candidate distance unit set, the optimal breathing distance unit is screened through an improved amplitude-phase correlation (MPC) index combined with a dynamic percentile threshold.

[0011] S2.3, in the candidate distance unit set, the optimal heartbeat distance unit is selected based on the maximum spectral kurtosis criterion.

[0012] S3, extracting the phase signal from the optimal breathing distance unit screened and applying a band-pass filter to extract the breathing signal; extracting the phase signal from the optimal heartbeat distance unit selected and differentiating, and then performing joint denoising through a dynamic double filter composed of a dynamic adaptive band-pass filter and a Savitzky-Golay filter, wherein the passband of the dynamic adaptive band-pass filter is dynamically adjusted according to the historical heart rate.

[0013] S4, performing spectral analysis on the breathing signal and the heartbeat signal obtained after filtering to obtain breathing and heart rate estimates, and correcting errors of the output results through statistical constraints to ensure that the estimates are within a physiological reasonable range.

[0014] In a possible implementation, the extraction of the candidate distance unit set is realized through the following steps:

[0015] (1) Calculate the standard deviation of the amplitude signal, and extract the initial target distance unit corresponding to the maximum standard deviation.

[0016] (2) Construct a symmetric window centered on the initial target distance unit, and screen to obtain a candidate distance unit set.

[0017] In a possible implementation, the screening of the optimal respiratory distance unit is achieved through the following steps:

[0018] (1) Calculate the amplitude-phase correlation MPC index in the candidate distance unit set.

[0019] (2) Calculate a dynamic threshold, construct a constraint function, and select the distance unit with the maximum MPC index as the optimal respiratory distance unit after applying the constraint function to the MPC index.

[0020] In a possible implementation, the selection of the optimal heartbeat distance unit specifically includes:

[0021] (1) In the candidate distance unit set, perform slow-time difference processing on the phase signal to generate a high-frequency enhanced phase difference sequence.

[0022] (2) Obtain the heartbeat signal corresponding to the phase difference sequence through the dynamic double filter at the previous moment, and then perform normalization and fast Fourier transform to obtain the frequency spectrum amplitude distribution.

[0023] (3) Calculate the spectral kurtosis value of each candidate distance unit, and select the distance unit with the maximum kurtosis value as the optimal heartbeat distance unit.

[0024] In a possible implementation, the implementation of the dynamic adaptive band-pass filter includes:

[0025] (1) First, determine whether the number of the sliding window at this time is greater than or equal to a preset threshold: if less than the preset threshold, use a band-pass filter with an initial passband, and skip the subsequent steps; if greater than or equal to the preset threshold, sequentially execute the subsequent steps.

[0026] (2) Calculate the mean value of the heart rate estimate value of the historical window before the current window to obtain the center frequency .

[0027] (3) According to the preset left and right margin parameters and , dynamically update the passband range to , and satisfy to match the physiological characteristics of the heart rate rise.

[0028] (4) Redesign the filter coefficients using a ripple optimization algorithm, with a passband ripple of ≤0.1 dB and a stopband attenuation of ≥40 dB.

[0029] In a possible implementation, the specific steps of the error correction are:

[0030] (1) Calculate the mean of the (respiration, heart rate) estimates of the last N history windows and the standard deviation .

[0031] (2) Calculate the lower limit and the upper limit according to the confidence coefficient .

[0032] (3) If the (respiration, heart rate) estimate of the current window is out of the range , correct it to the nearest limit value, otherwise directly output the original value.

[0033] In a second aspect, the embodiments of the present application provide a non-contact vital sign monitoring system based on FMCW radar, comprising the following modules:

[0034] A preprocessing module is configured to receive a radar echo signal and generate a three-dimensional data matrix, complete preprocessing through static clutter suppression and fast Fourier transform in the distance dimension, and extract an amplitude signal and a phase signal.

[0035] An adaptive distance cell locking module is configured to lock an optimal respiration distance cell and an optimal heartbeat distance cell, respectively, and comprises:

[0036] A candidate set selection submodule is configured to determine an initial target distance cell based on a standard deviation of the amplitude signal, and select a candidate distance cell set centered on the initial target distance cell.

