A continuous blood pressure monitoring method based on single-channel cardiac shock signal

Abstract

The present invention discloses a continuous blood pressure monitoring method based on a single-channel cardiac impact signal in the field of physiological signal detection, comprising collecting data of multiple subjects and constructing three blood pressure model data sets; using the three blood pressure model data sets to calculate Bayesian fusion model parameters; using the parameters of the Bayesian fusion model to perform blood pressure model fitting and systolic blood pressure calculation on the subjects. The present invention utilizes different signal components in the cardiac impact to extract pulse transmission time, and simultaneously measures the cardiac impact intensity and the shape characteristics of the cardiac impact signal as a basis for relative systolic blood pressure calculation, and utilizes the Bayesian model to fuse the three different indicators to further improve the accuracy of blood pressure calculation. The provided continuous blood pressure measurement method has the characteristics of no load, convenient measurement, low equipment complexity and low cost, overcomes the disadvantage of large errors commonly existing in single parameter models, and has great application potential.

Description

Technical Field

[0001] The invention relates to the field of physiological signal detection, and in particular to a continuous blood pressure monitoring method based on a single-channel cardiac shock signal. Background Art

[0002] Blood pressure is an important physiological parameter of the human body and an important factor affecting human health today. Existing blood pressure measurement methods can be divided into single blood pressure measurement and dynamic blood pressure measurement based on the measurement method. Among them, continuous dynamic blood pressure at night is a more sensitive cardiovascular prognosis indicator than daytime blood pressure. It helps to reveal the causes of latent hypertension and refractory hypertension, assists in the management of metabolic diseases, and can help evaluate sleep quality. Therefore, it has great scientific research value and social value.

[0003] The current dynamic blood pressure measurement method mainly uses a cuff-type blood pressure monitor, whose inflation and deflation operations are likely to cause interference to the subject, increase the subject's anxiety, and the measurement frequency is low, making it difficult to reflect the instantaneous fluctuations in blood pressure. Another common dynamic blood pressure measurement method is to use photoplethysmograph (PPG) to measure pulse transmission time (PTT) for blood pressure fitting and measurement. On the one hand, this measurement method is not accurate enough due to the existence of the pre-ejection period, and on the other hand, it also requires wearing a strap or finger cuff, which is not suitable for long-term use.

[0004] Continuous dynamic blood pressure measurement based on ballistocardiograph (BCG) signals has the advantages of non-contact and zero load, and is very suitable for blood pressure measurement for a long time or during sleep. The invention patent with application publication number CN108186000A discloses a method for measuring blood pressure using ballistocardiograph and PPG signals, but the user needs to wear a PPG finger cuff for measurement, and it is impossible to achieve complete zero load. The invention patent with application publication number CN108261193A discloses a method for measuring blood pressure based on calculating the pulse wave transmission time based on dual-channel ballistocardiograph signals, but the use of more than two sensing devices and a single transmission time parameter for blood pressure fitting and measurement still has the problem of complex equipment and insufficient measurement accuracy.

[0005] Based on this, the present invention designs a continuous blood pressure monitoring method based on a single-channel cardiac shock signal to solve the above problems. Summary of the invention

[0006] The purpose of the present invention is to provide a continuous blood pressure monitoring method based on a single-channel cardiac impulse signal, to achieve true zero-load, non-contact continuous dynamic blood pressure monitoring, and at the same time ensure the accuracy of the measurement results.

[0007] To achieve the above object, the present invention provides the following technical solution: a continuous blood pressure monitoring method based on a single-channel heartbeat signal, comprising the following steps:

[0008] S1, collect data from multiple subjects and construct three blood pressure model datasets;

[0009] S2, calculate the Bayesian fusion model parameters using three blood pressure model datasets;

[0010] S3, the parameters of the Bayesian fusion model are used to fit the blood pressure model and calculate the systolic blood pressure of the subjects;

[0011] Wherein S1 includes the following steps:

[0012] For each subject, a specific ballistocardiogram measurement method is used to obtain a ballistocardiogram signal at a specific location;

[0013] The cardiac impact signal of each subject is processed using two different filters, high frequency and low frequency, to obtain the limb vibration component and heart vibration component of the cardiac impact signal;

