High-precision monitoring system of bed sheet type heart rate based on fusion of non-contact electrocardio and heart vibration signals

Abstract

The application provides a high-precision heart rate monitoring system fusing non-contact electrocardio and heart vibration signals, which comprises a capacitive coupling type non-contact electrocardio and piezoelectric film heart vibration signal acquisition module, a heart rate channel selection module based on signal quality evaluation, and a heart rate calculation module.The specific implementation comprises the following steps: in the first step, the capacitive coupling type non-contact electrocardio and piezoelectric film heart vibration signal acquisition module is used for acquiring non-contact electrocardio signals and heart vibration signals; in the second step, the heart rate channel selection module based on signal quality evaluation is used for calculating the quality of the acquired signals; and in the third step, the heart rate calculation module is used for calculating the heart rate according to the signal type, and the calculation result comprises high-precision heart rate information based on RR interval and average heart rate information based on heart vibration; the system is oriented to long-time high-precision physiological signal acquisition in a sleep scenario, fuses the advantages of two modes of non-contact electrocardio and heart vibration, and can effectively acquire continuous, long-time and high-precision heart rate information.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of sleep and heart health monitoring, and relates to non-contact electrocardiogram and heart vibration monitoring technology, physiological signal processing technology, in particular, a non-invasive, non-contact cardiovascular disease and sleep rhythm high-precision monitoring method in a sleep scenario. BACKGROUND

[0002] In recent years, wearable cardiovascular monitoring technology has rapidly emerged in the medical field and become an important innovation in the field of non-contact heart monitoring. The non-contact physiological signal monitoring method has brought significant advantages and necessity for cardiovascular disease monitoring due to its non-allergic characteristics. Existing non-contact physiological signal monitoring methods include the use of millimeter wave radar, piezoelectric film, non-contact electrocardiogram, and optical fiber vibration sensing. The millimeter wave radar, piezoelectric film, and optical fiber vibration sensing methods can only measure the rhythm information of heart vibration, and the precision is not high enough. However, accurate RR interval is crucial for cardiovascular disease monitoring and sleep rhythm monitoring.

[0003] Non-contact electrocardiogram is a method that can measure electrocardiogram through a layer of clothes. Compared with existing commercial millimeter wave radar, piezoelectric film, and optical fiber vibration sensing methods, it has significantly higher precision. Therefore, the non-contact electrocardiogram heart rate monitoring method ensures comfort and monitoring precision, and is increasingly sought after. However, in actual use, the signal quality is not satisfactory due to the layer of clothes. At present, research mainly focuses on non-contact electrocardiogram hardware improvement methods. In many cases, due to the influence of fabric and individual differences, the non-contact electrocardiogram signal quality is poor, and it is not possible to obtain complete continuous heart rate. SUMMARY

[0004] The present application is truly aimed at the technical problems existing in the prior art, and provides a collection and analysis system integrating non-contact electrocardiogram and heart vibration. The existing solution is improved, and a heart rate calculation architecture combining heart vibration and non-contact electrocardiogram dual modalities is proposed, which can effectively record heart rate information during the whole night sleep and solve the problem of difficulty in obtaining long-term heart rate information during the whole night sleep. At the same time, it also ensures heart rate information with a certain precision.

[0005] In order to achieve the above purpose, the technical scheme of the present application is as follows: the present application provides a bedsheet type heart rate high-precision monitoring system combining non-contact electrocardiogram and heart vibration signals, which comprises: a capacitive coupling type non-contact electrocardiogram and piezoelectric film heart vibration signal collection module, a heart rate channel selection module based on signal quality evaluation, and a heart rate calculation module,

[0006] The capacitive coupling type non-contact electrocardiogram and piezoelectric film heart vibration signal collection module is used to collect heart rhythm related information of the human body in a non-contact manner;

[0007] The signal quality evaluation-based heart rate channel selection module is configured to evaluate non-contact electrocardio signal quality and heart shock signal quality, and to eliminate data without obvious physiological information, such as strong power frequency interference and motion artifacts.