[0037] A respiration distance cell screening submodule is configured to screen the optimal respiration distance cell from the candidate distance cell set by using an improved amplitude-phase correlation (MPC) index in combination with a dynamic percentile threshold.

[0038] A heartbeat distance cell selection submodule is configured to select the optimal heartbeat distance cell from the candidate distance cell set based on a maximum spectral kurtosis criterion.

[0039] A dynamic filtering module is configured to extract a phase signal from the optimal respiration distance cell screened and apply a band-pass filter to extract a respiration signal; extract a phase signal from the optimal heartbeat distance cell selected and difference, and then perform joint denoising by a dynamic double filter composed of a dynamic adaptive band-pass filter and a Savitzky-Golay filter to obtain a heartbeat signal, wherein a passband of the dynamic adaptive band-pass filter is dynamically adjusted according to a historical heart rate.

[0040] An error correction module is configured to perform spectral analysis on the respiration signal and the heartbeat signal obtained after filtering to obtain respiration and heart rate estimates, and perform error correction on the output results by statistical constraint to ensure that the estimates are within a physiological reasonable range.

[0041] In a third aspect, an electronic device is provided, comprising a processor and a memory;

[0042] The memory is configured to store a computer program;

[0043] The processor is configured to execute the computer program stored in the memory, thereby implementing any of the non-contact vital sign monitoring methods described herein.

[0044] In a fourth aspect, a computer readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, any of the non-contact vital sign monitoring methods described herein is implemented.

[0045] In a fifth aspect, a computer program product comprising instructions which, when executed on a computer, cause the computer to carry out any of the non-contact vital sign monitoring methods described herein.

[0046] Compared with the prior art, the present application achieves the following technical effects:

[0047] 1. Adaptive distance cell locking technology: The respiratory distance cell is locked by an improved MPC index (combined with a dynamic threshold), and an innovative heartbeat distance cell selection method based on spectral kurtosis is proposed, effectively solving the coupling interference problem caused by the spatial distribution difference between respiratory and heartbeat signals. Experiments show that this method can improve the accuracy of locking the heartbeat distance cell to more than 90% in complex environments.

[0048] 2. Dynamic dual-filter design: An adaptive band-pass filter (with a passband dynamically adjusted according to the heart rate) and a Savitzky-Golay filter are used for joint denoising, which suppresses respiratory harmonics and environmental noise while preserving the details of the heartbeat signal.

[0049] 3. Error correction mechanism: The mean and standard deviation of historical data are calculated by a sliding window to dynamically constrain the reasonable range of the current estimated value, avoiding abnormal output caused by sudden noise or distance cell drift. BRIEF DESCRIPTION OF DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and other embodiments can also be obtained by those skilled in the art based on these drawings.

[0051] Figure 1 is a flowchart of the non-contact vital sign monitoring method based on FMCW radar according to an embodiment of the present application;

[0052] Figure 2 is a radar echo signal matrix arrangement flowchart of an embodiment of the present application;

[0053] Figure 3 is dynamic double-filter extraction of heartbeat signals of an embodiment of the present application; (a) original phase signal; (b) after difference; (c) after dynamic band-pass filter; (d) after SG filter;

[0054] Figure 4 is an absolute error distribution diagram of output respiratory heart rate values and calibration values of an embodiment of the present application;

[0055] Figure 5 is output heartbeat and respiratory RMSE of an embodiment of the present application under different body postures. DETAILED DESCRIPTION

[0056] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art based on the present application belong to the scope of protection of the present application.

[0057] The present application solves the problems of mutual interference of respiratory and heartbeat signals, difficulty in extracting weak heartbeat signals, and insufficient suppression of environmental noise in existing radar monitoring through adaptive distance unit sorting, dynamic double-filter design, and error correction technology, and can be widely applied to medical monitoring, smart home, and health management scenarios.

[0058] EMBODIMENT

[0059] The present embodiment provides a non-contact vital sign monitoring method based on an FMCW radar, as shown in FIG. 1, including the following steps: Figure 1

[0060] S1, receive radar echo signals and generate a three-dimensional data matrix, complete preprocessing through static clutter suppression and distance dimension fast Fourier transform, and extract amplitude signals and phase signals.