[0014] For each subject's cardiac impact signal, find the cardiac impact characteristic points J of two different cardiac impact signal components, and calculate the time difference T between the two cardiac impact characteristic points J as the first relative systolic pressure, recorded as RSBP1, and then calculate the time difference T between point J in the "heart vibration component" and point J in the "limb vibration component". This time difference is the pulse transmission time for blood flow to be transmitted from the heart end to the limb end, and the parameter T is used as the second relative systolic pressure RSBP2;

[0015] For each subject, the cardiac impulse strength (BPS) of the two heartbeat components was calculated separately and recorded as the second category relative systolic blood pressure (RSBP2);

[0016] For each subject's cardiac impulse signal, locate the characteristic points I, J, and K, calculate the amplitude difference between points J and K and points I and J, and record it as relative systolic blood pressure RSBP3;

[0017] For each subject, the systolic blood pressure was measured using a standard sphygmomanometer, and a linear model was fitted with the three RSBPs as independent variables and the standard systolic blood pressure as the dependent variable to obtain three different SBP models. 3N linear blood pressure models can be generated for N subjects.

[0018] Preferably, the specific ballistocardiogram measurement method obtains the ballistocardiogram information at a specific location by collecting the ballistocardiogram signal through a specific sensor capable of collecting human micro-vibration signals, a specific medium between the sensor and the human body, and a specific hardware circuit, a signal sampling rate, etc.; placing the specific sensor at a specific location to obtain a signal that meets the requirements.

[0019] Preferably, the specific hardware circuit preferably has a circuit with noise filtering function, signal amplification function and analog-to-digital conversion function; the specific signal sampling rate is preferably a sampling rate above 250 Hz, and the sensor is placed in a specific position under the mattress, under the left upper arm, right upper arm or under the leg of the human body.

[0020] Preferably, the specific method for obtaining the limb vibration component and the heart vibration component of the cardiac impact signal is to use a Butterworth bandpass filter with a cutoff frequency of [2Hz, 6Hz] to process the cardiac impact signal to obtain the limb vibration component, and to use a Butterworth bandpass filter with a cutoff frequency of [8Hz, 24Hz] to process the cardiac impact signal to obtain the heart vibration component.

[0021] Preferably, the J method for finding the cardiac impact characteristic points of two different cardiac impact signal components is to obtain an initial mixed signal; separate a pure BCG cardiac impact signal from the initial mixed signal; select all signal segments with a duration of t in a cardiac impact signal with a duration of T; calculate the correlation coefficient between each signal segment and perform sequential clustering; select a type of signal segment with the highest median mean and average it as a heartbeat template; calculate the correlation coefficient between subsequent signal segments and the heartbeat template and construct a correlation coefficient function; locate the heartbeat position from the correlation coefficient function; and output the beat-by-beat heart rate J according to the heartbeat position sequence.

[0022] Preferably, the method for locating characteristic points I and K is to search forward from point J for the first trough point, which is point I, and search backward for the first trough point, which is point K, and calculate the amplitude difference between J, K and I, J as the third type of relative systolic blood pressure RSBP3.

[0023] Preferably, S2 includes the following steps: fitting the calculated errors of the three models of each subject into three error Gaussian models using the maximum likelihood method; and obtaining parameters of the Bayesian fusion model based on each error Gaussian model.

[0024] Preferably, the fitting error Gaussian model is to fit the same RSBP of N subjects into a Gaussian model, and under the assumption that each RSBP model has the same error distribution for different samples, the error of each RSBP model is fitted into a Gaussian model.

[0025] Preferably, S3 includes the following steps: measuring RSBP1, RSBP2, RSBP3 and the corresponding standard blood pressure of the subject for multiple times, fitting three linear models, and using the obtained parameters of the Bayesian fusion model to perform fusion calculations on the three models.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] The present invention utilizes different signal components in the cardiac impact to extract the pulse transmission time, and simultaneously measures the cardiac impact intensity and the shape characteristics of the cardiac impact signal as the basis for calculating the relative systolic pressure, and utilizes the Bayesian model to fuse the three different indicators to further improve the accuracy of blood pressure calculation. The provided continuous blood pressure measurement method has the characteristics of no load, convenient measurement, low equipment complexity and low cost, and overcomes the disadvantage of large errors that are common in single parameter models, and has great application potential. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0029] Figure 1 It is a flow chart of the continuous blood pressure measurement method based on a single-channel cardiac shock signal of the present invention;

