A contactless ECG signal monitoring method based on millimeter wave radar

Abstract

The application discloses a kind of non-contact ECG signal monitoring methods based on millimeter wave radar, belong to ECG signal monitoring field, through simultaneously sensing the chest vibration caused by heartbeat and carotid artery pulsation from neck, high-precision electrocardiogram reconstruction is realized by fusing space-time characteristics, solve the problem that traditional scheme is dependent on electrode and existing radar technology cannot restore complete ECG.This application carries out non-contact ECG signal monitoring in daily scene, with good privacy and comfort.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of millimeter wave radar vital sign monitoring, and particularly relates to a non-contact electrocardiogram (ECG) signal monitoring method based on a millimeter wave radar. BACKGROUND

[0002] Cardiovascular disease is the leading cause of death worldwide, and electrocardiogram (ECG) is the most commonly used means for diagnosing cardiovascular disease. Traditional ECG monitoring mainly relies on wearable devices, which has the problems of limited battery life and electrode falling off, affecting the continuity of monitoring and the reliability of data. In the non-contact ECG monitoring technology, using ultra-wideband and millimeter wave radar are two main schemes. However, ultra-wideband has the problems of high power consumption and being easily disturbed by multipath interference. Millimeter wave radar has good privacy, clothing penetration and non-absorption by the skin, and is low in cost, suitable for ECG monitoring in daily life.

[0003] Existing millimeter wave radar schemes can achieve non-contact heartbeat monitoring by detecting the subtle vibration of the human body surface, but existing schemes can only extract heart rate or part of the heartbeat features (such as for identity authentication), and cannot completely restore the ECG waveform, limiting its application in medical-level monitoring. Therefore, it is of great significance to study how to capture effective heartbeat signals and realize high-fidelity ECG waveform reconstruction in the field of vital sign monitoring. SUMMARY

[0004] In order to overcome the shortcomings of the prior art, the application provides a non-contact ECG signal monitoring method based on a millimeter wave radar, which simultaneously perceives the chest vibration caused by heartbeat and the carotid artery pulsation from the neck, and realizes high-precision electrocardiogram reconstruction by fusing spatial and temporal features, solving the problems of electrode dependence of traditional schemes and inability of existing radar technology to restore complete ECG.

[0005] The technical scheme adopted by the application to solve its technical problems is:

[0006] A non-contact ECG signal monitoring method based on a millimeter wave radar, comprising the following steps:

[0007] S1, the millimeter wave radar emits a linear frequency modulation continuous wave, and receives a human body reflection signal, generates an intermediate frequency signal through a frequency mixer, and performs a distance dimension Fourier transform on the intermediate frequency signal to construct a two-dimensional data matrix;

[0008] S2, by analyzing the distribution of the respiratory band energy of the two-dimensional data matrix at different distance bins, static and dynamic interference elimination is performed on the matrix, and the energy intensity of the interference-eliminated matrix is used to locate the position of the human thorax;

[0009] S3, based on the relative spatial relationship between the chest heart and the neck, further positioning the chest heart and the neck region, using the second-order difference method to eliminate the interference of body and respiratory activity on the target region, obtaining a candidate signal set of chest heartbeat and carotid artery pulse;

[0010] S4, using a variational mode decomposition segmentation algorithm based on the candidate signal, obtaining a series of heartbeat period signals, using a dynamic time planning based evaluation algorithm based on the similarity between the period signals, screening out high-quality chest heartbeat signals and carotid artery pulse signals;

[0011] S5, using the equal frequency binning method to process electrocardiogram data, balancing the amplitude distribution of ECG waveform, combining the screened chest heartbeat signal and carotid artery pulse signal with the synchronous collected electrocardiogram signal, and constructing a time-dependent data set through a sliding window sampling method;

[0012] S6, input the generated data set into the ECG prediction model, and the model captures the periodicity and waveform details of the heartbeat signal by extracting the time and space features between the signals, and predicts the ECG value.

[0013] Further, in the step S1, the millimeter wave radar is placed in front of the person to be detected, the related parameters of the transmitted linear frequency modulation electromagnetic wave are set, and the mixer generates an intermediate frequency signal after mixing the transmitted and received signals. The intermediate frequency signal data is stored according to the receiving time sequence, and the dimension is , wherein represents the number of sampling points of a single chirp, represents the number of chirps contained in each frame of data, represents the number of antennas receiving signals, and is the length of each frame of data; the data received by a single antenna is subjected to fast Fourier transform in the distance dimension to generate a two-dimensional matrix containing signal amplitude and phase information, and the dimension is .

