Human respiration detection method, electronic equipment, medium and computer program product

Abstract

The invention provides a human respiration detection method, electronic equipment, a medium and a computer program product. The method comprises the following steps: acquiring a plurality of reflection signals of a target area; adjusting the plurality of reflected signals to the same phase to obtain a signal sequence; preprocessing the signal sequence; determining an optimal variable variational mode decomposition parameter by adopting a subtraction averaging optimizer algorithm; decomposing the preprocessed signal sequence by adopting a multi-variable variational mode decomposition method, and extracting a vital sign signal; and performing time-frequency transformation on the vital sign signal to obtain a respiratory frequency and a heartbeat frequency. According to the invention, a plurality of reflection signals are adjusted to the same phase, the reflection signals are enhanced, the interference of signals outside a target area can be effectively removed, and the detection precision is improved; the optimal parameter of the multi-variable variational mode decomposition is obtained through the subtraction averaging optimizer algorithm, the influence of the initial parameter on the performance of the multi-variable variational mode decomposition is reduced, and the universality of the detection method is improved.

Description

Technical Field

[0001] The present invention belongs to the field of wireless sensing technology, and in particular relates to a human breathing detection method, electronic equipment, medium and computer program product. Background Art

[0002] Vital sign monitoring plays a key role in the assessment of human health status. Traditional monitoring systems usually require users to wear or install special sensors. These methods are often invasive, inconvenient to use, and limit their application in practical scenarios. For example, although electrocardiograms (ECGs) and respiratory belts can provide accurate physiological data, their use is limited by the user's willingness to be monitored in specific environments. If there is only one device, but if there are multiple users, they have to queue up and cannot be monitored at the same time, which causes certain obstacles. In addition, these contact monitoring devices may cause infection risks during long-term use, especially in hospital environments.

[0003] To overcome these limitations, researchers have begun to explore non-invasive methods for vital sign detection. Wireless sensing technologies, especially those based on RFID and WIFI, have attracted attention for their ability to monitor vital signs without physical contact. These technologies extract vital sign signals by analyzing wireless signals reflected from the human body. However, these methods face significant challenges in practical applications: wireless signals are easily interfered by environmental factors, such as furniture, walls, and other targets, which complicates the interpretation of signals; user body movements can cause changes in signal characteristics, which is a huge challenge for vital sign detection systems based on wireless signals; differences between different environments and individuals require monitoring systems to have good generalization capabilities, while existing systems often perform poorly in new environments; it is very important to accurately identify multiple users and ensure the stability of system performance.

[0004] To address these issues, researchers have tried a variety of approaches, including using more complex signal processing algorithms to distinguish vital sign signals from motion interference, or developing more advanced sensors to reduce the impact of environmental noise. However, these approaches often require high costs, complex hardware equipment, or a large amount of data training, limiting their practical applications. In addition, existing vital sign monitoring methods perform well in monitoring stationary users, but their performance degrades significantly when the user is in a non-stationary state (such as slight body movement or hand / head movement). Such movement interferes with the signal, making it difficult to extract information related to vital signs. To address this issue, signal decomposition techniques such as empirical mode decomposition (EMD), variational mode decomposition (VMD), or empirical mode decomposition integration (EEMD) are used to decompose the composite signal into different components and extract relevant vital sign signals from it. However, these methods rely on some initial parameters (such as the number of decomposed components, initial center frequency, and bandwidth balance parameters), which need to be determined experimentally, limiting the application of these methods in different environments. Summary of the invention

[0005] The purpose of the present invention is to address the deficiencies in the prior art and provide a human respiration detection method, electronic equipment, medium and computer program product to reduce reliance on prior knowledge and improve the versatility of the detection method.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is:

[0007] A human breathing detection method comprises the following steps:

[0008] S1: Acquire multiple reflection signals of the target area;

[0009] S2: Adjust multiple reflected signals in the target area to the same phase to obtain a signal sequence;

[0010] S3: preprocessing the signal sequence;

[0011] S4: using a subtraction mean optimizer algorithm to determine optimal multivariate variational modal decomposition parameters, wherein the multivariate variational modal decomposition parameters include a mode number and a penalty factor;

[0012] S5: Based on the optimal multivariate variational mode decomposition parameters, the multivariate variational mode decomposition method is used to decompose the preprocessed signal sequence and extract the vital sign signal;

[0013] S6: Perform time-frequency transformation on the vital sign signal to obtain the respiratory frequency and heart rate.

