Method for detecting heart rate under body movement based on millimeter wave radar

Abstract

The invention relates to a millimeter-wave radar sign detection technology, in particular to a method for rapidly detecting the heart rate under body movement based on a millimeter-wave radar. According to the method, when the distance bins are selected, the used time window is far smaller than that of frequency estimation, the adjacent distance bins obtained through selection are spliced through the distance bin splicing method, the signal length suitable for short-time frequency estimation is obtained, the distance bin where the heart of the human body in the random body movement state is located can be accurately selected through the operation, and the accuracy of the distance bin is improved. Interference of random body motion is eliminated, the signal length enough for frequency estimation is obtained, and the accuracy of heart rate frequency estimation is improved. Non-contact detection is achieved through the millimeter wave radar, and rapid and accurate heart rate detection under a short time window can be achieved while random body movement of the human body can be restrained.

Description

Technical Field

[0001] The present invention relates to a millimeter wave radar vital sign detection technology, and in particular to a method for rapid detection of heart rate during body movement based on millimeter wave radar. Background Art

[0002] Faced with the increasingly severe challenges of an aging population and the growing number of chronic heart disease patients, the demand for daily heart rate monitoring has grown significantly. Traditional contact monitoring tools, such as electrocardiographs, patch electrodes, and watches, while effective, are limited by the need for direct contact with the body, which can be uncomfortable. Watches and other monitoring devices only take a long time to measure average heart rate, lack real-time performance, and are unable to detect heart problems such as sudden heart rate changes. On the other hand, non-contact monitoring methods, such as camera surveillance, while offering the possibility of long-distance monitoring, often touch upon sensitive areas of personal privacy and are susceptible to environmental factors such as light and obstructions. Furthermore, while WiFi and continuous wave radar technologies offer the advantages of wireless monitoring, they cannot distinguish between multiple monitored objects in the same space. In contrast, frequency modulated continuous wave (FMCW) radar, with its compact design, low cost, high accuracy, and ability to seamlessly track multiple targets without infringing on personal privacy, offers an innovative solution to address current monitoring needs.

[0003] FMCW radar technology, thanks to its frequency modulation characteristics, can capture distance data of target objects and subtly reveal subtle movements of the target object through phase fluctuations, such as tiny chest displacements driven by heartbeats. Currently, using FMCW radar for heart rate detection faces two challenges: detecting rapid heart rates and suppressing random body motion.

[0004] Many applications require rapid heart rate detection, such as measuring heart rate variability and sudden heart rate changes within a short period of time and tracking human heart targets. To improve the accuracy of vital sign detection, many existing algorithms tend to analyze signals over long time windows and combine complex techniques such as modal decomposition or wavelet analysis to separate respiratory and heartbeat signals. While these strategies are effective to a certain extent, they often come with high computational costs and time consumption. Furthermore, frequency analysis of long-window signals only provides average heart rate information, which limits the ability to acquire and dynamically monitor vital sign signals in real time, bringing inconvenience and challenges to practical applications.

[0005] Random body motion (RBM) can significantly reduce accuracy. Many current methods for suppressing random body motion are complex to implement or require complex and expensive devices, and also require a long time window. Obviously, measuring heart rate in a short period of time can effectively reduce the impact of random body motion, because the probability of random body motion in the short term is relatively low compared to the long term. However, since the frequency resolution after the fast Fourier transform (FFT) is inversely proportional to the observation time, using a simple FFT cannot accurately measure the heart rate from the time window required for frequency estimation, especially in the presence of respiratory harmonics and system noise, which is also a difficulty. Summary of the Invention

[0006] The purpose of the present invention is to provide a method for rapid heart rate detection under body motion based on millimeter-wave radar, so as to overcome the shortcomings of existing methods, and significantly improve the accuracy of heart rate detection and the real-time performance of the system while suppressing random body motion interference.