[0014] Further, in the step S2, the two-dimensional data matrix is extracted , the phase signals of different distance bins are subjected to unwrapping and then pass through a classic respiratory frequency band pass filter to obtain a sub-band signal; the instantaneous energy of the sub-band signal is calculated, and is reorganized into a respiratory energy matrix , the dimension of which is the same as that of the original matrix , the distance dimension of the matrix is normalized to obtain ; then and the original matrix are subjected to Hadamard product operation to obtain a matrix , and The energy extreme point of the matrix in the distance dimension is used to locate the center position of the human thoracic cavity.

[0015] Further, in the step S3, the radiation diameter of the thoracic cavity region is about 15 to 25 cm, the carotid artery region is located 15 to 25 cm above the thoracic cavity, and the radiation range is about 8 to 12 cm; combined with the distance resolution of the millimeter wave radar of about 4 cm, 5 continuous distance bins in the thoracic cavity region and 3 distance bins near the carotid artery region are selected as the target signal extraction region; the phase signals of each distance bin are subjected to second-order difference processing to suppress the interference of the body and breathing, and a plurality of heartbeat candidate signals are obtained, which respectively constitute the candidate signal set of the thoracic cavity heartbeat and the carotid artery pulse.

[0016] In the step S4, the phase signal is decomposed and the main frequency signal of the heartbeat is extracted by using a variational mode decomposition algorithm; a peak value detection algorithm is performed based on an amplitude threshold value, the heartbeat period is segmented according to the position of the wave peak, and a series of period segments are obtained; the similarity between the period segments is calculated by using dynamic time warping (DTW), hierarchical clustering is performed by using the average linkage method, the optimal clustering number is determined according to the elbow rule, the cluster center with the largest number of samples is selected as the template signal of the candidate signal, and the average DTW value of the template signal and all period segments is calculated as a score P, and the candidate signal set is screened to obtain the thoracic cavity heartbeat signal and the carotid artery signal with the lowest score as high-quality signals.

[0017] In the step S5, in order to avoid model training bias, the ECG signal collected synchronously is preprocessed by using an equal-frequency binning method, the sampling points of the original ECG signal are arranged in ascending order of voltage value, and are divided into a plurality of bins to ensure that each bin contains the same number of data points, the ECG data in the bin is normalized to ensure that it is distributed in the interval [0, 1] while retaining the trend of the original waveform form, the thoracic cavity heartbeat signal, the carotid artery pulse signal and the ECG signal screened are aligned in time and sampling rate, and are stored in a joint data matrix with a size of 3*n according to the time stamp, where n is the length of the signal; the matrix is segmented by using a sliding window method, and the ECG voltage value data of the next time step under the current window are used as labels to construct a data set.

[0018] In the step S6, the ECG prediction model adopts a parallel network architecture. A time convolution network combined with a channel attention mechanism (TCN-SE) extracts spatial features, dynamically allocates the weights of the thoracic cavity heartbeat signal and the carotid artery pulse signal, a bidirectional gated recurrent unit combined with global attention (BiGRU-GA) extracts time features, and focuses on capturing the periodicity of the P-QRS-T wave group in the electrocardiogram.

[0019] The beneficial effects of the present application mainly include:

[0020] 1. Non-contact ECG signal monitoring can be performed in daily scenarios, with good privacy and comfort;

[0021] 2. The interference elimination based on the energy of the respiratory frequency band can effectively suppress the influence of dynamic and static interference in the environment, thereby accurately positioning the human body;

[0022] 3. The segmentation method based on variational mode decomposition and the evaluation method based on dynamic time planning can realize the extraction of high-quality thoracic heartbeat signals and carotid artery pulse signals;

[0023] 4. The spatiotemporal fusion model based on TCN and BiGRU effectively captures and reconstructs the local details and global periodicity of the ECG signal through the channel attention mechanism and the global attention mechanism. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 The flowchart of the non-contact ECG signal monitoring method based on the millimeter wave radar of the present application;

[0025] Figure 2 The network structure diagram of the spatiotemporal feature fusion of the present application. DETAILED DESCRIPTION

[0026] The present application will be further described below with reference to the accompanying drawings.