[0014] The present invention adjusts multiple reflected signals to the same phase, calibrates and enhances the reflected signals, which can effectively remove the interference of signals outside the target area and improve the detection accuracy; the present invention obtains the optimal parameters of multivariate variational mode decomposition through a subtraction average optimizer algorithm, reduces the influence of initial parameters on the performance of multivariate variational mode decomposition, and improves the versatility of the detection method.

[0015] Furthermore, the target area includes a human chest and a human hand.

[0016] Furthermore, the signal sequence S total The expression is as follows:

[0017]

[0018] Among them, A i is the amplitude of the reflected signal received by the i-th antenna of the millimeter-wave radar, S i is the reflected signal received by the i-th antenna of the millimeter-wave radar, e -j(i-1)Δφ is the phase adjustment coefficient, a is the imaginary unit, M is the number of millimeter-wave radar antennas, and Δφ is the phase difference between millimeter-wave radar antennas.

[0019] Further, in S3, a mean filter is used to filter the signal sequence.

[0020] After mean filtering, the interference of static targets is removed, while the reflected signals of moving targets can be retained and significantly enhanced, the signal-to-noise ratio is significantly improved, and the target signal amplitude is more obvious.

[0021] Furthermore, in S3, after filtering, the channel impulse response of the target area at each moment is shifted to align the channel impulse response with the maximum cross-correlation point, and the data 1 second before the channel impulse response mutation point and 2-3 seconds after the alignment are removed.

[0022] To eliminate the interference of body movement on the signal.

[0023] Furthermore, the implementation process of S4 includes:

[0024] Set the range of multivariate variational mode decomposition parameters;

[0025] Generate multiple agent positions, each agent position represents a combination of a modal number and a penalty factor;

[0026] For each agent position, the preprocessed signal sequence is decomposed into modes through multivariate variational mode decomposition, and the fitness value of each agent position is calculated. The expression of fitness value fitness is as follows:

[0027]

[0028] The maximum value among the fitness values ​​of the agent positions is selected as the optimal fitness value, and the mode number and penalty factor corresponding to the optimal fitness value are used as the optimal multivariate variational mode decomposition parameters;

[0029] Where k and c represent the mode index and channel index of multivariate variational mode decomposition, respectively, SE c,k is the sample entropy of the preprocessed signal sequence, ρ c,k is the Pearson correlation coefficient between the preprocessed signal sequence and the mode.

[0030] Furthermore, the implementation process of S5 includes:

[0031] By minimizing the cost function, the preprocessed signal sequence is decomposed into multiple modes;

[0032] Cost Function The expression is as follows:

[0033]

[0034] u ij (t) = A ij ·cos(2πf j t+φ j )

[0035] According to the frequency of the vital sign signal, a mode corresponding to the vital sign signal is screened out from multiple modes, and the mode corresponding to the vital sign signal is reconstructed into a new signal to obtain the vital sign signal;

[0036] The expression of the vital sign signal is as follows:

[0037]

[0038] Where N is the number of channels of the preprocessed signal sequence, x i (t) is the signal of the i-th channel, u ij (t) is the jth mode, K is the number of modes, α is the penalty factor, and f j is the center frequency of the jth mode, β is the alignment parameter, ∥∥ represents the norm, A ij is the amplitude of the jth mode, φ j is the phase of the jth mode; is the vital sign signal of the ith channel, [f min ,f max ] is the frequency range of vital sign signals.

[0039] Based on the same inventive concept, the present invention also provides an electronic device, including:

[0040] one or more processors;

[0041] A memory having one or more programs stored thereon, which, when executed by the one or more processors, enables the one or more processors to implement the steps of the human breathing detection method.

[0042] Based on the same inventive concept, the present invention also provides a computer-readable storage medium storing a computer program, which implements the steps of the human breathing detection method when executed by a processor.