[0061] S1.1, in an FMCW radar system, each transmission period (PRI) is represented by T, during which a chirp signal (also known as chirp) with a duration of T is transmitted. In each chirp, the frequency increases from f to f. The transmitted signal is reflected by the target and captured by the receiving antenna. The received signal is mixed with the transmitted signal to obtain an intermediate frequency signal:

[0062] ​​​​​

[0063] in It is the round-trip delay, and A is the beat frequency signal amplitude. This represents residual phase noise.

[0064] After the intermediate frequency signal is digitized by an ADC, a long-time discrete sequence is obtained. Each chirp in the long-time discrete sequence is generated by... It consists of 10 samples. A frame is made up of 1 chirp. Each frame constitutes a processing time window. Therefore, the complete dataset is constructed as a three-dimensional matrix. , Figure 2 The process of arranging the radar echo signal matrix is ​​demonstrated.

[0065] S1.2. Perform static clutter suppression and range-dimensional fast Fourier transform operations on the arranged three-dimensional matrix, and extract the amplitude and phase signals. First, to ensure time consistency and avoid phase changes caused by intra-frame motion, select the first chirp of each frame and perform static clutter suppression on it to obtain a two-dimensional matrix. The first dimension represents the fast time dimension, and the second dimension represents the slow time dimension:

[0066]

[0067] in Represents a set of time windows. To obtain distance information, the second dimension is fixed and an FFT transformation is performed along the first dimension. :

[0068]

[0069] in After FFT transformation, Represents distance unit, It reflects the time variation information within each distance unit. Amplitude signal Extraction, i.e., finding The amplitude; the phase signal needs to be... Perform phase demodulation and unwinding. It is a complex signal, and the initial phase signal can be obtained by arctangent demodulation. :

[0070]

[0071] Since the range of the arctangent function is Therefore, even if the actual phase exceeds this range, it will still be demodulated. range, causing phase wrapping, so an unwrapping operation is needed to restore the actual phase from the initial phase signal. The unwrapped phase is denoted as

[0072]

[0073] integer for handling phase wrapping and ensuring phase continuity based on the adjacent phase difference by iterative calculation. The integer is calculated as

[0074]

[0075] where is the floor function.

[0076] S2, adaptive distance cell locking technique is adopted to determine the optimal respiratory distance cell and the optimal heartbeat distance cell respectively.

[0077] S2.1, the initial target distance cell is determined based on the standard deviation of the amplitude signal, and a candidate distance cell set is selected around the initial target distance cell.

[0078] For each distance cell where is the total number of distance cells, the standard deviation of the amplitude signal is calculated as a detection indicator:

[0079]

[0080] The initial target distance cell is determined by selecting the maximum standard deviation, i.e. max, considering that the human body usually only affects a limited number of consecutive distance cells, and the maximum distance cell radius affected is , so a symmetric window is extracted from with a size of distance cells, forming a candidate distance cell set . This windowing operation focuses the analysis on the area where vital signs signals are most likely to exist, thereby excluding redundant distance cells that are unlikely to contain meaningful information:

[0081]

[0082] where is a positive integer.

[0083] ​S2.2, In the candidate distance cell set, the optimal respiratory distance cell is screened by the improved magnitude-phase correlation (MPC) index combined with a dynamic percentile threshold.

[0084] For each distance cell , the MPC (magnitude-phase consistency) index between its magnitude and phase signals is calculated . Distance cells containing vital sign signals exhibit a higher correlation between magnitude and phase:

[0085]

[0086] where are the mean values of the magnitude and phase signals, respectively, are the standard deviations of the magnitude and phase signals.

[0087] Although the MPC index has been verified as an effective method to identify respiratory-related distance cells, it is susceptible to false positives. Specifically, in distance cells dominated by static or near-DC signals, which have no physiological motion, the magnitude and phase signals can exhibit a higher correlation. This can lead to these distance cells being falsely identified as containing vital sign information. To address this issue, a dynamic threshold mechanism is introduced to suppress such false detections and enhance the robustness of distance cell selection, applying a percentile-based dynamic threshold to the MPC index . The statistical constraint function is constructed as follows:

[0088]

[0089] where is the screened MPC index; is the standard deviation of the magnitude signal, while the percentile-based dynamic threshold is given by the following equation:

[0090]

[0091] where denotes -th percentile operator. Next, the optimal respiratory distance cell can be obtained by simply finding the index of the maximum value in

[0092] S2.3, In the candidate distance cell set, the optimal heartbeat distance cell is selected based on the maximum spectral kurtosis criterion.