[0027] Reference Figure 1 and Figure 2 A non-contact ECG signal monitoring method based on a millimeter wave radar, comprising the following steps:

[0028] S1, the millimeter wave radar emits a linear frequency continuous wave, and receives a human body reflection signal, generates an intermediate frequency signal through a mixer, and performs a distance dimension Fourier transform to construct a two-dimensional data matrix;

[0029] The process of step S1 is as follows:

[0030] The millimeter wave radar is placed in front of the person to be detected, the relevant parameters of the emitted linear frequency electromagnetic wave are set, and the mixer generates an intermediate frequency signal after mixing the transmitted and received signals:

[0031] ;

[0032] wherein, represents the amplitude of the signal, , , represents the bandwidth, scanning period and wavelength of the transmitted signal, represents the speed of the electromagnetic wave, represents the distance from the object to the radar;

[0033] The intermediate frequency signal data is stored according to the receiving time sequence, and the dimension is wherein represents the number of sampling points of a single chirp, represents the number of chirps contained in each frame of data, represents the number of antennas of the received signal, and simultaneously, is the length of each frame of data;

[0034] The data received by a single antenna is subjected to a fast Fourier transform in the distance dimension to generate a two-dimensional matrix containing signal amplitude and phase information.

[0035] S2, by analyzing the distribution of the respiratory band energy of the two-dimensional data matrix at different distance bins, static and dynamic interference is eliminated from the matrix, and the energy intensity of the matrix after interference elimination is used to locate the position of the human thorax;

[0036] The process of the step S2 is as follows:

[0037] First, the two-dimensional data matrix is extracted The phase signals of the intermediate frequency signals in different distance bins are subjected to phase unwrapping to restore the phase signals. The phase signals are subjected to a band-pass filter of 0.1-0.5 Hz to obtain corresponding sub-band signals. The instantaneous energy of the sub-band signals is calculated, and the respiratory energy matrix is reorganized The distance dimension of the matrix is normalized to obtain . Subsequently , the Hadamard product operation is performed with the original matrix to obtain the matrix ( = ⊙ ). The energy extreme point of the matrix in the distance dimension is searched to locate the center position of the human thorax.

[0038] S3, based on the relative spatial relationship between the thoracic heart and the neck, the thoracic heart and the neck region are further located. The method of second-order difference is used to eliminate the interference of the body and the respiratory activity in the target region to obtain a candidate signal set of the thoracic heart and the carotid artery pulse.

[0039] The process of the step S3 is as follows:

[0040] ​To obtain a more comprehensive heartbeat signal from the target area, this example requires extracting all signals within that area. Since the radiation diameter of the thoracic cavity region is approximately 15 to 25 centimeters, and the carotid artery region is located 15 to 25 centimeters above the thoracic cavity, with a radiation range of approximately 8 to 12 centimeters, and considering the millimeter-wave radar's range resolution of approximately 4 centimeters, five consecutive range chambers within the thoracic cavity region (covering an area of ​​approximately 20 centimeters) and three range chambers near the carotid artery region (covering an area of ​​approximately 12 centimeters) are selected as the target signal extraction areas.

[0041] Next, the phase signal of each distance cell is processed using second-order differential processing:

[0042] ;

[0043] in, Indicates the first One sampling point, This represents the time interval between sampling points; second-order differential processing can effectively suppress low-frequency interference from the body and respiration, thereby separating the heartbeat signal. Performing second-order differential processing on all phase signals in the target area yields multiple candidate heartbeat signals, which respectively constitute candidate signal sets for chest heartbeat and carotid pulse.

[0044] S4. A variational mode decomposition-based segmentation algorithm is used on the candidate signals to obtain a series of heartbeat cycle signals. The similarity between the cycle signals is used to perform an evaluation algorithm based on dynamic time programming to screen out high-quality chest heartbeat signals and carotid pulse signals.

[0045] The process of step S4 is as follows:

[0046] The phase signal is decomposed using a variational mode decomposition algorithm to extract the dominant frequency signal of the heartbeat. In this example, 15 decomposition layers are selected, and the heartbeat range is set to 1.0~3.0Hz. A peak detection algorithm based on amplitude thresholding is used to segment the heartbeat period according to the location of the peak, resulting in a series of periodic segments. ;

[0047] Calculating the similarity between periodic segments using Dynamic Time Warping (DTW) Hierarchical clustering is performed using the average linking method, and the optimal number of clusters is determined according to the elbow rule. The cluster center with the largest number of samples is selected as the template signal for this candidate signal. The DTW average of the template signal and all periodic segments is calculated as the score. :

[0048] ;

[0049] The lower the average value of the DTW of the candidate signal, the higher the similarity between the periodic segments, and the less interference the signal is subjected to, and the better the signal quality. The signal with the lowest score in the candidate signal set is selected as the high-quality thoracic heartbeat signal and carotid artery signal to reconstruct the ECG signal.

