A linear regression system for remote real-time blood pressure monitoring

Abstract

The application discloses a kind of long-distance blood pressure real-time monitoring linear regression systems, belong to blood pressure measurement field, comprising the following steps: step 1: face is photographed by RGB camera, and the multiple face feature regions are obtained by detecting the face image obtained;Step 2: the IPPG pulse signal detection is carried out to each face feature region, and the multiple superimposed signals are obtained by superimposing the detection signals obtained;Step 3: the phase of each superimposed signal is extracted, and the pulse wave conduction velocity is calculated by the phase of each superimposed signal obtained and the relative position relationship;Step 4: the real-time blood pressure is calculated according to the blood flow rate and blood pressure relationship;The application utilizes multiple sampling signals and linear regression system, so that our system has the characteristics of strong anti-interference, test accuracy and the like.

Description

Technical Field

[0001] This invention belongs to the field of blood pressure measurement, specifically relating to a non-contact blood pressure acquisition system, and particularly to a long-distance real-time blood pressure monitoring linear regression system. Nanjing Yunsichuangzhi Information Technology Co., Ltd. Background Technology

[0002] Blood pressure is an important indicator of cardiovascular health, and currently, blood pressure measurement is still mainly based on traditional contact methods. Traditional methods have the following drawbacks: first, they require specialized equipment, making them unsuitable for on-call monitoring; second, they require operation by professionals, and improper operation can easily lead to errors. These shortcomings limit the widespread adoption of traditional blood pressure measurement methods, especially for daily monitoring in home settings.

[0003] Currently, non-contact blood pressure measurement research has begun both domestically and internationally. Much of this research is based on imaging photoplethysmography (IPPG) to measure the heart's pulse wave, the theoretical basis of which is a linear relationship between blood pressure and pulse wave conduction velocity. A common method involves calculating the conduction time by measuring the phase difference and pulse frequency between two fixed points, thus determining the pulse wave conduction velocity, and then obtaining blood pressure through a linear mapping. This method is computationally complex...

[0004] It is inconvenient to operate and easily affected by noise, and is currently still in the laboratory research stage. Summary of the Invention

[0005] To address the problems of inconvenience in use, inability to detect in real time, and computational complexity and susceptibility to interference in existing technologies, this invention combines facial hotspot region (ROI) detection, signal processing technology, and linear regression models to construct an accurate, robust, and easy-to-use blood pressure monitoring system.

[0006] To achieve the above objectives, the technical solution adopted in this invention is as follows: a long-distance real-time blood pressure monitoring linear regression system, which, based on machine vision methods, obtains the real-time blood pressure of the human body through signal processing analysis and the relationship between pulse signals and blood pressure, including the following steps:

[0007] Step 1: Capture a face using an RGB camera, and detect multiple facial feature regions from the acquired face image;

[0008] Step 2: Perform IPPG pulse signal detection on each facial feature region, and superimpose the detected signals to obtain multiple superimposed signals;

[0009] Step 3: Extract the phase of each superimposed signal, and calculate the pulse wave propagation velocity based on the phase and relative position relationship of each superimposed signal;

[0010] Step 4: Calculate the real-time blood pressure based on the relationship between blood flow velocity and blood pressure.

[0011] Step 1 specifically involves: detecting 468 feature points in the acquired face image using a face feature point detection algorithm; then, taking the surrounding area of ​​each feature point to obtain 468 feature regions, where each surrounding region is a square with a side length of 30*30 centered on the feature point.

[0012] Furthermore, the ROI region is selected based on the course of the facial arteries. The facial arteries extend upwards from both sides of the chin, pass through both sides of the nose, cross the eyebrows, and reach the forehead. Using the facial arteries, the sides of the chin, the origin of the facial arteries, can be selected as the starting area, and relevant facial feature areas along the path of the facial arteries can be considered as important regions.

[0013] Furthermore, a Cartesian coordinate system is established, with the area on both sides of the chin as the starting area, and the vertical position of this area as the origin of the horizontal axis. The vertical distance between other areas and the starting area is used as the coordinate of the X-axis.