[0093] To decouple the high-frequency heartbeat component from the low-frequency respiratory artifact in the phase signal, a slow-time difference is performed in the candidate distance cell set to obtain a phase difference sequence composed of phase difference signals. The calculation formula of the phase difference signal is:​

[0094]

[0095] This operation emphasizes the time-phase variation proportional to the micro-Doppler feature, and can suppress the quasi-static respiratory component while amplifying the heartbeat-induced modulation, since the latter exhibits a steeper phase gradient. Then, the dynamic dual filter of the last time instant is applied (see step S3.2 for the dynamic dual filter update rule) to obtain the heartbeat signal .

[0096] To reduce the amplitude variation in the subsequent spectral analysis, each column of is normalized to obtain the normalized heartbeat signal :

[0097]

[0098] Then, the FFT operation is applied to the normalized heartbeat signal to obtain the spectrum of the normalized heartbeat signal :

[0099]

[0100] where represents the frequency index value. The unbiased spectral kurtosis is calculated for the obtained spectrum . The spectral kurtosis can quantify the sharpness of the peaks in the spectrum, since the frequency bands containing clear heartbeat reflections exhibit more obvious spectral peaks, and thus their kurtosis values are higher. Therefore, can be used to accurately locate the heartbeat range:

[0101]

[0102] where represent the mean and standard deviation of each distance unit of , respectively, and the constant D represents the normalized frequency index corresponding to the upper limit of the heartbeat frequency in the spectrum. This index value does not directly represent the physical frequency, but a relative position, and the actual physical frequency corresponding to it is determined by the sampling rate. Finally, the optimal heartbeat distance unit is selected by maximizing the spectral kurtosis index, i.e., finding the maximum value index in .

[0103] This adaptive locking scheme can effectively separate the processing of respiratory and heartbeat signals, thereby achieving robust range bin selection under different environmental conditions.

[0104] S3, extract the phase signal from the optimal breathing distance unit obtained from the screening and apply a band-pass filter to extract the breathing signal; extract the phase signal from the selected optimal heartbeat distance unit and difference it, and then pass it through a dynamic double filter composed of a dynamic adaptive band-pass filter and an SG (Savitzky-Golay) filter to obtain the heartbeat signal, wherein the passband of the dynamic adaptive band-pass filter is dynamically adjusted according to the historical heart rate. Figure 3 The process of extracting the heartbeat signal using the dynamic double filter is shown.

[0105] In a possible implementation, the dynamic double filter specifically includes:

[0106] The is defined as the heart rate estimate in the time window. Although the heart rate estimates of adjacent windows usually exhibit temporal consistency ( ), the sudden change of the heart rate value caused by the change of the distance unit due to environmental changes or human micro-motions may occur. In order to reduce the sudden change of the heart rate value while maintaining the time resolution, a dynamic adaptive band-pass filter with a time-varying cutoff frequency is designed, and the passband thereof is defined as:

[0107] .

[0108] Since the change of the passband is based on the historical heart rate value, when the number of historical heart rate values is too small (i.e., when is less than a preset threshold ), an initial passband is used, and the initial passband contains the upper and lower limits of the human heartbeat frequency ( ); as more data is collected (i.e., when is greater than or equal to the preset threshold ), the passband will be updated according to the average heart rate. The motivation of the update process is that the heart rate is observed to change gradually, and the phase signal is susceptible to noise and harmonics after differentiation. The update steps are as follows:

[0109] (1) Calculate the center frequency: in order to capture the typical heart rate, the center frequency is calculated according to the average of historical heart rate values:

[0110]

[0111] This average process takes advantage of the smooth temporal variation of the heartbeat signal.

[0112] (2) Update the filter parameters: using the preset parameters and representing the left and right margins, update the passband:

[0113]

[0114] The asymmetry (wider right window) reflects the physiological tendency for heart rate to rise faster than to fall.