[0008] The heart rate calculation module is configured to include R-wave recognition and positioning using non-contact electrocardio, and to calculate accurate RR intervals and average heart rate of heart shock signals.

[0009] Further, the capacitive coupling type non-contact electrocardio and piezoelectric film heart shock signal acquisition module is configured to acquire human electrocardio signals and heart shock signals in a non-contact manner. The acquisition port includes a non-contact electrocardio electrode and a piezoelectric film heart shock sensor. Non-contact electrocardio can obtain ECG signals through clothes, and can be widely used in the evaluation of cardiovascular diseases and obtain clinical-level accuracy. When the non-contact electrocardio signal quality is poor, the heart shock signal can be used as a supplement to overcome the inaccuracy of non-contact electrocardio caused by the thickness difference of the clothes of the tester, so that the whole night heart rate information change can be obtained.

[0010] Further, the signal quality evaluation-based heart rate channel selection module is configured to calculate the signal quality and signal category selection evaluation of the acquired non-contact electrocardio and heart shock signals, to filter out non-contact physiological signals with poor signal quality, and to retain useful information. In addition, it is also used for evaluating the signal type and distinguishing whether the physiological signals collected for calculating the heart rate are non-contact electrocardio or heart shock signals.

[0011] Further, the heart rate calculation module includes a non-contact electrocardio R-wave positioning algorithm and a piezoelectric film-based heart shock heart rate calculation algorithm. The final output heart rate of the system is based on the signal quality evaluation-based heart rate channel selection module.

[0012] Further, the signal quality and signal category selection evaluation of the non-contact electrocardio and heart shock signals mainly optimizes time domain features, frequency domain features, R-wave features and nonlinear features, and uses machine learning or deep learning methods to effectively evaluate the signal quality of the above features.

[0013] Further, the non-contact electrocardio R-wave positioning mainly uses a variable mode decomposition enhanced R-wave detection algorithm to effectively and accurately identify the QRS complex of non-contact electrocardio and reduce the calculation complexity of monitoring. It includes low-frequency noise removal and QRS complex enhancement based on variable mode decomposition, first derivative, smooth Shannon energy envelope extraction and R-wave peak detection based on Hilbert transform.

[0014] Further, the low-frequency noise removal and QRS complex enhancement of the variational mode decomposition mainly includes using two-stage variational mode decomposition to enhance the QRS complex feature of the non-contact electrocardio, wherein for the first stage, the bandwidth control parameter and the mode number of the variational mode decomposition are respectively set to 50000 and 3 to filter out low-frequency noise. Then, the processed signal is subjected to the bandwidth control parameter and the mode number of the variational mode decomposition are respectively set to 8000 and 10 to enhance the R-wave feature in the range of 3-20 Hz. Then, the first derivative is performed again to further enhance the R-wave feature, and finally, the processed signal is normalized.

[0015] Further, the smooth Shannon energy envelope extraction mainly aims to enhance the R-wave feature, and the calculation process mainly includes calculating the Shannon energy and smoothing the moving average filter. For each signal point , the Shannon energy is calculated , and the calculated is subjected to the sliding average filter to obtain , and finally, the evaluation Shannon energy envelope is obtained. The energy trend of the signal is obtained through the smooth Shannon energy envelope extraction, so that the signal feature is more obvious.

[0016] Further, the R-wave peak detection based on Hilbert transform is performed on , and in the Hilbert-transformed signal, the zero-crossing points of the positive slope part are found, which are the preliminarily detected R-wave positions. A window containing 100 samples (50 samples in front and 50 samples behind) is constructed around each preliminarily detected R-wave position. The maximum value in the window is found and taken as the final R-wave position.

[0017] Further, the piezoelectric film heart vibration heart rate calculation algorithm includes using a 6th-order Butterworth band-pass filter of 0.7 Hz-10 Hz to denoise the heart vibration signal, effectively filtering out the respiratory noise and high-frequency vibration noise of the heart vibration, and then using a method including but not limited to multiple signal classification (MUSIC) to estimate the heart rate of the heart vibration signal.