[0030] Figure 2 A schematic diagram of different components of a cardiac impulse signal of the present invention;

[0031] Figure 3 It is a schematic diagram of characteristic points I, J, K and pulse amplitude of the heart ballistic waveform of the present invention;

[0032] Figure 4 It is a schematic diagram of the model of different Gaussian distributions fused through Bayesian in the present invention. DETAILED DESCRIPTION

[0033] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0034] See also Figure 1-4 The present invention provides a continuous blood pressure monitoring method based on a single-channel cardiac shock signal, which comprises:

[0035] S1, collect data from multiple subjects and construct three blood pressure model datasets.

[0036] S2, Calculate the Bayesian fusion model parameters using three blood pressure model datasets.

[0037] S3, the parameters of the Bayesian fusion model are used to fit the blood pressure model and calculate the systolic blood pressure of the subjects.

[0038] The three steps are described in detail below.

[0039] S1, collect data from multiple subjects and construct three blood pressure model data sets. This step is to calculate three BCG parameters for multiple subjects and fit them with the standard blood pressure to obtain three types of systolic blood pressure models, and each model contains data from multiple subjects. This step is divided into the following 6 sub-steps:

[0040] S101: For each subject, obtaining ballistocardiogram information at a specific location using a specific ballistocardiogram measurement method.

[0041] Specifically, the heart shock signal is collected through a specific sensor that can collect human micro-vibration signals, a specific medium between the sensor and the human body, and a specific hardware circuit, a signal sampling rate, etc.; and the sensor needs to be placed in a specific position to obtain a signal that meets the requirements. The reason why there are strict requirements for various heart shock measurement details is that the heart shock diagram is affected by various factors, and only under specific test conditions can an ideal signal be obtained for calculating blood pressure. The front-end sensor used is generally a piezoelectric ceramic sensor, and can also be a piezoelectric film sensor, an optical fiber pad sensor, etc.; the above-mentioned specific medium needs to be a mattress with a certain rigidity, preferably a palm mattress, an integral spring mattress, etc.; the above-mentioned specific hardware circuit preferably has a circuit with noise filtering function, signal amplification function and analog-to-digital conversion function; the above-mentioned specific signal sampling rate is preferably a sampling rate above 250Hz. The specific position of the sensor needs to be placed under the mattress, preferably under the left upper arm of the human body, or under the right upper arm or leg.

[0042] S102: The cardiac shock signal of each subject is processed using two different filters, high frequency and low frequency, to obtain a “limb vibration component” and a “heart vibration component” of the cardiac shock signal.

[0043] See attached Figure 2, this step is to separate the signal components generated by limb vibration and the signal components generated by heart vibration in the heart impact signal through certain processing. According to relevant theories, the micro-vibration signal components of the human body are complex, and mainly include respiratory signals and heart impact signals in a calm state, and the heart impact signal includes the beating of the myocardium, the vibration of the aorta impacted by blood flow, and the vibration of the limb artery impacted by blood flow. According to the signal acquisition method of step S101, the sensor can simultaneously collect the vibration components with lower frequencies (generally 2Hz-6Hz) generated by limb vibration and the vibration components with higher frequencies (generally 8Hz-24Hz) caused by heart beating. In a specific embodiment, this step is to use a Butterworth bandpass filter with a cutoff frequency of [2Hz, 6Hz] to process the heart impact signal to obtain the "limb vibration component", and use a Butterworth bandpass filter with a cutoff frequency of [8Hz, 24Hz] to process the heart impact signal to obtain the "heart vibration component". In addition, as an alternative, the filtering method of using the wavelet method to decompose the heart impact signal and then extracting wavelet components of different levels can also achieve sufficient results.

[0044] S103: for each subject's cardiac impulse signal, find the cardiac impulse characteristic points J of two different cardiac impulse signal components, and calculate the time difference T between the two cardiac impulse characteristic points J as the first relative systolic blood pressure (Relative Systolic Blood Pressure, RSBP), recorded as RSBP1.