[0050] S5, the ECG data is processed by using the equal frequency binning method to balance the amplitude distribution of the ECG waveform, and the thoracic heartbeat signal and the carotid artery pulse signal screened are combined with the ECG signal collected synchronously, and a time-dependent data set is constructed by using the sliding window sampling method.

[0051] The process of step S5 is as follows:

[0052] In order to avoid model training bias, the equal frequency binning method is used to preprocess the ECG signal collected synchronously. Specifically, the sampling points of the original ECG signal are arranged in ascending order of voltage value, divided into 256 bins, and ensured to contain the same number of data points. The ECG data in the bin is normalized to ensure the distribution in the interval [0, 1], while retaining the trend of the original waveform shape.

[0053] The thoracic heartbeat signal, the carotid artery pulse signal and the ECG signal screened are aligned in time by using the cross-correlation function and down-sampled to the same sampling rate of 100 Hz. The time stamp is stored in a joint data matrix with a size of 3*n, where n is the length of the signal, each table representing the thoracic vibration amplitude, carotid pulse amplitude and ECG voltage value at the same time. The matrix is divided by using the sliding window method, and the ECG voltage value data of the next time step under the current window is used as the label to construct the data set. In this example, the window length is 80 sampling points, and the step length is 10 sampling points.

[0054] S6, input the generated data set into the ECG prediction model, and the model captures the periodicity and waveform details of the heartbeat signal by extracting the time and space features between the signals, and predicts the value of the ECG.

[0055] The process of step S6 is as follows:

[0056] As Figure 2The example shown in this example constructs a parallel network architecture to reconstruct ECG signals. The time convolution network combines a channel attention mechanism (TCN-SE) to extract spatial features. TCN obtains current and past information through dilated causal convolution, which is suitable for prediction tasks; the SE mechanism can dynamically allocate weights between the chest heartbeat signal and the carotid pulse signal to obtain more favorable local detailed features. The bidirectional gated recurrent unit combines global attention (BiGRU-GA) to extract temporal features. Two layers of BiGRU obtain forward and backward time dependencies through bidirectional gated units; the GA mechanism obtains global time dependencies, focusing on capturing the periodicity of the P-QRS-T wave group in the electrocardiogram. After splicing and fusing the time and spatial features, the next time step ECG value is output through the fully connected layer.

[0057] The scheme of the present embodiment can realize high-precision non-contact ECG signal reconstruction, and exhibits excellent performance in both waveform and time. In terms of waveform morphology, the reconstructed and real ECG signals exhibit a high degree of consistency, with an average Pearson correlation coefficient (PCC) of 94.1% and an average root mean square error (RMSE) as low as 0.068 mV. In terms of time characteristics, the average time interval errors of key physiological parameters such as RR interval, QRS wave length, PR interval and QT interval are 0.18%, 0.99%, 2.60% and 1.34%, respectively, among which the error of RR interval is extremely low, indicating that the present example has extremely high timing accuracy in heart cycle recognition.

[0058] The content described in the embodiments of the present specification is only a list of implementation forms of the inventive concept, and is only for the purpose of description. The protection scope of the present application should not be regarded as being limited to the specific forms described in the present embodiments, and the protection scope of the present application also extends to equivalent technical means that can be thought of by those skilled in the art according to the inventive concept.

Claims

1. A contactless ECG signal monitoring method based on millimeter wave radar, characterized by, The method comprises the following steps: S1, the millimeter wave radar sends out a linear frequency modulation continuous wave, and receives a human body reflection signal, generates an intermediate frequency signal through a frequency mixer, and performs Fourier transform on the distance dimension to construct a two-dimensional data matrix; S2, by analyzing the distribution of the respiratory band energy of the two-dimensional data matrix at different distance bins, the matrix is subjected to static and dynamic interference elimination, and the energy intensity of the matrix after interference elimination is used to locate the position of the human chest cavity; S3, based on the relative spatial relationship between the chest cavity heart and the neck, the chest cavity heart and the neck region are further located, the second-order difference method is used to eliminate the interference of the body and the respiratory activity in the target region, and a candidate signal set of the chest cavity heartbeat and the carotid artery pulse is obtained; S4, the candidate signal is subjected to a variational mode decomposition segmentation algorithm to obtain a series of heartbeat period signals, and a dynamic time warping based evaluation algorithm is used according to the similarity between the period signals to screen high-quality chest cavity heartbeat signals and carotid artery pulse signals; S5, the electrocardiogram data is processed by using the equal frequency binning method to balance the amplitude distribution of the ECG waveform, the screened chest cavity heartbeat signal and carotid artery pulse signal are combined with the synchronous collected electrocardiogram signal, and a time-dependent data set is constructed by using the sliding window sampling method; S6, the generated data set is input into an ECG prediction model, the model captures the periodicity and waveform details of the heartbeat signal by extracting the time and space features between the signals, and predicts the ECG value.