[0018] A bedsheet type high-precision heart rate monitoring system fusing non-contact electrocardio and heart vibration signal mainly includes the following steps:

[0019] Step 1, the bed sheet type non-contact high-precision long-term heart rate monitoring system is placed on the bed, in which the capacitive coupling type non-contact electrocardiogram and piezoelectric film heart shock signal acquisition module is used to acquire non-contact electrocardiogram and heart shock signal. When the human body lies on the bed wearing cotton clothes, the non-contact electrocardiogram acquisition module can acquire electrocardiogram through the clothes, and the piezoelectric film heart shock signal acquisition module can acquire the heart shock of the human body. The two signals are synchronously acquired and transmitted to the heart rate channel selection module;

[0020] Step 2, the acquired non-contact electrocardiogram and heart shock signal are input to the signal quality evaluation heart rate channel selection module, which is used to calculate the signal quality of the acquired signal, screen out non-contact physiological signals with poor signal quality, and retain useful information. In addition, it is also used to evaluate the signal segment used to calculate the heart rate information, and judge whether to use non-contact electrocardiogram to calculate heart rate or use heart shock to calculate heart rate. The specific method is to calculate the time domain feature, frequency domain feature, R wave feature and nonlinear feature by using the non-contact electrocardiogram signal segment and the heart shock signal segment, and the different characteristic result values (such as time domain features including the following features: sSQI, kSQI, sdnSQI, pliSQI. The frequency domain includes the following features: pSQI, purSQI, basSQI, LpSQI, MpSQ, the nonlinear feature includes the following features: ApEn, SampEn, FuzzyEn, DisEn, MSEn, MFEn, and the R wave feature includes the following features: bsSQI, eSQI, hfSQI, rsdSQI, PiCASQI, pcaSQI). And input the above features into the traditional machine learning model (including but not limited to support vector machine, random forest, XGboost, etc.) to classify the signal quality by using deep learning model. The classification result includes judging whether the non-contact electrocardiogram signal quality of the segment is reliable, and whether the heart shock signal quality is reliable. If the non-contact electrocardiogram quality is reliable, the non-contact electrocardiogram is used to calculate the heart rate, if it is not reliable, the heart shock signal is used to calculate the heart rate. If both are not reliable, the segment signal is discarded.

[0021] Step 3, the heart rate calculation module is used for heart rate calculation according to the signal type, and the calculation result includes high-precision heart rate information based on RR interval and heart rate information based on heart shock. According to the calculation result of step 2, if the non-contact electrocardiogram is reliable, the high-precision heart rate information based on RR interval is used, and if the non-contact electrocardiogram is unreliable, the heart rate information calculation method based on heart shock is used. For the non-contact electrocardiogram heart rate calculation method, a R wave detection algorithm based on variational mode decomposition enhancement is mainly used to effectively and accurately identify the QRS complex of the non-contact electrocardiogram, and the calculation complexity of the monitoring is reduced. It includes low-frequency noise removal and QRS wave group enhancement based on variational mode decomposition, first derivative, smooth shannon energy envelope extraction and R wave peak detection based on Hilbert transform. For the heart rate information calculation method based on heart shock, a 6th order Butterworth band-pass filter of 0.7Hz~10Hz is used to denoise the heart shock signal, effectively filtering out the respiratory noise and high-frequency vibration noise of the heart shock, and then the heart rate is estimated by segmenting the signal and using methods including but not limited to multiple signal classification (MUSIC).

[0022] Compared with the prior art, the present application has the following advantages:

[0023] (1) The mattress type non-contact electrocardiogram and heart shock acquisition device of the system can collect the electrocardiogram, heart shock and respiratory coupled signal of the patient through a layer of clothes. This non-contact measurement method has the advantages of long-term monitoring, high comfort and the like compared with the traditional contact type wet electrode measurement method;

[0024] (2) The non-contact electrocardiogram signal quality evaluation algorithm of the system extracts signal features according to the characteristics of the non-contact electrocardiogram signal, and has the advantage of high accuracy;

[0025] (3) In the case of poor signal quality, the system adaptively uses the heart shock signal to calculate the heart rate, which can restore the whole night heart rate information as much as possible;