[0043] Based on the same inventive concept, the present invention also provides a computer program product, including a computer program / instruction; when the computer program / instruction is executed by a processor, the steps of the human breathing detection method are implemented.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] The present invention adjusts multiple reflected signals to the same phase, calibrates and enhances the reflected signals, which can effectively remove the interference of signals outside the target area and improve the detection accuracy; the present invention obtains the optimal parameters of multivariate variational mode decomposition through a subtraction average optimizer algorithm, reduces the influence of initial parameters on the performance of multivariate variational mode decomposition, and improves the versatility of the detection method. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 Schematic diagram of the process of the human breathing detection method of the present invention;

[0047] Figure 2 A schematic diagram of the phase adjustment of the reflected signal of the present invention;

[0048] Figure 3 is a schematic diagram of a millimeter wave radar device;

[0049] Figure 4 A schematic diagram comparing the respiratory frequency errors obtained by the method of the present invention and other methods;

[0050] Figure 5 The figure is a schematic diagram comparing the heart rate errors obtained by the method of the present invention and other methods. DETAILED DESCRIPTION

[0051] The present invention will be described in detail below in conjunction with the embodiments. It should be noted that the embodiments and features of the embodiments of the present invention can be combined with each other without conflict. For the convenience of description, if the words "upper", "lower", "left" and "right" appear in the following, they only indicate that the upper, lower, left and right directions are consistent with the drawings themselves, and do not limit the structure.

[0052] Example

[0053] like Figure 1 The human breathing detection method of this embodiment includes the following steps:

[0054] 1. Connect the Texas Instruments TI AWR1642 millimeter-wave radar module to the DCA1000EVM, and transfer the collected data of the human body's environment to the computer via Ethernet. The radar operates in the 77-81GHz frequency range, with a maximum bandwidth of 4GHz, and can provide a distance resolution of about 3.5cm and a displacement resolution of 4mm. The radar's transmission power complies with FCC regulations, with a maximum effective isotropic radiated power (EIRP) of 21dBm. Equipment such as Figure 3 shown.

[0055] Use multiple antenna arrays to receive signals, and enhance the reflected signals in the target area through digital beamforming; adjust the phase of the signal of each antenna through the phase adjustment factor, and then multiply and sum the signals to enhance the reflected signals in the target area to obtain a signal sequence.

[0056] In order to enhance the reflected signal of the target area (such as the chest and hands) at an angle θ, it is first necessary to define the target area as the area where the radar focuses on monitoring, such as the chest. The determination of the target area is usually set based on the radar field of view and the position of the body, and because the antenna array of the millimeter-wave radar has multiple antennas that receive the reflected signal of the same target area, but these signals usually have different phase differences due to the antenna position (relative to the target area). In order for these signals to enhance each other when synthesized, it is necessary to phase adjust the signal received by each antenna so that the signal from the target area has the same phase when received by all antennas. In this way, the signals of all antennas can have a stronger signal after superposition. Specifically, the reflected signal received by the i-th antenna is multiplied by the phase adjustment coefficient (e -a(i-1)Δφ ), the physical meaning of the phase adjustment coefficient is to correct the phase of the signal according to the position of each antenna, so that the reflected signals of the target area received by all antennas have the same phase, where Δφ represents the phase difference between antennas, which is calculated based on the geometric configuration of the antenna array and the relative position of the target area:

[0057]

[0058] Among them, d i is the propagation distance from the ith antenna to the target area, d j is the propagation distance from the jth antenna to the target area, and λ is the wavelength of the transmitted signal.

[0059] Then add the signals from each antenna to get the signal sequence S total .

[0060]

[0061] Among them, A i is the amplitude of the reflected signal received by the i-th antenna of the millimeter-wave radar, S i is the reflected signal received by the i-th antenna of the millimeter-wave radar, e -j(i-1)Δφ is the phase adjustment coefficient, a is the imaginary unit, M is the number of millimeter-wave radar antennas, and Δφ is the phase difference between millimeter-wave radar antennas.

[0062] During signal synthesis, only the antenna signals of the target area are phase adjusted to ensure that the phase of the target area is consistent on all antennas. Therefore, the vector addition in the same direction enhances the reflected signal from the target area. In contrast, the signals from other areas of the body have different reflection angles, and the adjusted phases are not the same. Therefore, these signals will cancel or subtract each other during the addition process, and the resulting synthetic vector is much smaller than the reflected signal from the target area. In this way, the reflected signal from the target area is enhanced at an angle θ, while the interfering reflected signals from other areas of the body are effectively attenuated. This phase adjustment process is the core principle of beamforming technology. Through beamforming, the received signal of the antenna array can be concentrated in a specific direction, thereby enhancing the signal strength of the target area. Since beamforming makes the phase of the signals from the target area consistent, a stronger synthetic signal is formed, such as Figure 2 .