[0115] (3) Redesign the filter: use the ripple optimization algorithm to design the impulse response of the band-pass filter , whose frequency response satisfies the following requirements:

[0116]

[0117] The passband is limited to:

[0118]

[0119] where and represent the passband ripple and stopband ripple, respectively, is the sampling frequency of the heartbeat signal, which is the frame frequency in the present invention, i.e. , represents the length of the filter, whose value is equal to , while is the transition bandwidth.

[0120] The phase difference signal after dynamic adaptive band-pass filtering still needs to be filtered by an SG filter to further filter out noise and high-frequency harmonics. This double filtering method combines dynamic adaptive band-pass filtering with SG smoothing, which not only preserves the basic characteristics of the heartbeat signal, but also reduces the influence of noise and harmonic interference.

[0121] S4, perform spectrum analysis on the filtered respiratory signal and heartbeat signal to obtain respiratory and heart rate estimates, and use sliding window statistics to constrain the output results to correct errors and ensure that the estimates are within the physiological reasonable range. The following takes the heartbeat signal estimate as an example, and the respiratory signal is the same (i.e. replace the heartbeat signal with the respiratory signal):

[0122] First, the discrete Fourier transform (DFT) is used to obtain the frequency spectrum of the heartbeat signal, and the maximum value of the frequency spectrum is selected as the current time window heart rate estimate , but the frequency resolution of its frequency spectrum is limited by , where is the duration of the window . In order to suppress false outliers and improve the robustness of real-time monitoring, a simple statistical constraint is performed based on the last heart rate values in the current time window. Select the last heart rate values , define:

[0123]

[0124] wherein is the confidence parameter. The corrected heart rate estimate is:

[0125]

[0126] wherein:

[0127] : the heart rate estimate (in bpm) for the current time window w.

[0128] : the corrected heart rate estimate for the current time window w.

[0129] : the mean and standard deviation of the previous heart rate values for the current time window w.

[0130] : the corrected lower and upper limits for the current time window w.

[0131] : the custom threshold value.

[0132] Experimental results and analysis

[0133] This example was conducted in a standard laboratory environment, using the same experimental subject and the same calibration equipment, to collect 10 minutes of continuous respiration and heart rate data, and to calibrate it synchronously with a contact reference device. The experimental results were collated and plotted as Figure 4 and Figure 5 .

[0134] 1. Absolute error distribution (see Figure 4 ): The horizontal axis is the "output estimate - calibration value" error, in bpm; the vertical axis is the cumulative proportion of the corresponding error interval. The proportion of respiration rate error within ±1 bpm reached 85%, and within ±2 bpm reached 95%; the proportion of heart rate error within ±1 bpm was about 78%, and within ±2 bpm was about 92%. The results show that the method of the present application can achieve high-precision monitoring at the sub-physiological jitter level most of the time.

[0135] 2. RMSE under different body postures (see Figure 5): For the six postures of face, back, left side, right side, supine and prone, the RMSE of the respiration rate and the heart rate is calculated respectively. The respiration RMSE range is 0.73 bpm (face)~2.29 bpm (right side); the heart rate RMSE range is 0.94 bpm (left side)~1.89 bpm (prone). Compared with the prior art, the method has better stability and precision under multi-posture conditions.

[0136] The application is suitable for medical monitoring, smart home and sports health monitoring scenes, and has significant advantages in monitoring sensitive groups such as bedridden patients and newborns.

[0137] The application embodiment also provides a non-contact vital sign monitoring system based on an FMCW radar, comprising the following modules:

[0138] A preprocessing module is used to receive a radar echo signal and generate a three-dimensional data matrix, complete preprocessing through static clutter suppression and distance dimension fast Fourier transform, and extract amplitude signals and phase signals.

[0139] An adaptive distance cell locking module is used to lock optimal respiration distance cells and optimal heartbeat distance cells respectively, and comprises:

[0140] A candidate set selection submodule is used to determine an initial target distance cell based on an amplitude signal standard deviation, and select a candidate distance cell set with the initial target distance cell as the center.