[0007] The technical method of the present invention is as follows:

[0008] A method for detecting heart rate during body movement based on millimeter-wave radar, the method comprising the following steps:

[0009] Step 1: Collect millimeter-wave radar echo signals, locate the range bin where the human heart is located within a 0.05-second time window using the energy variance method under random body motion interference, and calculate the phase of all echo signals in each range bin within the time window;

[0010] Step 2: Perform phase difference processing on the phase characteristics in each time window to obtain a phase difference signal, and perform time domain splicing on the phase difference signals of adjacent range bins within a 5-second time window based on the range bin splicing method to form an enhanced phase signal;

[0011] Step 3: Using a median filter with a step size of 5 to suppress pulse noise and inter-pulse interference on the enhanced phase signal to obtain a denoised phase signal;

[0012] Step 4, decomposing the denoised phase signal into a respiratory signal component and a heartbeat signal component through a bandpass filter bank, wherein the respiratory signal component is filtered using a 0.1-0.5 Hz bandpass filter, and the heartbeat signal component is filtered using a 0.8-2 Hz bandpass filter;

[0013] Step 5, performing short-time frequency estimation on the respiratory and heartbeat signal components respectively: using the MUSIC algorithm to perform short-time frequency estimation on the respiratory signal component to generate a respiratory frequency spectrum, and using the MUSIC algorithm to perform short-time frequency estimation on the heartbeat signal component to generate a heartbeat frequency spectrum;

[0014] Step 6: Eliminate respiratory harmonic interference in the heart rate spectrum based on the respiratory frequency spectrum, and then extract accurate heart rate characteristic values ​​through peak search.

[0015] In step 1, the time window of 0.05 seconds used by the range bin is selected for example. The specific time window needs to be determined according to the radar frame interval, which is an integer multiple of the radar frame interval. The radar needs to transmit more than or equal to two chirp signals at equal intervals in each frame.

[0016] In step 1, the energy variance method used to select the distance bin is specifically:

[0017] Calculate the energy variance of all echo signals in each time window for each range bin, and select the range bin with the largest variance as the range bin where the heart is located in that time window. Note that due to random body motion, the range bins selected for each time window may not be consistent.

[0018] Other methods that can accurately select the distance bin where the heart is located can also be used instead, such as the energy mean method, the phase variance method, etc.

[0019] In step 2, the reason why the distance bin splicing method can directly splice signals from different time windows and distance bins is specifically explained as follows:

[0020] The phase signal received by the radar can be specifically expressed as:

[0021]

[0022] Here, R(t) represents the displacement of the signal reflection surface, which can be decomposed into the fixed distance d0 between the radar and the target, the different sizes of d0 represent the different range bins, and the chest movement r(t) caused by breathing and heartbeat.

[0023] After differentiating the phase signal, the signal becomes:

[0024] φ(t)=[φ1,φ2,…,φ M ]

[0025] φ′(t)=[φ2-φ1,φ3-φ2,…,φ M -φ M-1 ]

[0026] It can be seen that signal differencing eliminates the fixed distance d0 between the radar and the target, that is, eliminates the influence of different distance bins on the signal, leaving only the required information, that is, the chest movement x(t) caused by breathing and heartbeat. Therefore, the signals of multiple different distance bins can be spliced ​​into a single signal.

[0027] In step 3, the specific operation of the median filter used is:

[0028] Median filter the phase signal with a window of step size m = 5,

[0029]

[0030] where N is the number of phase signals.

[0031] In step 5 described above, the MUSIC algorithm is used for short-time frequency estimation, specifically:

[0032] The radar echo signal is x(t), and its expression is:

[0033]

[0034] In the formula, A k , f k and φ k respectively represent the signal amplitude, frequency, and phase, and K is the number of frequency components (excluding noise). Sample it according to the following matrix

[0035]

[0036] where N represents the number of sampling points of the signal x(t), M represents the order of the sample matrix, which can be freely determined within a certain range and satisfies K < M < N. If M is too large, there will be too few columns in the sampling matrix, reducing the number of observed samples and resulting in inaccurate estimation results. On the contrary, if M is too small, signals with similar frequencies cannot be effectively distinguished, resulting in broadened or even overlapping spectral peaks.

[0037] Calculate the autocorrelation matrix of the sample matrix and decompose its eigenvalues

[0038]

[0039] where H represents the conjugate transpose, D represents the eigenvalue matrix, and E represents the eigenvector matrix.

[0040] Subsequently, the signal spectrum can be calculated

[0041]

[0042] In the formula, a(f) = [1, e -j2πf , …, e -j2πf(M-1) . The noise subspace G is obtained from the eigenvector matrix E according to the number K of signal frequencies, and G takes the K + 1 column to the M column of E.