[0063] 2. Preprocess the signal sequence

[0064] The signal sequence is filtered using a mean filter to filter the static target reflection signal (environmental noise) and enhance the moving target reflection signal. Since the distance from the static target to the radar is constant, the static target receives the signal R[m 1 ,n 1 ] is also constant, and it can be considered that these signals do not change in time. In contrast, the reflected signal of the moving target is affected by the Doppler effect, and its delay, frequency and phase will change with time. Therefore, among multiple pulse signals, the reflected signal of the moving target shows instability. 1 ,n 1 ] and average it to get the reference pulse receiving signal Z c [i]:

[0065]

[0066] Where, L is the total number of pulse received signals, m 1 and n 1 are the sampling points of fast time and slow time respectively.

[0067] The calculated reference pulse received signal represents the reflected signal of the static target, which is consistent with the noise in the static environment. In order to remove the interference of these static targets, it is only necessary to add the reference pulse received signal to each pulse received signal R[m 1 ,n 1 ] to obtain the reflected signal of the moving target:

[0068] X′(t)=R[m 1 ,n 1 ]-Z c [i]

[0069] After mean filtering, the interference of static targets is removed, while the reflected signals of moving targets can be retained and significantly enhanced, and the signal-to-noise ratio is significantly improved.

[0070] Although the signal has been enhanced by filtering, these steps fail to completely eliminate the interference caused by body motion. Therefore, the signal is further processed, which includes the alignment of the chest region signal and the removal of interference data. Considering that body motion may cause the position of the chest to shift, it is impossible to extract effective vital sign signals in the subsequent process. In order to ensure that the ranging point of the channel impulse response (CIR) corresponding to the measured chest remains constant, it is necessary to calculate the two-dimensional correlation between consecutive CIRs. By periodically translating the CIR at each moment to align it with the maximum cross-correlation point, the alignment of the chest region frame is achieved. However, alignment alone is not enough because there are still parts of limb movement in the desired fast time index sequence. Therefore, in order to process these parts of limb movement, it is necessary to remove the data 1 second before the CIR mutation point and 2-3 seconds after alignment. This is because motion perturbations will introduce error deviations in these data segments. After these processes, the interference caused by body movement is effectively removed.

[0071] The process of obtaining the channel impulse response is as follows:

[0072] The radar transmits a series of waveforms, called chirps (frequency modulated pulses), which are emitted periodically and the frequency of each chirp signal increases linearly. f is the duration of each chirp signal. A chirp is a single transmission. For the nth chirp, its transmission signal can be expressed as:

[0073]

[0074] Among them, f c is the initial frequency of the chirp, t∈[0,T c ], T c is the duration of the chirp, BW is the bandwidth, AT is the amplitude of the transmitted signal. When the electromagnetic wave signal is transmitted to the human chest and reflected, the reflected signal It can be expressed as:

[0075]

[0076] Among them, A R is the amplitude of the received signal, represents the round trip delay of the pth reflection path, d p is the distance from the radar to the target, c is the speed of light, and P represents the total number of reflection points. By mixing the received reflected signal with the transmitted signal, the resulting IF signal can be expressed as:

[0077]

[0078] In addition, by utilizing multiple antennas to improve the angular resolution, the channel information can be further expressed as:

[0079]

[0080] Among them, f p is the frequency of the IF signal, A is the amplitude of the IF signal, λ c represents the wavelength of the chirp signal, d l is the relative distance introduced by the lth antenna, θ 1 is the azimuth of the target. m (t, l) is subjected to fast Fourier transform (FFT) to convert the channel information into channel impulse response (CIR), which is expressed as:

[0081]

[0082] Among them, h r,l (m) represents the channel impulse response of the lth antenna at the rth distance tap (range resolution unit) at time m, and N1 is the total number of chirp signals.