[0141] A respiration distance cell screening submodule is used to screen optimal respiration distance cells in the candidate distance cell set through an improved amplitude-phase correlation (MPC) index combined with a dynamic percentile threshold.

[0142] A heartbeat distance cell selection submodule is used to select optimal heartbeat distance cells in the candidate distance cell set based on a spectrum kurtosis maximization criterion.

[0143] A dynamic filtering module is used to extract phase signals from the screened optimal respiration distance cells and apply a band-pass filter to extract respiration signals; extract phase signals from the selected optimal heartbeat distance cells and difference, and then perform joint denoising through a dynamic double filter composed of a dynamic adaptive band-pass filter and a Savitzky-Golay filter to obtain heartbeat signals, wherein the passband of the dynamic adaptive band-pass filter is dynamically adjusted according to historical heart rates.

[0144] An error correction module is used to perform spectrum analysis on the filtered respiration signals and heartbeat signals to obtain respiration and heart rate estimates, and perform error correction on the output results through statistical constraints to ensure that the estimates are within a physiological reasonable range.

[0145] In a possible implementation, the candidate set selection sub-module operates as follows:

[0146] (1) Calculate the standard deviation of the amplitude signal Extract the initial target distance unit corresponding to the maximum standard deviation.

[0147] (2) Construct a symmetric window centered on the initial target distance unit, and screen to obtain a candidate distance unit set.

[0148] In a possible implementation, the respiratory distance unit screening sub-module operates as follows:

[0149] (1) In the candidate distance unit set, calculate the amplitude-phase correlation MPC index

[0150] (2) Calculate the dynamic threshold value and construct the constraint function After applying the constraint function to the MPC index, select the distance unit with the maximum MPC index value as the optimal respiratory distance unit.

[0151] In a possible implementation, the heartbeat distance unit selection sub-module operates as follows:

[0152] (1) In the candidate distance unit set, perform slow-time difference processing on the phase signal to generate a high-frequency enhanced phase difference sequence.

[0153] (2) Obtain the heartbeat signal corresponding to the phase difference sequence through the dynamic double filter of the previous moment, and then perform normalization and fast Fourier transform to obtain the frequency spectrum amplitude distribution.

[0154] (3) Calculate the spectral kurtosis value of each candidate distance unit, and select the distance unit with the maximum kurtosis value as the optimal heartbeat distance unit.

[0155] In a possible implementation, the dynamic adaptive band-pass filter of the dynamic filtering module includes:

[0156] (1) First, determine whether the number of sliding windows at this time is greater than or equal to a preset threshold value If it is less than the preset threshold value, a band-pass filter with an initial passband is used, and the subsequent steps are skipped; if it is greater than or equal to the preset threshold value, the subsequent steps are sequentially executed.

[0157] (2) Calculate the mean value of the heart rate estimate value of the historical window before the current window to obtain the center frequency .

[0158] (3) According to the preset left and right margin parameters and , dynamically update the passband range to , and satisfy to match the physiological characteristics of the heart rate rise.

[0159] (4) Redesign the filter coefficient by using ripple optimization algorithm, passband ripple ≤0.1 dB, stopband attenuation ≥40 dB.

[0160] In a possible implementation, the error correction module specifically operates as follows:

[0161] (1) Calculate the mean value of the (respiration, heart rate) estimated value of the last N historical windows and the standard deviation .

[0162] (2) Calculate the lower limit and the upper limit according to the confidence coefficient .

[0163] (3) If the (respiration, heart rate) estimated value of the current window is out of the range , correct it to the nearest limit value, otherwise directly output the original value.

[0164] The embodiment of the application further provides an electronic device, and the embodiment of the application provides an electronic device, including a processor and a memory.

[0165] The memory is used to store a computer program.

[0166] The processor is used to execute the program stored on the memory, and realize any method described in the application.

[0167] In a possible implementation, the electronic device of the embodiment of the application further includes a communication interface and a communication bus, wherein the processor, the communication interface, and the memory complete mutual communication through the communication bus.

[0168] The communication bus mentioned in the above electronic device can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0169] The communication interface is used for communication between the above electronic device and other devices.

