Heartbeat waveform reconstruction method based on MIMO millimeter wave radar signal

Abstract

The invention designs a heartbeat waveform reconstruction method based on MIMO millimeter wave radar signals. Firstly, a phase difference signal corresponding to the estimated heart rate is filtered through a band-pass filter. Then, the filtered phase difference signal is segmented through trough detection. Then, a heartbeat waveform template is constructed and the template is resampled on the basis of the segmentation. And finally, finely adjusting the starting point and the ending point of each segment to obtain a reconstructed heartbeat waveform. According to the method, the problem that the heartbeat waveform reconstruction error is large when the target is far away from the radar and deviates from the radar is solved, and the accuracy of the heartbeat waveform reconstruction based on the radar signal is improved.

Description

Technical Field

[0001] The present invention relates to the fields of millimeter wave radar, wireless sensor and heartbeat monitoring. Background Art

[0002] As one of the most important life indicators, heartbeat not only reflects a person's physical condition, but also reflects a person's psychological changes. Traditionally, continuous heartbeat monitoring requires contact sensors such as electrocardiograms or pulse oximeters. Although these sensors can obtain very accurate heartbeat waveforms, they bring great inconvenience to humans, especially those with burns and infections. In order to overcome the shortcomings of contact sensing and realize ubiquitous heartbeat monitoring, many non-contact heartbeat monitoring systems have been proposed in recent years.

[0003] Non-contact heartbeat monitoring can be based on different signals, such as sound, visible light, infrared, electromagnetic waves, etc. Compared with other types of signals, electromagnetic waves have obvious advantages in privacy protection and environmental robustness. Various systems use electromagnetic waves to detect heartbeats, such as WiFi, ultra-wideband radar, and millimeter-wave radar. Among them, millimeter-wave radar has the characteristics of the shortest wavelength, the largest bandwidth, and the most granular perception ability, which can achieve the best heartbeat monitoring accuracy.

[0004] In recent years, many heartbeat monitoring systems based on mm-wave radar have been proposed. For example, Hajar Abedi et al. from a Canadian public university used a sequential deep learning network to restore the heartbeat waveform in a paper titled Non-Visual and Contactless Wellness Monitoring for Long Term Care Facilities Using mm-Wave Radar Sensors published in IEEE Sensors. Li Tao et al. from Beijing University of Posts and Telecommunications reconstructed the heartbeat signal by simultaneously optimizing the decomposition of all extracted composite vital signs signals in different ranges and angles in a paper titled A Novel Heart Rate Estimation Method Exploiting Heartbeat Second Harmonic Reconstruction Via Millimeter Wave Radar published in IEEE Internet of Things Journal. Xiangyu Xu et al. from Shanghai Jiao Tong University used a method based on hierarchical variational mode decomposition (VMD) to extract heartbeat signals from mm-wave radar data in a paper titled mmecg: Monitoring human cardiac cycle in driving environments leveraging millimeter wave published in IEEE Conference on Computer Communications.

[0005] However, when the research object is far away from the radar and deviates from the radar, the above method will produce a large error in reconstructing the heartbeat waveform. In actual situations, the position of the research object often changes, and in most cases it is far away from the radar and deviates from the radar. The purpose of the present invention is to reconstruct a more accurate heartbeat waveform in such a situation where the human body is far away from the radar and deviates from the radar. Summary of the invention

[0006] When the research object is far away from and deviates from the millimeter-wave radar, the existing methods cannot guarantee the accuracy of the reconstruction of the heartbeat waveform. In order to reconstruct an accurate heartbeat waveform at a long distance and a large deviation angle, the present invention further proposes a heartbeat waveform reconstruction method based on a MIMO millimeter-wave radar signal on the basis of the heart rate estimation method MMHeart based on MIMO millimeter-wave radar. The method designed by the present invention first filters the phase difference signal corresponding to the estimated heart rate through a bandpass filter. Then, the filtered phase difference signal is segmented by trough detection. Next, a heartbeat waveform template is constructed and the template is resampled based on the segmentation. Finally, the starting point and end point of each segment are fine-tuned to obtain a reconstructed heartbeat waveform. The specific technical scheme is as follows:

[0007] The present invention is a heartbeat waveform reconstruction method based on MIMO millimeter wave radar signal, which specifically comprises the following steps:

[0008] Step 1: Filter the phase difference signal corresponding to the estimated heart rate through a bandpass filter;

[0009] Step 2: Segment the filtered phase difference signal through trough detection;

[0010] Step 3: Construct a heartbeat waveform template and resample the template based on the segmentation;

[0011] Step 4: Fine-tune the starting point and ending point of each segment to obtain a reconstructed heartbeat waveform.