[0083] The maximum cross-correlation point refers to the time point corresponding to the maximum value obtained when calculating the two-dimensional correlation between two signals (for example, the channel impulse response at consecutive moments). The two-dimensional correlation is a function used to measure the similarity and alignment between two signals, and is usually used for signal delay estimation and synchronization.

[0084] The two-dimensional correlation between consecutive CIRs is performed by evaluating the similarity between the two signals (CIRs), usually obtained using a cross-correlation method. Specifically, for two channel impulse responses h i (t) and h j (t), the two-dimensional correlation can be expressed as:

[0085]

[0086] Where Δt is the time offset, which represents h i (t) relative to the signal h j (t), T is the length of the signal. This step is to calculate the similarity of the two CIR signals under different time delays.

[0087] By calculating the maximum value of the cross-correlation function, we can find the maximum similarity point Δt between the two signals. max , which is the maximum cross-correlation point, and this maximum point corresponds to the best alignment position of the signal:

[0088]

[0089] According to Δt max , periodically translate the CIR at each moment to achieve alignment of the chest region frames. However, alignment alone is not enough, because there are still parts of limb movement in the desired fast time index sequence. Therefore, in order to process these parts of limb movement, it is necessary to remove the data 1 second before the CIR mutation point and 2-3 seconds after alignment. This is because in these data segments, motion disturbances will introduce error deviations. After these processes, the interference caused by body movement is effectively removed.

[0090] 3. The subtraction average optimizer (SABO) method is used to obtain the optimal parameters of the multivariate variational mode decomposition (MVMD), eliminating the need for prior knowledge of the initial parameters and improving the universality of the system.

[0091] SABO uses the subtraction mean of multiple agents to update the position of population members in the search space, so as to find the optimal modal number K and penalty factor α of MVMD decomposition, and the MVMD decomposition is optimal. The population member represents each individual in the optimization process, which is usually a combination of parameters (K and α in MVMD), and the search space represents the set of all possible solutions, that is, the range of all possible values ​​of parameters K and α, and the population members search in this space.

[0092] Among them, SABO is based on a special operation "-v", called v-subtraction of subtracting search agent B from search agent C, and is defined as follows:

[0093]

[0094] Where v is a random number generated from [1,2]. is a random value that follows a normal distribution; F(C) and F(B) are the values ​​of the objective function of the search agents C and B, respectively, which are used to evaluate the quality of the current solution of each search agent. This objective function is usually expressed as a function that needs to be minimized (or maximized), depending on the setting of the optimization problem. sign is the signum function. In the SABO algorithm, any search agent X i The displacement in the search space is done by each search agent X j From Search Agent X i Therefore, the new position of each search agent is calculated using the following formula:

[0095]

[0096] Among them, T 1 represents the total number of individuals, X i represents the position of the i-th search agent, X j represents the position of the jth search agent, represents a random value that follows a normal distribution, is the new position of the ith search agent.

[0097] If the new position has a higher fitness value, then accept the new position, otherwise keep it unchanged:

[0098]

[0099] Among them, F i and Respectively represent agent X i and The objective function value of .

[0100] In order to optimize MVMD, a suitable objective function needs to be defined. In MVMD, the goal is to decompose the process variable into regular oscillations, so sample entropy (SE) is used to measure the regularity of the signal. SE is mainly used to describe the complexity of time series. It has the advantages of fewer data points required for calculation and strong anti-noise and anti-interference capabilities. It is suitable for deterministic signals and random signals.

[0101] Specifically, SE distinguishes different processes by calculating the marginal probability distribution of the signal sequence. It can quantitatively describe the randomness and irregularity of the signal sequence. The more complex the sequence, the greater its approximate entropy. The calculation process of SE is as follows:

[0102] (1) Assume that the signal x has been c (t) is decomposed by MVMD, and the kth mode is obtained as:

[0103] u ck(t)=u(n)={u(1),u(2),...,u(i),...,u(Q)}

[0104] Among them, x c (t) is the c-th channel signal of signal X(t), Q is the length of the data, and represents the total length of the time series.

[0105] (2) Constructing the data modal vector: Given τ, the original time series is transformed into a τ-dimensional vector:

[0106] U τ (i)={u(i),u(i+1),…,u(i+τ-1)},i=1,2,…,Q-τ+1

[0107] After the above steps, a sequence consisting of multiple τ-dimensional vectors is constructed from the original time series.