[0170] The memory can include a random access memory (RAM) and can also include a non-volatile memory (NVM), such as at least one disk memory. Optionally, the memory can also be at least one storage device located remotely from the aforementioned processor.

[0171] The processor described above can be a general processor, including a central processing unit (CPU), a network processor (NP), etc.; can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component.

[0172] In yet another embodiment provided in the present application, a computer readable storage medium is provided, and the computer readable storage medium stores a computer program. The computer program is executed by a processor to implement any of the methods provided in the present application.

[0173] In yet another embodiment provided in the present application, a computer program product containing instructions, which, when run on a computer, causes the computer to execute any of the methods provided in the present application.

[0174] In the embodiments described above, all or some of the steps can be implemented by using software, hardware, firmware or any combination thereof. When implemented by using software, all or some of the steps can be implemented in the form of one or more computer programs. The computer program is stored in a computer readable storage medium, and includes all the procedures which can be performed by the mobile terminal or the base station, and / or specific device used by the mobile terminal or the base station. The computer readable storage medium includes: any tangible storage device which can store the programs and data, and which can be accessed by the computer, such as random access memory (RAM), flash memory, readonly memory (ROM), programmable read-only memory (PROM), electrically programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), magnetic devices, optical devices, and / or any other suitable device. The computer readable storage medium is distributed to computers, servers or data centers, and the programs or all or some of the programs included in the computer readable storage medium are downloaded from one computer readable storage medium to another computer readable storage medium. The computer readable storage medium can be accessed by the computer, and the programs included in the computer readable storage medium are executed by the computer. The computer can be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatuses. The programs stored in the computer readable storage medium can be transmitted to other computer readable storage media via a wired (for example, a coaxial cable, an optical fiber, a digital subscriber line (DSL)) or wireless (for example, infrared, wireless, microwave, and the like) channel, and can be transmitted between websites, computers, servers or data centers.

[0175] It should be noted that the relative terms, such as first and second, etc., are used only to differentiate one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. In addition, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device. Without more limitations, the element defined by the statement "including a" does not exclude the presence of additional identical elements in the process, method, article or device including the element.

[0176] Each of the embodiments in the specification is described in a related manner, and each embodiment focuses on the difference from other embodiments. The same and similar parts between the embodiments can be referred to each other.

[0177] The above only describes the preferred embodiments of the present application, and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A FMCW radar based non-contact vital sign monitoring method, characterized in that, The method comprises the following steps: S1, receiving a radar echo signal and generating a three-dimensional data matrix, completing preprocessing through static clutter suppression and distance dimension fast Fourier transform, and extracting amplitude signals and phase signals; S2, determining an optimal breathing distance unit and an optimal heartbeat distance unit respectively by using an adaptive distance unit locking technology, wherein: S2.1, determining an initial target distance unit based on amplitude signal standard deviation, and selecting a candidate distance unit set centered on the initial target distance unit; S2.2, in the candidate distance unit set, screening the optimal breathing distance unit by using an improved amplitude-phase correlation MPC index combined with a dynamic percentile threshold; S2.3, in the candidate distance unit set, selecting the optimal heartbeat distance unit based on a maximum spectral kurtosis criterion; S3, extracting phase signals from the screened optimal breathing distance unit and applying a band-pass filter to extract a breathing signal; extracting phase signals from the selected optimal heartbeat distance unit and differentiating, and then performing joint denoising by using a dynamic double filter composed of a dynamic adaptive band-pass filter and a Savitzky-Golay filter to obtain a heartbeat signal, wherein a passband of the dynamic adaptive band-pass filter is dynamically adjusted according to a historical heart rate; S4, performing spectral analysis on the filtered breathing signal and heartbeat signal to obtain breathing and heart rate estimation values, and correcting errors of output results by using statistical constraints to ensure that the estimation values are within a physiological reasonable range.

2. The FMCW radar-based non-contact vital sign monitoring method of claim 1, wherein, The extraction of the candidate distance unit set is realized by the following steps: (1) calculating the standard deviation of the amplitude signal extracting an initial target distance unit corresponding to the maximum standard deviation (2) constructing a symmetric window centered on the initial target distance unit to screen the candidate distance unit set.