[0012] In step 1, the phase difference signal corresponding to the estimated heart rate is filtered by a bandpass filter, specifically:

[0013] Step 1-1: Get the estimated heart rate f i,j,n Corresponding phase difference signal, where i,j represents the channel (T i ,R j ), n is the channel (T i ,R j ) in the nth distance bin, for each channel (T i ,R j ), T i represents the i-th transmitting antenna, R j represents the jth receiving antenna, (T i ,R j ) represents the channel between the i-th transmitting antenna and the j-th receiving antenna;

[0014] Step 1-2: Use a bandpass filter to filter f i,j,n The corresponding phase difference signal is filtered to obtain a purer signal, that is, to obtain the estimated heart rate f i,j,n Nearby signals with smaller frequency ranges.

[0015] Furthermore, in step 2, the filtered phase difference signal is segmented by valley detection, specifically:

[0016] Step 2-1: Find all the troughs in the filtered phase difference signal. Each trough can be regarded as the starting point of the current segment and the end point of the previous segment.

[0017] Step 2-2: Calculate the time difference between adjacent troughs in sequence. If the time difference between two adjacent troughs is less than the preset threshold Δt v , then the next trough is considered to be a false trough and is removed from the trough set.

[0018] Step 2-3: Based on the remaining troughs, a segment set S = {s1, s2, ..., s G}, where s1, s2 and s G They represent the filtered phase difference signals of the first, second and Gth segments respectively, and the length of each segment is {|s1|, |s2|, ..., |s G |}, where |s g |=l g -l g-1 , l g-1 and l g are the start time and end time of the g-th segment, respectively, where 1≤g≤G, and G is the number of all segments.

[0019] Further, in step 3, a heartbeat waveform template is constructed and the template is resampled based on the segmentation, specifically:

[0020] Step 3-1: Use a pulse oximeter to measure a real human heartbeat waveform, and capture one cycle of the waveform as a heartbeat waveform template μ;

[0021] Step 3-2: For each segment of S, resample at the resampling rate |s g | / |μ|Resample μ to get μ g , that is, the gth segment of the reconstructed heartbeat waveform, can be expressed as μ g =ω(μ,|s g |), where ω(μ,|s g |) means resampling the template μ to segment s g Length, 1≤g≤G.

[0022] Furthermore, in step 4, the starting point and the ending point of each segment are fine-tuned to obtain a reconstructed heartbeat waveform, specifically:

[0023] Step 4-1: Based on trough detection, only the rough starting and ending points of each cardiac cycle can be obtained, so that the reconstructed heartbeat waveform is not aligned with the filtered phase difference signal, such asFigure 2 As shown. In order to align the reconstructed heartbeat waveform with the filtered phase difference signal, we fine-tune the starting and ending points of each segment in S. Fine-tuning can be constructed as an optimization problem, which can be expressed as Among them l g Indicates g+1 The starting point and s g The end point, t peak (μ g )-t peak (s g ) represents μ g With s g The time difference at the peak;

[0024] Step 4-2: In each iteration, the reconstructed heartbeat waveform is updated according to the latest segmentation, and then the time difference between the peak value of the reconstructed heartbeat waveform and the filtered phase difference signal is updated. After multiple iterations, the reconstructed heartbeat waveform is finally obtained, such as Figure 3 shown.

[0025] The beneficial effect of the present invention is that the present invention makes full use of radar data of multiple channels of MIMO millimeter waves, and solves the problem of large error in reconstructing the heartbeat waveform when the research object is far away and deviates from the radar with lower complexity. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 Schematic diagram of a heartbeat waveform reconstruction method based on MIMO millimeter-wave radar signals.

[0027] Figure 2 Schematic diagram of the initial reconstructed heartbeat waveform and the filtered phase difference signal.

[0028] Figure 3 Schematic diagram of the final reconstructed heartbeat waveform and the filtered phase difference signal.

[0029] Figure 4 Schematic diagram of the accuracy of the cardiac cycle duration of the mmEGC of the present invention and the system at different distances.

[0030] Figure 5 Schematic diagram of the accuracy of the cardiac cycle duration of the mmEGC of the present invention and the system at different angles. DETAILED DESCRIPTION

[0031] The following will disclose the embodiments of the present invention with drawings. For the purpose of clear description, many practical details will be described together in the following description. However, it should be understood that these practical details should not be used to limit the present invention. That is to say, in some embodiments of the present invention, these practical details are not necessary.