[0108] (3) Calculate the distance between vectors: Define U τ (i) and U τ The distance between (j) is:

[0109]

[0110] This distance metric considers the maximum difference between two vectors in each dimension and can reflect the similarity between them.

[0111] (4) Calculate the proportion of distances less than the threshold: Assume that a similarity tolerance r is given y , calculate the distance less than r y U τ Specifically, count how many j such that d[U τ (i),U τ (j)] <r y , and define the following ratios:

[0112]

[0113] Among them, num represents the number of j that meets the conditions, and Q-τ+1 is the total number of all possible i.

[0114] (5) Calculate the above ratio The average value of:

[0115]

[0116] This is a global similarity measure that describes the similarity between all points in the data.

[0117] (6) Increase the dimension of the data pattern vector to τ+1, and then repeat steps (2) to (5) to obtain a new ratio B τ+1 (r y ). This can reflect the similarity in higher dimensions.

[0118] (7) Finally, the sample entropy SE of the preprocessed signal sequence is c,k Defined as:

[0119]

[0120] Sample entropy measures the variation of similar patterns in data under different dimensions. A lower sample entropy value indicates a stronger regularity of the data, while a higher sample entropy value indicates a higher complexity of the data.

[0121] However, sample entropy can only measure the complexity of the pattern and cannot reflect the relationship between the mode and the signal. Therefore, PCC (Pearson correlation coefficient) is introduced to make up for this deficiency and further optimize the effect of MVMD:

[0122]

[0123] Among them, ρ c,k is the Pearson correlation coefficient of the preprocessed signal sequence and mode, Cov(·) and Var(·) represent the covariance and variance respectively, u ck (t) is the mode, x c (t) is the c-th channel signal of the multi-channel signal X(t).

[0124] The specific steps of obtaining the optimal parameters of multivariate variational mode decomposition (MVMD) using the subtraction average optimizer (SABO) method are as follows:

[0125] (1) Initialize the parameters of the SABO algorithm, including the population size and the maximum number of iterations, and set the parameter range of the MVMD, including the minimum and maximum values ​​of the modal number and the minimum and maximum values ​​of the penalty factor; α∈[5e 2 ,4e 3 ], K = 3, 4, ..., 9;

[0126] (2) Generate initial agent positions: First, randomly generate an initial position matrix in the search space, where the position of each agent represents a possible combination of K and α;

[0127] (3) Calculate the initial fitness value (objective function value): By performing MVMD decomposition, calculate the fitness value (fitness) of each agent position and evaluate its optimization effect. The fitness function uses an SE-based objective function, which reflects the complexity of the signal and determines the quality of the optimization:

[0128]

[0129] Where k and c represent the mode index and channel index of multivariate variational mode decomposition, respectively, SE c,k is the sample entropy of the preprocessed signal sequence, which is obtained by calculating the marginal probability distribution of the signal sequence and is mainly used to describe the complexity of the signal sequence; ρ c,k is the Pearson correlation coefficient between the preprocessed signal sequence and the mode.

[0130] (4) Perform iterative optimization: starting from the first iteration and ending at the maximum number of iterations. For each agent, use the subtraction mean method in SABO to update its position. At the new position, perform MVMD decomposition and calculate the new fitness value. If the fitness value of the new position is better (i.e., the objective function value is larger), then update the position and fitness value of the agent.

[0131] (5) In all iterations, the agent position with the maximum fitness value is finally selected as the optimal solution, and the corresponding K and α values ​​are returned as the optimal parameters of the optimized MVMD decomposition.

[0132] 4. The optimal decomposition parameters K and α of MVMD, where K determines the number of intrinsic mode functions (IMFs) extracted during the decomposition process, and α controls the bandwidth of the mode, which usually refers to a parameter used to control overfitting or oversmoothing during the decomposition process. The introduction of α can help balance the accuracy and generalization ability of the decomposition. It is usually added to the optimization problem as a regularization term to avoid overly complex or overly smooth decomposition.

[0133] The signal X(t) obtained after preprocessing is decomposed using MVMD.

[0134] The dimension of X(t) is N×M, where N is the number of channels of the signal and M is the time length of the signal.

[0135] X(t) = [x 1 (t),x 2 (t),…,x i (t),…,x N (t)]

[0136] Among them, x i (t) represents the signal of the i-th channel.