[0043] In step 5 described above, the generated respiration frequency spectrum is:

[0044]

[0045] In the formula, The noise subspace G1 is obtained from the eigenvector matrix E1 according to the number of respiration signal frequencies K1, and G1 takes the (K1 + 1)-th column to the M1-th column of E1; where H represents the conjugate transpose;

[0046] The eigenvector matrix E1 comes from eigenvalue decomposition:

[0047]

[0048] where D1 represents the eigenvalue matrix, E1 represents the eigenvector matrix, N1 represents the number of sampling points of the respiration signal x1(t), M1 represents the order of the sample matrix, which can be freely determined within a certain range and satisfies K1 < M1 < N1. If M1 is too large, there will be too few columns in the sampling matrix, thus reducing the number of observed samples and resulting in inaccurate estimation results. On the contrary, if M1 is too small, signals with similar frequencies cannot be effectively distinguished, resulting in broadened or even overlapping spectral peaks.

[0049] The matrix X1 comes from sampling the respiration signal:

[0050]

[0051] where the respiration signal is x1(t), and its expression is:

[0052]

[0053] In the formula, and respectively represent the signal amplitude, frequency and phase, and K1 is the number of frequency components (excluding noise);

[0054] In step 5, the generated heartbeat frequency spectrum is:

[0055]

[0056] In the formula, The noise subspace G2 is obtained from the eigenvector matrix E2 according to the number of respiration signal frequencies K2, and G2 takes the (K2 + 1)-th column to the M2-th column of E2; where H represents the conjugate transpose.

[0057] The eigenvector matrix E2 comes from eigenvalue decomposition:

[0058]

[0059] Among them, D2 represents the eigenvalue matrix, E2 represents the eigenvector matrix, N2 represents the number of sampling points of the respiratory signal x2(t), M2 represents the order of the sample matrix, which can be freely determined within a certain range and satisfies K2 < M2 < N2. If M2 is too large, it will result in too few columns in the sampling matrix, thereby reducing the number of observed samples and leading to inaccurate estimation results. On the contrary, if M2 is too small, signals with similar frequencies cannot be effectively distinguished, resulting in broadened or even overlapping spectral peaks.

[0060] The matrix X2 is obtained by sampling the respiratory signal:

[0061]

[0062] Among them, the respiratory signal is x2(t), and its expression is:

[0063]

[0064] In the formula, and respectively represent the signal amplitude, frequency, and phase, and K2 is the number of frequency components (excluding noise);

[0065] In the said step 6, based on the respiratory frequency spectrum, the respiratory harmonic interference is eliminated in the heartbeat frequency spectrum. Specifically, the respiratory fundamental frequency is extracted through peak search in the respiratory frequency spectrum, and then peak search is performed in the heartbeat frequency spectrum. If the peak frequency obtained from the search is equal to or close to a multiple of the respiratory fundamental frequency, then this frequency is discarded, and then the frequency of the second peak is judged until this condition is not met. At this time, the selected frequency is the heartbeat frequency.

[0066] Compared with the prior art, the technical method provided by the present invention has the following beneficial effects:

[0067] The rapid heart rate detection method of millimeter-wave radar for suppressing random body movement interference provided by the present invention only requires one transmitting antenna and one receiving antenna, with a simple structure and low cost; it can accurately select the distance bin where the heart is located under the condition of a measurement time window of 0.05 seconds; it can accurately estimate the heart rate when the signal is interfered by random body movement; it can accurately and quickly give a real-time heart rate estimate under the condition that the measurement time window is only 5 seconds. The method provided by the present invention effectively improves the accuracy and real-time performance of the millimeter-wave radar system for measuring the heart rate under random body movement conditions, and has a low cost and low system complexity. Brief Description of the Drawings

[0068] Figure 1 The figure shows a schematic flow chart of the rapid heart rate detection method for suppressing random body movement interference provided by the present invention;

[0069] Figure 2Schematic diagram of performing range FFT on each received chirp signal before selecting the range bin;

[0070] Figure 3 The time window required for estimating the extreme frequency is spliced ​​with the signal of the time window required for estimating the relatively high frequency;

[0071] Figure 4 is the effect of phase difference on the signal spectrum. DETAILED DESCRIPTION

[0072] To further illustrate the principles, processes, and advantages of the present invention, exemplary embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be noted that the verification methods mainly adopted in the embodiments described below are only used to further explain the present invention and do not represent all implementation methods of the method proposed in the present invention.