[0032] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings. Figure 1 After MIMO millimeter waves are transmitted to the human body and receive reflected signals, radar data of multiple channels are generated. For all channels, the present invention first filters the phase difference signal corresponding to the estimated heart rate through a bandpass filter. Then, the filtered phase difference signal is segmented by trough detection. Next, a heartbeat waveform template is constructed and the template is resampled based on the segmentation. Finally, the starting point and end point of each segment are fine-tuned to obtain a reconstructed heartbeat waveform. The measurement method of the present invention specifically includes the following steps:

[0033] Step 1: Filter the phase difference signal corresponding to the estimated heart rate through a bandpass filter. The specific steps are:

[0034] Step 1-1: Get the estimated heart rate f i,j,n Corresponding phase difference signal, where i,j represents the channel (T i ,R j ), n is the channel (T i ,R j ) in the nth distance bin, for each channel (T i ,R j ), T i represents the i-th transmitting antenna, R j represents the jth receiving antenna, (T i ,R j ) represents the channel between the i-th transmitting antenna and the j-th receiving antenna;

[0035] Step 1-2: Use a bandpass filter to filter f i,j,n The corresponding phase difference signal is filtered to obtain a purer signal, that is, to obtain the estimated heart rate f i,j,n Nearby signals with smaller frequency ranges.

[0036] Step 2: Segment the filtered phase difference signal through valley detection. The specific steps are as follows:

[0037] Step 2-1: Find all the troughs in the filtered phase difference signal. Each trough can be regarded as the starting point of the current segment and the end point of the previous segment.

[0038] Step 2-2: Calculate the time difference between adjacent troughs in sequence. If the time difference between two adjacent troughs is less than the preset threshold Δt v , then the next trough is considered to be a false trough and is removed from the trough set.

[0039] Step 2-3: Based on the remaining troughs, a segment set S = {s1, s2, ..., s G}, where s1, s2 and s GThey represent the filtered phase difference signals of the first, second and Gth segments respectively, and the length of each segment is {|s1|, |s2|, ..., |s G |}, where |s g |=l g -l g-1 , l g-1 and l g are the start time and end time of the g-th segment, respectively, where 1≤g≤G, and G is the number of all segments.

[0040] Step 3: Construct a heartbeat waveform template and resample the template based on the segmentation. The specific steps are as follows:

[0041] Step 3-1: Use a pulse oximeter to measure a real human heartbeat waveform, and capture one cycle of the waveform as a heartbeat waveform template μ;

[0042] Step 3-2: For each segment of S, resample at the resampling rate |s g | / |μ|Resample μ to get μ g , that is, the gth segment of the reconstructed heartbeat waveform, can be expressed as μ g =ω(μ,|s g |), where ω(μ,|s g |) means resampling the template μ to segment s g Length, 1≤g≤G.

[0043] Step 4: Fine-tune the starting point and ending point of each segment to obtain the reconstructed heartbeat waveform. The specific steps are as follows:

[0044] Step 4-1: Based on trough detection, only the rough starting and ending points of each cardiac cycle can be obtained, so that the reconstructed heartbeat waveform is not aligned with the filtered phase difference signal, such as Figure 2 As shown. In order to align the reconstructed heartbeat waveform with the filtered phase difference signal, we fine-tune the starting and ending points of each segment in S. Fine-tuning can be constructed as an optimization problem, which can be expressed as Among them l g Indicates g+1 The starting point and s g The end point, t peak (μ g )-t peak (s g ) represents μ g With s g The time difference at the peak;

[0045] Step 4-2: In each iteration, the reconstructed heartbeat waveform is updated according to the latest segmentation, and then the time difference between the peak value of the reconstructed heartbeat waveform and the filtered phase difference signal is updated. After multiple iterations, the reconstructed heartbeat waveform is finally obtained, such as Figure 3 shown.

[0046] In order to verify the effect of the present invention, the inventors conducted two sets of experiments. The present invention uses the MIMO millimeter wave radar developed by TI as the sensing front end, and uses a laptop computer to process data in the background. The radar model used is IWR6843, which has 3 transmitting antennas and 4 receiving antennas.

[0047] The environment in which the present invention collects data is an indoor environment. During the experiment, the distance between the experimental subject and the radar increased from 1m to 4m. At each distance, the chest of the experimental subject was basically at the same height as the radar. The deviation angle between the experimental subject and the radar gradually increased from 0° to 60°. While measuring the radar data, the experimental subject wore a pulse oximeter to obtain the real heartbeat waveform of the human body to further verify the accuracy of the system designed by the present invention. In order to evaluate the performance of the heartbeat waveform reconstruction, the reconstruction results of the present invention were compared with the reconstruction results of mmEGC. The evaluation index is the heart cycle duration error, which is calculated as the average value of the difference between the respective heart cycle durations in the reconstructed heartbeat waveform and the real heartbeat waveform.