[0045] The three systolic blood pressure judgment bases adopted in the present invention all satisfy the linear relationship of SBP=a×RSBPx+b with systolic blood pressure, wherein x is 1-3, a is the proportionality coefficient, and b is the offset. For each type of RSBP and each subject, the values ​​of a and b are different.

[0046] Refer to the attached Figure 2, through the processing of the previous step, we obtained two different signal components. The dotted line indicates the location of the R wave of the electrocardiogram, which serves as a reference for the heartbeat diagram. In this step, it is necessary to first locate the characteristic point J of each heartbeat in the two signals by a certain method to obtain the initial mixed signal; separate the pure BCG heartbeat signal from the initial mixed signal; select all signal segments with a duration of t in a heartbeat signal with a duration of T; calculate the correlation coefficient between each signal segment and perform sequential clustering; select the signal segment with the highest mean value of the median value and average it as the heartbeat template; calculate the correlation coefficient of each subsequent signal segment with the heartbeat template and construct a correlation coefficient function; locate the heartbeat position from the correlation coefficient function; output the beat-by-beat heart rate J according to the heartbeat position sequence, and then calculate the time difference T between point J in the "heart vibration component" and point J in the "limb vibration component". This time difference is the pulse transmission time for blood flow to be transmitted from the heart end to the limb end, and the parameter T is used as the second relative systolic pressure RSBP2.

[0047] S104, for each subject, respectively calculate the ballistocardiogram pulse strength (BPS) of the two heartbeat components and record them as the second type relative systolic blood pressure RSBP2.

[0048] Among them, BPS is obtained by calculating the peak value of the cardiac impact energy signal. Taking the cardiac impact signal with an AD sampling rate of 250hz as an example, the cardiac impact energy signal is calculated according to the following formula:

[0049]

[0050] Among them, X′(ni) represents the value of the cardiac impact signal at the coordinate ni. That is, the cardiac impact signal is squared and then the result of 80-point sliding average filtering is taken.) The peak value of E(n) per heartbeat is BPS, which is used as the second type of relative systolic blood pressure parameter RSBP2.

[0051] S105, for each subject's cardiac signal, locate characteristic points I, J, and K. Calculate the amplitude difference between point JK and point IJ, and record it as relative systolic blood pressure RSBP3.

[0052] Refer to the attached Figure 3 , through the method of S103, the J point of the heart attack signal is located, the first trough point searched forward is point I, and the first trough point searched backward is point K. The amplitude difference between JK and IJ is calculated as the third relative systolic pressure RSBP3.

[0053] S106, for each subject, the systolic blood pressure was measured using a standard sphygmomanometer, and a linear model was fitted with the three RSBPs as independent variables and the standard systolic blood pressure as the dependent variable to obtain three different SBP models.

[0054] Specifically, for each subject, three different SBP models were fitted with relative systolic blood pressure parameters RSBP1, RSBP2, and RSBP3 as independent variables and standard systolic blood pressure SBP as dependent variable. For each subject, three linear models of the following forms were obtained:

[0055] SBPx=a x ×RSBPx+b x

[0056] Among them, SBPx represents the x-th category systolic blood pressure, RSBPx is the x-th category relative systolic blood pressure, and a x , b x are the x-th type systolic blood pressure model coefficients respectively.

[0057] The fitting method used to determine the a and b parameters of each model for each subject is the least squares method, according to the following formula:

[0058]

[0059]

[0060] Here y is the standard systolic blood pressure SBP, and x is the measured RSBP. and are the mean values ​​of each systolic blood pressure and each RSBP for each subject under each type of model.

[0061] After step S1, 3N linear blood pressure models can be generated for N subjects.

[0062] S2: Calculate the Bayesian fusion model parameters using three blood pressure model data sets. Each model calculation step is to fit the three types of models generated by S1 into three Gaussian models by maximum likelihood method, and use the fitted Gaussian model parameters to build the Bayesian fusion model of three RSBP. It is divided into the following two sub-steps:

[0063] S201: The calculated errors of the three models of each subject are fitted into three error Gaussian models using the maximum likelihood method.