[0026] (4) The non-contact electrocardiogram high-precision R wave positioning monitoring algorithm based on variational mode decomposition enhancement involved in the project can accurately calculate and restore the heart rate under the condition of low signal-to-noise ratio of non-contact electrocardiogram;

[0027] (5) The project uses the method of non-contact electrocardiogram and heart shock sensing fusion to realize the whole night high-precision calculation of heart rate in the sleep scene. The system application scene is bed sheet type, and the heart rate during the night sleep process is monitored. The high-precision monitoring of the heart rate during the night sleep process is of great significance for cardiovascular disease monitoring. In the future, it can effectively monitor arrhythmia events, sleep rhythm analysis, etc. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1is a system layout of the present application,

[0029] Figure 2 is a system block diagram of the present application,

[0030] Figure 3 is a system algorithm overall flowchart of the present application,

[0031] Figure 4 is a non-contact ECG R wave detection algorithm based on variational mode decomposition enhancement. DETAILED DESCRIPTION

[0032] In order to better understand the above technical solutions of the present application, further detailed description is made below in combination with the drawings and examples.

[0033] Example 1:

[0034] The present application provides a high-precision bedsheet type heart rate monitoring system fusing non-contact ECG and heart vibration signals, comprising: a capacitive coupling type non-contact ECG and piezoelectric film heart vibration signal acquisition module, a heart rate channel selection module based on signal quality evaluation, and a heart rate calculation module,

[0035] The capacitive coupling type non-contact ECG and piezoelectric film heart vibration signal acquisition module is used to acquire information related to the cardiac rhythm of the human body in a non-contact manner.

[0036] The heart rate channel selection module based on signal quality evaluation is used to evaluate the quality of the non-contact ECG signal and the quality of the heart vibration signal, and to eliminate data without obvious physiological information including strong power frequency interference and motion artifacts.

[0037] The heart rate calculation module is used to include R wave recognition and positioning using non-contact ECG, calculation of accurate RR interval, and average heart rate calculation of heart vibration signal.

[0038] The capacitive coupling type non-contact ECG and piezoelectric film heart vibration signal acquisition module is used to acquire the ECG signal and the cardiac vibration signal of the human body in a non-contact manner. Its acquisition port includes two parts of non-contact ECG electrode and piezoelectric film cardiac vibration sensor. Non-contact ECG can obtain ECG signal through clothes and can be widely used for evaluation of cardiovascular diseases and obtain clinical level precision. When the quality of non-contact ECG signal is poor, the cardiac vibration signal can be used as a supplement to overcome the inaccuracy of non-contact ECG due to the difference in thickness of the clothes of the tester, so that the whole night heart rate information change can be obtained.

[0039] The signal quality assessment-based heart rate channel selection module is configured to calculate signal quality and signal type selection assessment of the collected non-contact electrocardiogram and cardiac vibration signals, and to filter out non-contact physiological signals with poor signal quality and to retain useful information. In addition, the signal quality assessment-based heart rate channel selection module is configured to assess the signal type and determine whether the physiological signals collected for calculating the heart rate are non-contact electrocardiogram signals or cardiac vibration signals.

[0040] The heart rate calculation module includes a non-contact electrocardiogram R-wave positioning algorithm and a piezoelectric film-based cardiac vibration heart rate calculation algorithm. The final output heart rate of the system is determined according to the signal quality assessment-based heart rate channel selection module.

[0041] The signal quality and signal type selection assessment of the non-contact electrocardiogram and cardiac vibration signals mainly involves preferred time domain features, frequency domain features, R-wave features and nonlinear features, and machine learning or deep learning methods are used to effectively assess the signal quality of the above features.

[0042] The non-contact electrocardiogram R-wave positioning mainly uses a variational mode decomposition-enhanced R-wave detection algorithm to effectively and accurately identify the QRS complex of the non-contact electrocardiogram and reduce the calculation complexity of the monitoring. The algorithm includes low-frequency noise removal and QRS complex enhancement based on variational mode decomposition, first derivative, smooth Shannon energy envelope extraction and R-wave peak detection based on Hilbert transform.