[0048] The first set of experiments: In order to evaluate whether the present invention can effectively improve the accuracy of cardiac cycle duration at different distances, the present invention faces the target directly to the radar and then compares it with the latest system mmEGC at different distances. The experimental results are as follows Figure 4 As shown (the estimated results under each distance setting are the average values ​​under the corresponding distance).

[0049] The second set of experiments: In order to evaluate whether the present invention can effectively improve the accuracy of cardiac cycle duration in different orientations, the present invention fixed the distance between the object to be measured and the radar at 4 meters, and then compared it with the latest system mmEGC in different orientations. The experimental results are as follows Figure 5 As shown (the estimated results under each angle setting are the average values ​​under the corresponding angle).

[0050] The above description is only an embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A heartbeat waveform reconstruction method based on MIMO millimeter wave radar signal, characterized in that: The heartbeat waveform reconstruction method comprises the following steps: Step 1: Filter the phase difference signal corresponding to the estimated heart rate through a bandpass filter; Step 2: Segment the filtered phase difference signal through trough detection; Step 3: Construct a heartbeat waveform template and resample the template based on the segmentation; Step 4: Fine-tune the starting point and ending point of each segment to obtain a reconstructed heartbeat waveform.

2. A heartbeat waveform reconstruction method based on MIMO millimeter wave radar signal according to claim 1, characterized in that: The step 1 specifically includes the following steps: Step 1-1: Get the estimated heart rate f i,j,n Corresponding phase difference signal, where i,j represents the channel (T i ,R j ), n is the channel (T i ,R j ) in the nth distance bin, for each channel (T i ,R j ), T i represents the i-th transmitting antenna, R j represents the jth receiving antenna, (T i ,R j ) represents the channel between the i-th transmitting antenna and the j-th receiving antenna; Step 1-2: Use a bandpass filter to filter f i,j,n The corresponding phase difference signal is filtered to obtain a purer signal, that is, to obtain the estimated heart rate f i,j,n Nearby signals with smaller frequency ranges.

3. The heartbeat waveform reconstruction method based on MIMO millimeter wave radar signal according to claim 1, characterized in that: The step 2 is specifically as follows: Step 2-1: Find all the troughs in the filtered phase difference signal. Each trough can be regarded as the starting point of the current segment and the end point of the previous segment. Step 2-2: Calculate the time difference between adjacent troughs in sequence. If the time difference between two adjacent troughs is less than the preset threshold Δt v , then the next trough is considered a false trough and is removed from the trough set. Step 2-3: Based on the remaining troughs, a segment set S = {s1, s2, ..., s G }, where s1, s2 and s G They represent the filtered phase difference signals of the first, second and Gth segments respectively, and the length of each segment is {|s1|, |s2|, ..., |s G |}, where |s g |=l g -l g-1 , l g-1 and l g are the start time and end time of the g-th segment, respectively, where 1≤g≤G, and G is the number of all segments.

4. The heartbeat waveform reconstruction method based on MIMO millimeter wave radar signal according to claim 1, characterized in that: The step 3 specifically includes the following steps: Step 3-1: Use a pulse oximeter to measure a real human heartbeat waveform, and capture one cycle of the waveform as a heartbeat waveform template μ; Step 3-2: For each segment of S, resample at the resampling rate |s g | / |μ|Resample μ to get μ g , that is, the gth segment of the reconstructed heartbeat waveform, can be expressed as μ g =ω(μ,|s g |), where ω(μ,|s g |) means resampling the template μ to segment s g Length, 1≤g≤G.

5. The heartbeat waveform reconstruction method based on MIMO millimeter wave radar signal according to claim 1, characterized in that: The step 4 specifically includes the following steps: Step 4-1: Based on trough detection, only the rough start and end points of each cardiac cycle can be obtained, which makes the reconstructed heartbeat waveform misaligned with the filtered phase difference signal. In order to align the reconstructed heartbeat waveform with the filtered phase difference signal, we fine-tune the start and end points of each segment in S. Fine-tuning can be constructed as an optimization problem, which can be expressed as Among them l g Indicates g+1 The starting point and s g The end point, t peak (μ g )-t peak (s g ) represents μ g With s g The time difference at the peak; Step 4-2: In each iteration, the reconstructed heartbeat waveform is updated according to the latest segmentation, and then the time difference between the peak value of the reconstructed heartbeat waveform and the filtered phase difference signal is updated. After multiple iterations, the reconstructed heartbeat waveform is finally obtained.

Images