[0137] The goal of modal decomposition is to transform each signal channel x i (t) is decomposed into several modes u ij (t), where j is the number of the mode, and each mode contains different frequency components. MVMD minimizes a cost function that includes reconstruction error, bandwidth constraint, and frequency alignment. To achieve this decomposition:

[0138]

[0139] Among them, the first term is the reconstruction error, which represents the difference between the decomposed mode and the original signal. The second term is the bandwidth constraint term, which uses the second-order derivative to measure the smoothness of each mode. The parameter α controls the degree of smoothness. The third term is the frequency alignment constraint; u ij (t) is the jth mode, f j represents the center frequency of the jth mode; β is the alignment parameter, which controls the strictness of the frequency alignment, and ∥∥ represents the norm.

[0140] By minimizing the above cost function, MVMD decomposes the signal into K modes u ij (t), each mode has a different center frequency, bandwidth and phase. Each mode can be expressed as:

[0141] u ij (t) = A ij ·cos(2πf j t+φ j )

[0142] Among them, A ij is the amplitude of the jth mode, φ j is the phase of the jth mode.

[0143] By screening the frequency components of each mode, the mode related to the vital sign signal is selected. Vital sign signals (such as heart rate, breathing rate, etc.) are usually within a specific frequency range (breathing: 0.15-0.5Hz, heart rate: 1-2Hz). Therefore, the mode corresponding to the frequency component related to the vital sign signal is selected and reconstructed into a new signal.

[0144] For each channel signal x i (t), reconstructed vital sign signal for:

[0145]

[0146] Among them, [f min ,f max ] is the frequency range of vital sign signals.

[0147] MVMD is not only able to extract modalities from a single signal, but also align frequency components across multiple channels. This is particularly useful for multi-channel signals. With the help of MVMD, it is possible to ensure that modal frequency components in multiple signal channels are aligned, thereby improving the matching of cross-channel signals. For physiological parameters such as heart rate and respiratory rate, MVMD is able to identify the same frequency patterns in multiple signal channels, which helps to accurately identify and analyze these vital signs signals.

[0148] 5. Convert the time domain signal into a frequency domain signal through Fourier transform, and then judge the target user's breathing rate and heart rate according to the typical frequency range of breathing and heart rate (breathing: 0.15-0.5Hz, heart rate: 1-2Hz).

[0149] According to the human breathing detection method based on exercise proposed in the embodiment of the present invention, the breathing error is less than 0.3rpm and the heart rate error is less than 2bpm in three different environments. Figure 4 , Figure 5 , superior to other vital sign monitoring methods (1. Run Zhao, Dong Wang, Qian Zhang, Haonan Chen, and Anna Huang. 2018. CRH: AContactless Respiration and Heartbeat Monitoring System with COTS RFIDTags. In 15th Annual IEEE International Conference on Sensing, Communication, and Networking, SECON 2018, Hong Kong, China, June 11-13, 2018. IEEE, 325–333.2, Fengyu Wang, Feng Zhang, Chenshu Wu, Beibei Wang, and KJRay Liu. 2021. ViMo: Multiperson Vital Sign Monitoring Using Commodity Millimeter-Wave Radio. IEEE Internet Things J. 8, 3 (2021), 1294–1307.333.).

[0150] This method studies the method of eliminating body motion interference and separating signals based on millimeter waves, and designs an adaptive optimization scheme to intelligently extract high-quality vital sign signals from the environment, while using alternating beamforming to maximize the performance for multiple targets. This method makes it possible to accurately monitor multi-target human vital signs in different scenarios. This method can not only significantly reduce the body motion interference during monitoring, but also has relatively good generalization performance, and finally achieves high detection accuracy and strong practicality.

[0151] Another embodiment of the present invention provides an electronic device, including:

[0152] one or more processors;

[0153] A memory stores one or more programs, and when the one or more programs are executed by one or more processors, the one or more processors implement the steps of the human breathing detection method.

[0154] In some implementations, the memory may be a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory.

[0155] In some other implementations, the processor may be a central processing unit (CPU), a digital signal processor (DSP), or other general-purpose processors of various types, which are not limited herein.

[0156] Another embodiment of the present invention provides a computer-readable storage medium storing a computer program, which implements the steps of the human breathing detection method when executed by a processor.