[0043] The low-frequency noise removal and QRS complex enhancement based on variational mode decomposition mainly involves using two-stage variational mode decomposition to enhance the QRS complex features of the non-contact electrocardiogram. For the first stage, the bandwidth control parameter and the number of modes are set to 50,000 and 3, respectively, to filter out low-frequency noise. Then, the processed signal is subjected to variational mode decomposition with the bandwidth control parameter and the number of modes set to 8,000 and 10, respectively, to enhance the R-wave features in the range of 3-20 Hz. Then, the first derivative is calculated to further enhance the R-wave features, and finally, the processed signal is normalized.

[0044] The smooth Shannon energy envelope extraction mainly involves enhancing the R-wave features. The calculation process mainly includes calculating the Shannon energy and performing moving average filtering. For each signal point , the Shannon energy is calculated as . The calculated is subjected to sliding average filtering to obtain , and finally, the evaluation Shannon energy envelope The energy change trend of the signal is extracted by smoothing the Shannon energy envelope, so that the signal characteristics are more obvious.

[0045] The R wave peak detection based on Hilbert transform is used to The Hilbert transform is performed, and in the signal after the Hilbert transform, the zero-crossing points of the positive slope part are found, which are the preliminary detected R wave positions. A window containing 100 samples (50 samples in front and 50 samples behind) is constructed around each preliminary detected R wave position. The maximum value in the window is found and used as the final R wave position. The heart rate calculation algorithm of the piezoelectric film heart vibration comprises the following steps: using a 6th order Butterworth band-pass filter of 0.7Hz-10Hz to reduce the noise of the heart vibration signal, effectively filtering out the respiratory noise and high-frequency vibration noise of the heart vibration, and then using a method including but not limited to multiple signal classification (MUSIC) to estimate the heart rate of the heart vibration signal.

[0046] Embodiment 2:

[0047] Figure 1 The system arrangement diagram of the application is composed of two parts, namely a bed sheet sensor 1 and a rear-end circuit 5. For the sensor end, from top to bottom are a first non-contact electrocardio electrode 2, a piezoelectric film sensor 3 and a second non-contact electrocardio electrode 4, the two non-contact electrocardio electrodes are connected to a non-contact electrocardio analog front end, and the piezoelectric film sensor 3 is connected to a heart vibration analog front end. The rear-end circuit includes the non-contact electrocardio analog front end and the heart vibration analog front end, and an analog-digital converter to finally obtain the non-contact electrocardio and heart vibration signals.

[0048] Figure 2 The specific implementation steps are as follows: first, input the collected non-contact electrocardio signal and heart vibration signal into a signal quality evaluation heart rate channel selection module, which is used to calculate the signal quality of the collected signal, filter out the non-contact physiological signal with poor signal quality, and retain useful information. In addition, it is also used to evaluate the signal type and determine the subsequent heart rate calculation method. For the non-contact electrocardio signal with good quality, a high-precision heart rate information based on RR interval is used, and for the electrocardio signal with poor quality and the heart vibration signal with good quality, a heart rate calculation method based on MUSIC spectrum is used.

[0049] Figure 3is the overall flowchart of the algorithm, first for the electrocardiogram signal quality assessment 6, using time-frequency domain indicators, frequency domain indicators, nonlinear indicators, (such as time domain features first the following features sSQI, kSQI, sdnSQI, pliSQI. Frequency domain includes but is not limited to the following features pSQI, purSQI, basSQI, LpSQI, MpSQ, nonlinear features include the following features first: ApEn, SampEn, FuzzyEn, DisEn, MSEn, MFEn, R wave feature-based includes the following features first: bsSQI, eSQI, hfSQI, rsdSQI, PiCASQI, pcaSQI), using machine learning or deep learning method to complete the electrocardiogram signal quality assessment, heart shock signal quality assessment, eliminate invalid data without obvious electrocardiogram signal. For better non-contact electrocardiogram signal quality, use variational mode decomposition to enhance QRS complex 7, use R wave detection algorithm based on variational mode decomposition enhancement to identify R wave position 8 and calculate heart rate.