[0073] Figure 1 The flowchart of the method for rapid heart rate detection with suppression of random body motion interference provided by the present invention is shown, including:

[0074] Step 1: Collect millimeter-wave radar echo signals, locate the range bin where the human heart is located within a 0.05-second time window using the energy variance method under random body motion interference, and calculate the phase of all echo signals in each range bin within the time window;

[0075] Step 2: Perform phase difference processing on the phase characteristics in each time window to obtain a phase difference signal, and perform time domain splicing on the phase difference signals of adjacent range bins within a 5-second time window based on the range bin splicing method to form an enhanced phase signal;

[0076] Step 3: Using a median filter with a step size of 5 to suppress pulse noise and inter-pulse interference on the enhanced phase signal to obtain a denoised phase signal;

[0077] Step 4, decomposing the denoised phase signal into a respiratory signal component and a heartbeat signal component through a bandpass filter bank, wherein the respiratory signal component is filtered using a 0.1-0.5 Hz bandpass filter, and the heartbeat signal component is filtered using a 0.8-2 Hz bandpass filter;

[0078] Step 5, performing short-time frequency estimation on the respiratory and heartbeat signal components respectively: using the MUSIC algorithm to perform short-time frequency estimation on the respiratory signal component to generate a respiratory frequency spectrum, and using the MUSIC algorithm to perform short-time frequency estimation on the heartbeat signal component to generate a heartbeat frequency spectrum;

[0079] Step 6: Eliminate respiratory harmonic interference in the heart rate spectrum based on the respiratory frequency spectrum, and then extract accurate heart rate characteristic values ​​through peak search.

[0080] Figure 2 The figure shows the preprocessing process of each chirp data before selecting the distance bin where the heart is located in step 1, that is, performing distance FFT on each received chirp signal. The fast time dimension represents 1-N, which represents N different distance bins, and the slow time dimension represents 1-M, which represents M time windows.

[0081] In step 1, the time window of 0.05 seconds used by the range bin is selected for example. The specific time window needs to be determined according to the radar frame interval, which is an integer multiple of the radar frame interval. The radar needs to transmit more than or equal to two chirp signals at equal intervals in each frame.

[0082] In step 1, the energy variance method used to select the distance bin is specifically:

[0083] Calculate the energy variance of all echo signals in each time window for each range bin, and select the range bin with the largest variance as the range bin where the heart is located in that time window. Note that due to random body motion, the range bins selected for each time window may not be consistent.

[0084] Other methods that can accurately select the distance bin where the heart is located can also be used instead, such as the energy mean method, the phase variance method, etc.

[0085] In step 2, the reason why the distance bin splicing method can directly splice signals from different time windows and distance bins is specifically explained as follows:

[0086] The phase signal received by the radar can be specifically expressed as:

[0087]

[0088] Here, R(t) represents the displacement of the signal reflection surface, which can be decomposed into the fixed distance d0 between the radar and the target, the different sizes of d0 represent the different range bins, and the chest movement r(t) caused by breathing and heartbeat.

[0089] After differentiating the phase signal, the signal becomes:

[0090] φ(t)=[φ1,φ2,…,φ M ]

[0091] φ′(t)=[φ2-φ1,φ3-φ2,…,φ M -φ M-1 ]

[0092] It can be seen that signal differencing eliminates the fixed distance d0 between the radar and the target, that is, eliminates the influence of different distance bins on the signal, leaving only the required information, that is, the chest movement x(t) caused by breathing and heartbeat. Therefore, the signals of multiple different distance bins can be spliced ​​into a single signal.

[0093] In step 2, the specific process of signal splicing is as follows Figure 3 as shown

[0094] In the described step 3, the specific operation of the median filter used is as follows:

[0095] Perform median filtering on the phase signal with a window of step size m = 5

[0096]

[0097] where N is the number of phase signals

[0098] In step 3, in addition to being able to eliminate the influence of range bins on the signal, the difference can also suppress respiratory harmonics and at the same time amplify the heartbeat signal, specifically as follows Figure 4 as shown. It can be clearly seen in the figure that the amplitudes of signals with relatively small frequencies in the spectrum of the signal after differentiation (such as respiratory signals and low-order harmonics) are suppressed, and the amplitudes of signals with relatively large frequencies are amplified (such as heartbeat signals).

[0099] In step 5, since the time window used is only 5 s, sufficient spectral resolution cannot be obtained using FFT, so MUSIC is used for short-time frequency estimation

[0100] In the described step 5, the MUSIC algorithm is used for short-time frequency estimation, specifically as follows

[0101] The radar echo signal is x(t), and its expression is

[0102]

[0103] where A k , f k and φ k respectively represent the signal amplitude, frequency and phase, and K is the number of frequency components (excluding noise). Sample it according to the following matrix

[0104]

[0105] where N represents the number of sampling points of the signal x(t), M represents the order of the sample matrix, which can be freely determined within a certain range, satisfying K < M < N. If M is too large, there will be too few columns in the sampling matrix, resulting in a reduction in the number of observed samples and inaccurate estimation results. On the contrary, if M is too small, signals with similar frequencies cannot be effectively distinguished, resulting in broadened or even overlapping spectral peaks

[0106] Calculate the autocorrelation matrix of the sample matrix and decompose its eigenvalues

[0107]

[0108] Where H represents the conjugate transpose, D represents the eigenvalue matrix, and E represents the eigenvector matrix.