[0157] Another embodiment of the present invention provides a computer program product, including a computer program / instruction; when the computer program / instruction is executed by a processor, the steps of the human breathing detection method are implemented.

[0158] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0159] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices, and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as a combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0160] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0161] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0162] The contents explained in the above embodiments should be understood as these embodiments are only used to more clearly illustrate the present invention, and are not used to limit the scope of the present invention. After reading the present invention, various equivalent forms of modifications to the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

Claims

1. A human breathing detection method, characterized in that: The following steps are involved: S1: Acquire multiple reflection signals of the target area; S2: Adjust multiple reflected signals in the target area to the same phase to obtain a signal sequence; S3: preprocessing the signal sequence; S4: using a subtraction mean optimizer algorithm to determine optimal multivariate variational modal decomposition parameters, wherein the multivariate variational modal decomposition parameters include a mode number and a penalty factor; S5: Based on the optimal multivariate variational mode decomposition parameters, the multivariate variational mode decomposition method is used to decompose the preprocessed signal sequence and extract the vital sign signal; S6: Perform time-frequency transformation on the vital sign signal to obtain the respiratory frequency and heart rate.

2. The human breathing detection method according to claim 1, characterized in that: The target area includes a human chest and a human hand.

3. The human breathing detection method according to claim 1, characterized in that: Signal sequence S total The expression is as follows: Among them, A i is the amplitude of the reflected signal received by the i-th antenna of the millimeter-wave radar, S i is the reflected signal received by the i-th antenna of the millimeter-wave radar, e -j(i-1)Δφ is the phase adjustment coefficient, a is the imaginary unit, M is the number of millimeter-wave radar antennas, and Δφ is the phase difference between millimeter-wave radar antennas.

4. The human breathing detection method according to claim 1, characterized in that: In S3, a mean filter is used to filter the signal sequence.

5. The human breathing detection method according to claim 4, characterized in that: In S3, after filtering, the channel impulse response of the target area at each moment is shifted to align the channel impulse response with the maximum cross-correlation point, and the data 1 second before the channel impulse response mutation point and 2-3 seconds after the alignment are removed.

6. The human breathing detection method according to claim 1, characterized in that: The implementation process of S4 includes: Set the range of multivariate variational mode decomposition parameters; Generate multiple agent positions, each agent position represents a combination of a modal number and a penalty factor; For each agent position, the preprocessed signal sequence is decomposed into modes through multivariate variational mode decomposition, and the fitness value of each agent position is calculated. The expression of fitness value fitness is as follows: The maximum value among the fitness values ​​of the agent positions is selected as the optimal fitness value, and the mode number and penalty factor corresponding to the optimal fitness value are used as the optimal multivariate variational mode decomposition parameters; Where k and c represent the mode index and channel index of multivariate variational mode decomposition, respectively, SE c,k is the sample entropy of the preprocessed signal sequence, ρ c,k is the Pearson correlation coefficient between the preprocessed signal sequence and the mode.

7. The human breathing detection method according to claim 1, characterized in that: The implementation process of S5 includes: By minimizing the cost function, the preprocessed signal sequence is decomposed into multiple modes; Cost Function The expression is as follows: u ij (t)=A ij ·cos(2πf j t+φ j ) According to the frequency of the vital sign signal, a mode corresponding to the vital sign signal is screened out from multiple modes, and the mode corresponding to the vital sign signal is reconstructed into a new signal to obtain the vital sign signal; The expression of the vital sign signal is as follows: Where N is the number of channels of the preprocessed signal sequence, x i (t) is the signal of the ith channel, u ij (t) is the jth mode, K is the number of modes, α is the penalty factor, and f j is the center frequency of the jth mode, β is the alignment parameter, ∥∥ represents the norm, A ij is the amplitude of the jth mode, φ j is the phase of the jth mode; is the vital sign signal of the ith channel, [f min ,f max ] is the frequency range of vital sign signals.

8. An electronic device, characterized in that: include: one or more processors; A memory having one or more programs stored thereon, which, when the one or more programs are executed by the one or more processors, enables the one or more processors to implement the steps of the method according to any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that: The computer program is stored therein, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer program product comprising a computer program / instructions; characterized in that When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images