[0050] For poor non-contact electrocardiogram signal quality, and good heart shock quality, use piezoelectric film-based heart shock heart rate calculation algorithm, including using 0.7Hz~10Hz 6-order Butterworth band-pass filter 9 to denoise heart shock signal, effectively filtering out respiratory noise and high-frequency vibration noise of heart shock, including but not limited to multiple signal classification (MUSIC) method 10 to estimate heart rate of heart shock signal. Finally, get the whole night heart rate information 11.

[0051] Figure 4 is the flowchart of the R wave detection algorithm based on variational mode decomposition enhancement, first using variational mode decomposition to remove low-frequency noise of the signal and enhance QRS complex, then taking the derivative of the processed signal, and then smoothing the Shannon energy envelope to extract the last R wave peak based on Hilbert transform.

[0052] Among them, the variational mode decomposition includes: initializing a set of frequency centers And modal function And Lagrange multiplier Update the modal function by minimizing the variational model ,

[0053]

[0054] Among them, is the weight of the smoothing term.

[0055] Then update the center frequency And Lagrange multiplier ,

[0056] ;

[0057] ,

[0058] is a step parameter, the process of updating the modal function, updating the frequency center, updating the Lagrange multiplier is repeated until the convergence condition is met, and finally the modal function and its corresponding frequency are output , the above steps complete the variational modal decomposition, and the non-contact electrocardio signal processed by the variational modal decomposition is output as , the first-order difference of is carried out and normalized to obtain ;

[0059] ,

[0060] Then, the smooth shannon energy envelope is extracted:

[0061] ,

[0062] Finally, the R wave peak detection method based on Hilbert transform is performed on the signal.

[0063] The main steps of the R wave peak detection method based on Hilbert transform include: first, calculate the discrete Fourier transform (DFT) of the signal s(n), denoted as ; second, process the signal by inverse discrete Fourier transform (IDFT); then perform Hilbert transform on the processed signal to determine the positive slope zero crossing point of the signal as the R peak position; after the preliminary detection of the R wave position, the R peak position detected by the above steps may deviate from the true R peak position; finally, peak correction: in order to correct these deviations, a window containing 82 samples is constructed around each detected R peak position, and the maximum peak value in this window is found to determine a more accurate R peak position.

[0064] Example 3:

[0065] The hardware and algorithms described above are used to analyze the heart rate variability (HRV) of the patient and output the results. HRV refers to the degree of variation in heart rate over a certain period of time, and can reflect the functional status of the autonomic nervous system. These results can be used to assess the user's heart health and the function of the autonomic nervous system. Table 1 gives several typical HRV indicators, including: mean RR interval (MeanNN), standard deviation of all RR intervals (SDNN), standard deviation of adjacent RR interval differences (SDSD), power of RR interval signal in 0.04 to 0.15 Hz frequency band (LF), power of RR interval signal in 0.15 to 0.4 Hz frequency band (HF), standard deviation of short-term variability in figure analysis (SD1), standard deviation of long-term variability in figure analysis (SD2), sample entropy (SampEn). By comparing the non-contact system with the traditional contact electrocardio acquisition system, it can be found that the system has high precision; Table 1 is the comparison result of heart rate variability analysis of our system and standard equipment. It can be seen that the error of the device compared with the standard device is very small, reaching the advanced level;

[0066] .

[0067] It should be noted that the above embodiments are not intended to limit the scope of protection of the present application, and equivalent transformations or substitutions made on the basis of the above technical solutions fall within the scope of protection of the claims of the present application.