3. The FMCW radar-based non-contact vital sign monitoring method according to claim 1 or 2, characterized in that, The screening of the optimal breathing distance unit is realized by the following steps: (1) calculating an amplitude-phase correlation MPC index in the candidate distance unit set; (2) Calculate dynamic threshold, build constraint function After applying the constraint function to the MPC index, the distance unit with the maximum MPC index is selected as the optimal breathing distance unit.

4. The FMCW radar-based non-contact vital sign monitoring method according to claim 1 or 2, characterized in that, The selection of the optimal heartbeat distance unit specifically comprises: (1) performing slow-time difference processing on the phase signals in the candidate distance unit set to generate a high-frequency enhanced phase difference sequence; (2) obtaining heartbeat signals corresponding to the phase difference sequence by using the dynamic double filter at the last moment, and then performing normalization and fast Fourier transform to obtain a spectral amplitude distribution; (3) calculating spectral kurtosis values of each candidate distance unit, and selecting a distance unit with the maximum kurtosis value as the optimal heartbeat distance unit.

5. The FMCW radar-based non-contact vital sign monitoring method of claim 1, wherein, The implementation of the dynamic adaptive band-pass filter comprises: (1) first determine whether the number of sliding windows at this time is greater than or equal to a preset threshold If less than the preset threshold, a bandpass filter with an initial passband is adopted, and the subsequent steps are skipped. If greater than or equal to the preset threshold, the subsequent steps are sequentially executed. (2) calculating a mean of the heart rate estimate values of the history windows before the current window, obtaining a center frequency ; (3) according to the preset left and right margin parameters and , the passband range is dynamically updated , and satisfies to match the physiological characteristics of heart rate rise; (4) redesigning filter coefficients by using a ripple optimization algorithm, wherein a passband ripple is less than or equal to 0.1 dB, and a stopband attenuation is greater than or equal to 40 dB.

6. The FMCW radar-based non-contact vital sign monitoring method of claim 1, wherein, The specific steps of the error correction are: (1) statistics of the mean of the respiratory or heart rate estimates of the last N historical windows and the standard deviation ; (2) according to the confidence coefficient lower limit of calculation and upper limit ; (3) If the current window's respiration or heart rate estimate is outside the range, it is corrected to the nearest limit, otherwise the raw value is output directly.

7. A FMCW radar based non-contact vital sign monitoring system, characterized in that, The method comprises the following modules: A preprocessing module is used for receiving a radar echo signal and generating a three-dimensional data matrix, completing preprocessing through static clutter suppression and distance dimension fast Fourier transform, and extracting amplitude signals and phase signals; An adaptive distance unit locking module is used for locking an optimal breathing distance unit and an optimal heartbeat distance unit respectively, and comprises: A candidate set selection submodule is used for determining an initial target distance unit based on amplitude signal standard deviation, and selecting a candidate distance unit set centered on the initial target distance unit; The breathing distance unit screening submodule screens the optimal breathing distance unit from the candidate distance unit set by using the improved amplitude-phase correlation MPC index combined with a dynamic percentile threshold; The heartbeat distance unit selection submodule selects the optimal heartbeat distance unit from the candidate distance unit set based on a spectral kurtosis maximization criterion; The dynamic double-filtering module extracts the phase signal from the optimal breathing distance unit screened and applies a band-pass filter to extract the breathing signal; extracts the phase signal from the optimal heartbeat distance unit selected and differentiates it, and then performs joint denoising by using a dynamic double-filtering module composed of a dynamic adaptive band-pass filter and a Savitzky-Golay filter to obtain the heartbeat signal, wherein the passband of the dynamic adaptive band-pass filter is dynamically adjusted according to the historical heart rate; The error correction module is used for performing spectral analysis on the breathing signal and the heartbeat signal obtained after filtering to obtain the breathing and heart rate estimates, and performing error correction on the output results by statistical constraint to ensure that the estimates are within the physiological reasonable range.

8. An electronic device, comprising: The processor and the memory are included; The memory is used for storing a computer program; The processor is used for executing the program stored on the memory to implement the non-contact vital sign monitoring method of any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer program is stored in the computer readable storage medium, and the computer program is executed by the processor to implement the non-contact vital sign monitoring method of any one of claims 1-6.

Images