2. A contactless ECG signal monitoring method based on millimeter wave radar according to claim 1, characterized in that, In the step S1, the millimeter wave radar is placed in front of the person to be detected, the related parameters of the emitted linear frequency modulation electromagnetic wave are set, the mixer generates an intermediate frequency signal after mixing the emitted and received signals, the intermediate frequency signal data is stored according to the receiving time sequence, and the dimension is wherein represents the number of sampling points of a single chirp, represents the number of chirps contained in each frame of data, represents the number of antennas of the received signal, and simultaneously, is the length of each frame of data; the distance dimension fast Fourier transform is performed on the data received by a single antenna to generate a two-dimensional matrix containing signal amplitude and phase information, and the dimension is also .

3. A contactless ECG signal monitoring method based on millimeter wave radar according to claim 1 or 2, characterized in that, In the step S2, a two-dimensional data matrix is extracted The phase signals of different distance bins are subjected to band-pass filtering of a classical respiratory frequency band to obtain sub-band signals after unwrapping the phase signals. Calculate the instantaneous energy of its sub-band signal and reconstruct it into a breathing energy matrix. Its dimensions are the same as the original matrix. Same, for The matrix is ​​obtained by normalizing the distance dimension. ; then With the original matrix Obtain the matrix by performing the Hadamard product operation. By searching The energy extremum point of the matrix in the distance dimension locates the center of the human chest cavity.

4. A contactless ECG signal monitoring method based on millimeter wave radar according to claim 1 or 2, characterized in that, In the step S3, according to the radiation diameter of the chest cavity region being 15 to 25 cm, the carotid artery region is located 15 to 25 cm above the chest cavity, and the radiation range is 8 to 12 cm; combined with the distance resolution of the millimeter wave radar being 4 cm, 5 continuous distance bins in the chest cavity region and 3 distance bins near the carotid artery region are selected as the target signal extraction region; the second-order difference processing is performed on the phase signal of each distance bin to suppress the interference of the body and the respiration, and a plurality of heartbeat candidate signals are obtained to form a candidate signal set of the chest cavity heartbeat and the carotid artery pulse, respectively.

5. A contactless ECG signal monitoring method based on millimeter wave radar according to claim 1 or 2, characterized in that, In the step S4, the variational mode decomposition algorithm is used to decompose the phase signal and extract the main frequency signal of the heartbeat; the peak value detection algorithm is performed based on the amplitude threshold value, the heartbeat period is segmented according to the position of the wave peak, and a series of period segments are obtained; the similarity between the period segments is calculated by using the dynamic time warping DTW, the hierarchical clustering is performed by using the average linkage method, the best clustering number is determined according to the elbow rule, the cluster center with the most sample quantity is selected as the template signal of the candidate signal, and the average DTW value of the template signal and all period segments is calculated as the score P, and the signal with the lowest score in the candidate signal set is selected as the high-quality chest cavity heartbeat signal and the carotid artery signal.

6. A contactless ECG signal monitoring method based on millimeter wave radar according to claim 1 or 2, characterized in that, In the step S5, in order to avoid model training bias, the equal frequency binning method is used to preprocess the synchronously collected ECG signal, the sampling points of the original ECG signal are arranged in ascending order of voltage value, divided into several bins, and each bin contains the same number of data points, the ECG data in the bin is normalized to ensure that it is distributed in the interval [0, 1], while retaining the trend of the original waveform form, the screened thoracic heartbeat signal, carotid pulse signal and ECG signal are aligned in time and sampling rate, and stored in a joint data matrix with a size of 3*n according to the time stamp, where n is the length of the signal; the matrix is segmented by using the sliding window method, and the ECG voltage value data of the next time step in the current window is used as the label to construct the data set.

7. A contactless ECG signal monitoring method based on millimeter wave radar according to claim 1 or 2, characterized in that, In the step S6, the ECG prediction model adopts a parallel network architecture, a time convolution network combines a channel attention mechanism to extract spatial features, dynamically allocates the weights of the thoracic heartbeat signal and the carotid pulse signal, a bidirectional gated recurrent unit combines a global attention to extract time features, and focuses on capturing the periodicity of the P-QRS-T wave group in the electrocardiogram.

Images