Claims

1. A bed sheet-type high-precision heart rate monitoring system integrating non-contact electrocardiogram and cardiac vibration signals, characterized in that: The monitoring system includes a capacitively coupled non-contact ECG and piezoelectric film echocardiogram signal acquisition module, a heart rate channel selection module based on signal quality assessment, and a heart rate calculation module. The capacitively coupled non-contact ECG and piezoelectric thin film oscillation signal acquisition module is used to acquire human heart rhythm-related information in a non-contact manner. The heart rate channel selection module based on signal quality assessment is used to evaluate the quality of non-contact electrocardiogram signals and cardiac vibration signals, and to remove data including strong power frequency interference, motion artifacts and data with no obvious physiological information. The heart rate calculation module is used to include R wave identification and localization using non-contact ECG, calculating accurate RR intervals, and calculating average heart rate from the oscillation signal. The heart rate calculation module includes a non-contact ECG R-wave localization algorithm and a heart vibration heart rate calculation algorithm based on piezoelectric film. The final output heart rate of the system is determined by the heart rate channel selection module based on signal quality assessment. Non-contact ECG R-wave localization mainly uses variational mode decomposition-enhanced R-wave detection algorithms to effectively and accurately identify QRS complex waves in non-contact ECG and reduce the computational complexity of monitoring. These algorithms include low-frequency noise removal and QRS complex enhancement based on variational mode decomposition, first derivative and smoothed Shannon energy envelope extraction, and R-wave peak detection based on Hilbert transform. Low-frequency noise removal and QRS complex enhancement by variational mode decomposition mainly involve using two-stage variational mode decomposition to enhance the QRS complex characteristics of non-contact ECG. In the first stage, the bandwidth control parameter (α) and the number of modes (K) of variational mode decomposition are set to 50000 and 3, respectively, to filter out low-frequency noise. Then, the processed signal is subjected to variational mode decomposition again with the bandwidth control parameter (α) and the number of modes (K) set to 8000 and 10, respectively, to enhance the R-wave characteristics in the 3–20 Hz range. Then, the first derivative is applied to further enhance the R-wave characteristics. Finally, the processed signal is normalized.

2. The bed sheet-type high-precision heart rate monitoring system integrating non-contact electrocardiogram and cardiac vibration signals according to claim 1, characterized in that: The capacitively coupled non-contact ECG and piezoelectric film cardiac vibration signal acquisition module is used to acquire human electrocardiogram (ECG) and cardiac vibration signals in a non-contact manner. Its acquisition port includes non-contact ECG electrodes and a piezoelectric film cardiac vibration sensor. Non-contact ECG acquires ECG signals through clothing for the assessment of cardiovascular diseases and achieves clinical-level accuracy. When the quality of the non-contact ECG signal is poor, the cardiac vibration signal is used as a supplement to overcome the inaccuracy caused by the difference in clothing thickness of the test subject in non-contact ECG and obtain the heart rate information changes throughout the night.

3. The bed sheet-type high-precision heart rate monitoring system integrating non-contact electrocardiogram and cardiac vibration signals according to claim 1, characterized in that: The heart rate channel selection module based on signal quality assessment is used to calculate and evaluate the signal quality and signal category of the acquired non-contact ECG and cardiac vibration signals. It filters out non-contact physiological signals with poor signal quality and retains useful information. It is also used to evaluate the signal type and determine whether the physiological signal acquired for calculating heart rate is a non-contact ECG or cardiac vibration signal.

4. The bed sheet-type high-precision heart rate monitoring system integrating non-contact electrocardiogram and cardiac vibration signals according to claim 3, characterized in that: The signal quality and signal category selection evaluation of non-contact electrocardiogram and cardiac vibration signals includes calculating time-domain features, frequency-domain features, R-wave features, and nonlinear features, and using machine learning methods to effectively evaluate the signal quality of the above features.

5. The bed sheet-type high-precision heart rate monitoring system integrating non-contact electrocardiogram and cardiac vibration signals according to claim 4, characterized in that: Smoothing the Shannon energy envelope extraction primarily aims to enhance R-wave characteristics. The calculation process mainly includes calculating the Shannon energy and smoothing using moving average filtering. For each signal point x(n), the Shannon energy E(n) is calculated as follows: Then, a moving average filter is applied to the calculated E(n) to obtain... Finally, the Shannon energy envelope s[n] is obtained, and the energy change trend of the signal is obtained by smoothing the Shannon energy envelope extraction, making the signal characteristics more obvious.

6. The bed sheet-type high-precision heart rate monitoring system integrating non-contact electrocardiogram and cardiac vibration signals according to claim 5, characterized in that: R-wave peak detection based on Hilbert transform involves performing a Hilbert transform on s[n]. In the signal after the Hilbert transform, the zero-crossing points of the positive slope portion are found. These points are the initially detected R-wave positions. A window containing 100 samples is constructed around each initially detected R-wave position, with 50 samples in front and 50 samples behind. The maximum value is found in this window and taken as the final R-wave position.