[0109] The signal spectrum can then be calculated

[0110]

[0111] Where a(f)=[1,e -j2πf ,…,e -j2πf(M-1) ]. The noise subspace G is obtained from the eigenvector matrix E according to the number of signal frequencies K, and G takes the K+1 column to the M column of E.

[0112] In step 6, based on the respiratory frequency spectrum, respiratory harmonic interference is eliminated in the heart frequency spectrum. Specifically, the respiratory fundamental frequency is extracted by peak search in the respiratory spectrum, and then a peak search is performed in the heart frequency spectrum. If the peak frequency obtained by the search is equal to or close to a multiple of the respiratory fundamental frequency, the frequency is discarded, and the frequency of the second peak is judged again until the condition is no longer met. At this time, the selected frequency is the heart frequency.

[0113] The validation indicator used to verify the heart rate results of step 6 is the mean absolute percentage error (MAPE).

[0114] MAPE is calculated as:

[0115]

[0116] The MAPE calculated in this embodiment is the average value of the real-time heart rate of the 60-s signal;

[0117] The actual heart rate is measured using an ECG sensor.

[0118] Example

[0119] Based on the above method, the specific implementation process S1-S6 was carried out on 5 subjects in the presence and absence of random body movement. In each case, each subject measured 5 sets of data, each set of data lasting 1 minute.

[0120] Table 1 MAPE of 25 sets of experimental data:

[0121] Data Number MAPE (%) under random body motion MAPE (%) without random body motion 1 2.85 5.79 2 2.71 2.95 3 1.16 1.67 4 1.29 3.42 5 3.16 2.67 6 2.06 2.06 7 2.86 5.09 8 1.81 1.95 9 1.69 4.85 10 0.98 3.60 11 1.72 3.20 12 4.46 4.21 13 2.14 2.94 14 1.96 6.14 15 4.37 2.49 16 0.42 4.58 17 8.10 4.95 18 2.19 2.41 19 6.21 1.68 20 5.90 4.11 21 4.61 4.20 22 2.80 5.31 23 5.08 4.63 24 1.24 1.51 25 3.71 1.96 Average results 3.02 3.54

[0122] As can be seen from the results, the average error with random body motion is 3.02%, and the average error without random body motion is 3.54%, which shows the effectiveness of this method.

[0123] This embodiment provides a method for detecting heart rate during body movement based on millimeter-wave radar, which can quickly and accurately detect heart rate in real time even in the presence of random body movement. Compared with existing technologies, this method requires fewer measurements, greatly improves accuracy, is low-cost, and is easy to implement.

[0124] It should be noted that, in the description of the present invention, terms such as "first" and "second" are used for illustrative purposes only and do not indicate or imply relative importance. In addition, in the description of the present application, unless otherwise explicitly stated, the term "plurality" refers to at least two.

[0125] Any process or method described in any flowchart or other manner described in the present invention can be understood as a code module, segment, or portion including one or more executable instructions for performing steps of a specific logical function or process. The scope of the preferred embodiments of the present invention also includes other implementations, which may not be performed in the order shown, as understood by those skilled in the art to which the present invention relates.

[0126] It should be understood that when describing the present invention, terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" refer to at least one embodiment or example having specific features, structures, materials, or characteristics. The use of these terms does not necessarily refer to the same embodiment or example, and the specific features, structures, materials, or characteristics described in the description may be appropriately combined in one or more embodiments or examples.

[0127] The above specific embodiments further describe the purpose, technical methods and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent replacements or improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