[0064] The systolic blood pressure calculation model obtained in step S1 is applied to each data point to obtain the error between the model and the true value. We fit the same RSBP of N subjects into a Gaussian model, based on the assumption that each RSBP model has the same error distribution for different samples. This error of each RSBP model is fitted into a Gaussian model. The maximum likelihood method is used to fit the error of each RSBP model. The specific formula is as follows:

[0065]

[0066]

[0067] Among them, E i is the error of the ith data point, is the expected error, is the variance of the error. Normally, the expectation of the least squares fit is Close to 0, while different SBP calculation models have different variances

[0068] S202: Based on the variance of the errors of the Gaussian models, the SBP calculation models are fused using the Bayesian method.

[0069] Reference Figure 4 , Bayesian information fusion has been proven to be able to fuse two or more models with large measurement variances into a model with smaller measurement errors. The embodiment of the present invention uses the Bayesian model to fuse the three SBP calculation models to obtain a more accurate systolic blood pressure. Specifically, the variance σ of the three SBP error Gaussian models above is obtained. 1 , σ 2 , σ 3 are the variances of the error distributions of the three different models.

[0070] S3: Use the parameters of the Bayesian fusion model to fit the blood pressure model and calculate the systolic blood pressure of the subjects.

[0071] The calculations in steps S1 and S2 are to obtain the Bayesian fusion model. Step S3 is to re-measure the standard blood pressure and relative systolic blood pressure RSBP for a new user to be tested, fit three new linear models, and then use the Bayesian fusion model obtained above to perform data fusion to obtain a more accurate blood pressure value. This step also includes the following two sub-steps:

[0072] S301: Measure RSBP1, RSBP2, RSBP3 and the corresponding standard blood pressure of the subject multiple times according to the aforementioned method, and fit three linear models.

[0073] Following the process of step S1 again, a new subject to be tested performs RSBP1-3 measurement. The difference is that S1 is for generating a data set for calculating the parameters of the Bayesian fusion model, and this step is to achieve continuous dynamic systolic blood pressure measurement of the new subject to be tested.

[0074] The three models generated are:

[0075] SBPx=a x ×RSBPx+b x

[0076] The value of x is 1-3.

[0077] S302: Using the obtained parameters of the Bayesian fusion model, the three models are fused.

[0078] This step improves the accuracy of the calculation results by fusing the calculation results of the three models. The fusion is performed according to the following formula:

[0079] For this embodiment, Bayesian fusion is performed according to the following formula:

[0080]

[0081] in, is the final calculated blood pressure value, σ 1 , σ 2 , σ 3 is a constant, and is the variance of the error distribution of the three different models obtained in step S2, SBP 1 , SBP 2 , SBP 3 The three different RSBPs are respectively calculated by linear models to obtain the human systolic blood pressure (SBP).

[0082] In summary, the present invention utilizes different signal components in the cardiac impact to extract the pulse transmission time, and simultaneously measures the cardiac impact intensity and the shape characteristics of the cardiac impact signal as the basis for calculating the relative systolic blood pressure, and utilizes the Bayesian model to fuse the three different indicators to further improve the accuracy of blood pressure calculation. The provided continuous blood pressure measurement method has the characteristics of no load, convenient measurement, low equipment complexity and low cost, and overcomes the disadvantage of large errors that are common in single parameter models, and has great application and promotion potential.