7. The bed sheet-type high-precision heart rate monitoring system integrating non-contact electrocardiogram and cardiac vibration signals according to claim 4, characterized in that: The algorithm for calculating heart rate from cardiac vibration using piezoelectric thin films includes using a 6th-order Butterworth bandpass filter (0.7Hz~10Hz) to denoise the cardiac vibration signal, effectively filtering out respiratory noise and high-frequency vibration noise. Then, the heart rate is estimated by segmenting the signal and using the Multiple Signal Classification (MUSIC) method.

8. A high-precision bed sheet-based heart rate monitoring method integrating non-contact electrocardiogram and cardiac vibration signals, characterized in that, The bed sheet-type high-precision heart rate monitoring system, which integrates non-contact electrocardiogram and cardiac vibration signals, as described in any one of claims 1-7, The method includes the following steps: Step 1: The bed sheet-type non-contact high-precision long-term heart rate monitoring system is placed on the bed. The capacitively coupled non-contact ECG and piezoelectric film vibratory signal acquisition modules are used to collect non-contact ECG and vibratory signals. When the person is lying on the bed wearing cotton clothing, the non-contact ECG acquisition module collects the ECG signal through the clothing, while the piezoelectric film vibratory signal acquisition module collects the heart vibration. Both signals are collected simultaneously and transmitted to the heart rate channel selection module. Step 2: The acquired non-contact ECG and cardiac vibration signals are input into the heart rate channel selection module for signal quality assessment. This module calculates the signal quality, filters out non-contact physiological signals with poor signal quality, and retains useful information. It also evaluates the signal segments used to calculate heart rate information, determining whether to use non-contact ECG or cardiac vibration for heart rate calculation. Specifically, it calculates time-domain, frequency-domain, R-wave, and nonlinear characteristics using non-contact ECG and cardiac vibration signal segments. The results of these different features are then input into a machine learning model to classify the signal quality. The classification results include whether the non-contact ECG signal quality and the cardiac vibration signal quality are reliable. If the non-contact ECG quality is reliable, it is used to calculate the heart rate; if unreliable, cardiac vibration is used; if neither is reliable, the signal segment is discarded. Step 3: The heart rate calculation module is used to calculate the heart rate according to the signal type. The calculation results include high-precision heart rate information based on the RR interval and heart rate information based on cardiac vibration. According to the calculation results of Step 2, if the non-contact ECG is reliable, the high-precision heart rate information based on the RR interval is sampled. If the non-contact ECG is unreliable, the heart rate information calculation method based on cardiac vibration is used. For the non-contact ECG heart rate calculation method, the variational mode decomposition-enhanced R-wave detection algorithm is mainly used to effectively and accurately identify the QRS complex wave of the non-contact ECG and reduce the computational complexity of monitoring. The R-wave detection algorithm based on variational mode decomposition enhancement is as follows: First, low-frequency noise in the signal is removed using variational mode decomposition and QRS group enhancement is performed. Then, the processed signal is subjected to a first derivative, followed by smoothing the Shannon energy envelope extraction. Finally, R-wave peak detection is performed based on Hilbert transform. Variational mode decomposition includes: initializing a set of frequency centers { } and modal functions { } and the Lagrange multiplier λ, updating the mode function by minimizing the variational model. : Where α is the weight of the smoothing term, Then update the center frequency. And the Lagrange multiplier λ: The step size parameter is used to repeatedly update the modal functions, frequency centers, and Lagrange multipliers until the convergence condition is met, and finally, the modal functions are output. and their corresponding frequencies The above steps complete the variational mode decomposition. The output of the non-contact ECG signal after variational mode decomposition is y[n]. First-order difference is performed on y[n]. And normalization is performed to obtain : = y[n+1]- y[n] Then, smooth the Shannon energy envelope. extract: Finally, an R-wave peak detection method based on Hilbert transform was used to detect the signal.

Images