[0128] In summary, the above are only preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for detecting heart rate during body movement based on millimeter wave radar, characterized in that The steps of the method include: Step 1: Real-time acquisition of millimeter-wave radar echo signals. Under random body motion interference, the distance bin where the human heart is located is selected every other time window, and then the phase of all echo signals of each distance bin within the time window is calculated. Step 2: Perform phase difference processing on the phases of all echo signals in the time window of each range bin obtained in step 1 to obtain the phase difference signal of the corresponding range bin, and perform time domain splicing on the phase difference signals of adjacent range bins in the time window to form an enhanced phase signal; Step 3, performing median filtering on the enhanced phase signal obtained in step 2 to obtain a denoised phase signal; Step 4, decomposing the denoised phase signal obtained in step 3 into a respiratory signal component and a heartbeat signal component; Step 5, performing short-time frequency estimation on the respiratory signal component obtained in step 4 to generate a respiratory frequency spectrum, and performing short-time frequency estimation on the heartbeat signal component obtained in step 4 to generate a heartbeat frequency spectrum; Step 6: Eliminate respiratory harmonic interference from the heartbeat frequency spectrum generated in step 5 according to the respiratory frequency spectrum generated in step 5, and then extract accurate heartbeat frequency characteristic values ​​through peak search.

2. The method for detecting heart rate during body movement based on millimeter-wave radar according to claim 1, characterized in that: In step 1, the time window is an integer multiple of the radar frame interval, and the radar transmits greater than or equal to 2 chirp signals at equal intervals in each frame.

3. The method for detecting heart rate during body movement based on millimeter-wave radar according to claim 2, characterized in that: The time window is 0.05s.

4. The method for detecting heart rate during body movement based on millimeter-wave radar according to claim 1, characterized in that: In step 1, the range bin is selected using the energy variance method, the energy mean method or the phase variance method.

5. The method for detecting heart rate during body movement based on millimeter-wave radar according to claim 4, characterized in that: The specific method of selecting distance bins using the energy variance method is: Calculate the energy variance of all echo signals in each range bin within each time window, and select the range bin with the largest variance value as the range bin where the heart is located in that time window.

6. The method for detecting heart rate during body movement based on millimeter-wave radar according to claim 1, characterized in that: In step 3, the denoised phase signal is: f″ i =Med{φ′ i-v ,…,φ′ i ,…,φ′ i+v } Among them, φ″ i is the phase value of the i-th denoised phase signal, N is the number of phase signals, m is a natural number, m is the window step size, φ′ i is the phase value of the i-th phase difference signal.

7. The method for detecting heart rate during body movement based on millimeter-wave radar according to claim 1, characterized in that: In step 4, the denoised phase signal is decomposed into a respiratory signal component and a heartbeat signal component by a bandpass filter bank. The respiratory signal component is filtered by a 0.1-0.5 Hz bandpass filter, and the heartbeat signal component is filtered by a 0.8-2 Hz bandpass filter.

8. The method for detecting heart rate during body movement based on millimeter-wave radar according to claim 1, characterized in that: In step 5, the respiratory frequency spectrum generated is: Where, The noise subspace G1 is obtained from the eigenvector matrix E1 according to the number of respiratory signal frequencies K1. G1 takes the K1+1 column to the M1 column of E1; where H represents the conjugate transpose; The eigenvector matrix E1 comes from the eigenvalue decomposition: Among them, D1 represents the eigenvalue matrix, E1 represents the eigenvector matrix, N1 represents the number of sampling points of the respiratory signal x1(t), M1 represents the order of the sample matrix, K1 <M1<N1; The matrix X1 comes from sampling the respiratory signal: Among them, the breathing signal is x1(t), and its expression is: Where, and represent the signal amplitude, frequency and phase respectively, and K1 is the number of frequency components.

9. The method for detecting heart rate during body movement based on millimeter-wave radar according to claim 1, characterized in that: In step 5, the heart rate spectrum generated is: Where, The noise subspace G2 is obtained from the eigenvector matrix E2 according to the number of respiratory signal frequencies K2. G2 takes the K2+1 column to the M2 column of E2; where H represents the conjugate transpose; The eigenvector matrix E2 comes from the eigenvalue decomposition: Among them, D2 represents the eigenvalue matrix, E2 represents the eigenvector matrix, N2 represents the number of sampling points of the respiratory signal x2(t), M2 represents the order of the sample matrix, K2 <M2<N2; The matrix X2 comes from sampling the respiratory signal: Among them, the breathing signal is x2(t), and its expression is: Where, and represent the signal amplitude, frequency and phase respectively, and K2 is the number of frequency components.

10. The method for detecting heart rate during body movement based on millimeter-wave radar according to claim 1, characterized in that: In step 6, the respiratory fundamental frequency is extracted by peak search in the respiratory frequency spectrum, and then a peak search is performed in the heart rate spectrum. If the peak frequency obtained by the search is equal to or close to a multiple of the respiratory fundamental frequency, the frequency is discarded, and the frequency of the second peak is judged again until the condition is no longer met. At this time, the selected frequency is the heart rate.

Images