[0083] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0084] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A continuous blood pressure monitoring method based on a single-channel heartbeat signal, comprising the following steps: S1, collect data from multiple subjects and construct three blood pressure model data sets; calculate three BCG parameters for multiple subjects and fit them with standard blood pressure to obtain three types of systolic blood pressure models, and each model contains data from multiple subjects; Wherein S1 includes the following steps: S10, for each subject, obtaining cardiac ballistic signals at different locations using different cardiac ballistic measurement methods; S11, using two different filters, high frequency and low frequency, to process the cardiac shock signal of each subject, and obtain a limb vibration component and a heart vibration component of the cardiac shock signal; S12, for each subject's cardiac shock signal, find two cardiac shock feature points J of different cardiac shock signal components, and calculate the amplitude difference T of the two cardiac shock feature points J, and use this T as the first type relative systolic pressure RSBP1; S2, calculate the Bayesian fusion model parameters using three blood pressure model datasets; S201: fitting the calculated errors of the three models of each subject into three error Gaussian models using the maximum likelihood method; S202: using the Bayesian method to fuse the SBP calculation models according to the variance of the errors of the Gaussian models; S3, the parameters of the Bayesian fusion model are used to fit the blood pressure model and calculate the systolic blood pressure of the subjects; S301: According to the aforementioned method, the RSBP1, RSBP2, RSBP3 of the measured object and the corresponding standard blood pressure are measured for multiple times, and three linear models are fitted; the three models generated are SBPn=an×RSBPn+bn, where n is 1-3, SBPn represents the nth type of systolic blood pressure, RSBPn is the nth type of relative systolic blood pressure, and an and bn are the nth type of systolic blood pressure model coefficients respectively; The fitting method used to determine the a and b parameters of each model for each subject is the least squares method, according to the following formula: Here y is the standard systolic blood pressure SBP, and x is the measured RSBP. and are the mean of each systolic blood pressure and the mean of each RSBP of each subject under each type of model. After step S1, 3N linear blood pressure models can be generated for N subjects; S302: Using the obtained parameters of the Bayesian fusion model, the three models are fused; the Bayesian fusion is performed according to the following formula: in, To calculate the final blood pressure value, σ1, σ2, and σ3 are constants, which are the variances of the error distributions of the three different models obtained in step S2, and SBP1, SBP2, and SBP3 are the human systolic blood pressure SBP calculated by the linear model using three different RSBPs; S13, for each subject, respectively calculating the records of the two different cardiac impact signal components recorded in S12 as the second type relative systolic blood pressure RSBP2; The heart beat energy signal is calculated according to the following formula: Wherein, X′(ni) represents the value of the heartbeat signal at the coordinate ni, and the peak value of E(n) per heartbeat is BPS, which is used as the second type of relative systolic blood pressure parameter RSBP2; S14, for each subject's cardiac shock signal, locate the characteristic points I, J, K, calculate the difference between the amplitude of JK and the amplitude of IJ, and record it as the third category relative systolic blood pressure RSBP3.

2. The continuous blood pressure monitoring method based on a single-channel heartbeat signal according to claim 1, characterized in that: The S10 includes collecting the heartbeat signal through different sensors capable of collecting human body micro-vibration signals, different media between the sensor and the human body, and different hardware circuits and signal sampling rates; placing different sensors in different parts to obtain signals that meet the requirements.

3. The continuous blood pressure monitoring method based on a single-channel heartbeat signal according to claim 2, characterized in that: The hardware circuit includes a circuit with noise filtering function, signal amplification function and analog-to-digital conversion function; the sampling rate of the signal is above 250Hz, and the different locations where the sensor is placed include under the mattress, under the left upper arm, under the right upper arm or under the leg.

4. The continuous blood pressure monitoring method based on a single-channel heartbeat signal according to claim 1, characterized in that: The specific method for obtaining the limb vibration component and the heart vibration component of the cardiac impact signal in S11 is to use a Butterworth bandpass filter with a cutoff frequency of [2Hz, 6Hz] to process the cardiac impact signal to obtain the limb vibration component, and use a Butterworth bandpass filter with a cutoff frequency of [8Hz, 24Hz] to process the cardiac impact signal to obtain the heart vibration component.

5. The continuous blood pressure monitoring method based on a single-channel heartbeat signal according to claim 1, characterized in that: S2 includes the following steps: the calculation errors of the three models of each subject are fitted into three error Gaussian models using the maximum likelihood method, and the parameters of the Bayesian fusion model are obtained according to each error Gaussian model.

6. The continuous blood pressure monitoring method based on a single-channel heartbeat signal according to claim 5, characterized in that: The maximum likelihood method is to fit the same RSBP of N subjects into a Gaussian model, based on the assumption that each RSBP model has the same error distribution for different samples, and fit the error of each RSBP model into a Gaussian model.

7. The continuous blood pressure monitoring method based on a single-channel heartbeat signal according to claim 1, characterized in that: S3 includes the following steps: measuring RSBP1, RSBP2, RSBP3 and the corresponding standard blood pressure of the subject for multiple times, fitting three linear models, and using the obtained parameters of the Bayesian fusion model to perform fusion calculation on the three models.

Images