[0014] Step 2 specifically involves: along the facial arteries, a series of characteristic regions can be obtained, and pulse signals with the same frequency can be sampled from each region: X1, X2, ..., X... n Where X1 is the initial region signal, then X2, ..., X n The distances to X1 are L1, ..., L n Plotted in a Cartesian coordinate system, where the X-axis represents the distance between the pulse signal and the initial pulse signal, and the Y-axis represents the amplitude, or intensity, of the pulse signal.

[0015] Step 3 specifically involves superimposing the signals from other regions and the starting region to obtain the superimposed signal for each key region. The phase of the signal in each region represents the phase difference with the signal from the starting region. Assuming the starting signal is X1 and the signal from another region is X2, the superimposed signal is X... out A schematic diagram of their interactions is shown below. Figure 4 As shown:

[0016] Figure 4 The signal X is obtained by superimposing signals X1 and X2. out Its expression is as follows:

[0017]

[0018]

[0019]

[0020] Step 4 specifically involves: assuming blood pressure is y and pulse conduction time is x. p Their relationship is expressed by the following formula:

[0021]

[0022] Where a and b are constants, which are related to the elasticity of blood vessel walls in the human body.

[0023] Assuming the distance between X2 and X1 is L, that the same pulse signal travels through X2 and X1, that the time to travel through X1 is t1, and the time to travel through X2 is t2, and that the pulse wave propagation velocity is v, then we have: =L / v

[0024]

[0025]

[0026]

[0027] If the distance between two points is short, then the Taylor first-order expansion can be used to obtain...

[0028]

[0029] As can be seen from the above formula, Blood pressure y is directly proportional to distance L and inversely proportional to pulse wave velocity v. Since conduction time is inversely proportional to velocity, blood pressure y can be expressed in terms of v.

[0030]

[0031] The slope w / v is obtained by regressing the linear regression system. The frequency w can be converted from the heart rate obtained by the IPPG method. Therefore, dividing the slope by w gives 1 / v. Substituting this into the blood pressure calculation formula above will give the blood pressure.

[0032] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention utilizes multiple sampling signals and a linear regression system, thus enabling the system to have:

[0033] 1. Because it is an end-to-end output, the system is simple to implement and easy to promote.

[0034] 2. It performs regression calculations across multiple regions, exhibiting strong anti-interference capabilities and accurate testing.

[0035] 3. Blood pressure is calculated based on pulse wave conduction velocity and signal processing methods, which has mature technical support and strong adaptability. Attached Figure Description

[0036] Figure 1This is a schematic diagram of the linear regression method for real-time long-distance blood pressure monitoring in this embodiment.

[0037] Figure 2 This is a schematic diagram illustrating the selection of facial feature regions in this embodiment.

[0038] Figure 3 This is a schematic diagram of the arteries and veins involved in the human face in this embodiment.

[0039] Figure 4 This is a schematic diagram illustrating the interaction between the two starting signals and the superimposed signal in this embodiment. Detailed Implementation

[0040] The technical solution of the present invention will be further explained below with reference to the accompanying drawings and specific embodiments.

[0041] like Figure 1 As shown in the diagram, this embodiment illustrates a linear regression method for real-time long-distance blood pressure monitoring, which includes the following steps:

[0042] Step 1: Capture a face using an RGB camera, and detect multiple facial feature regions from the acquired face image;

[0043] Step 2: Perform IPPG pulse signal detection on each facial feature region, and superimpose the detected signals to obtain multiple superimposed signals;

[0044] Step 3: Extract the phase of each superimposed signal, and calculate the pulse wave propagation velocity based on the phase and relative position relationship of each superimposed signal;

[0045] Step 4: Calculate the real-time blood pressure based on the relationship between blood flow velocity and blood pressure.

[0046] Further, step 1 specifically involves: detecting 468 feature points in the acquired face image using a facial feature point detection algorithm; then, taking the surrounding area of ​​each feature point to obtain 468 feature regions, where each surrounding region is a square with a side length of 30*30 centered on the feature point.

[0047] like Figure 2 As shown, further, the ROI region is selected based on the course of the facial arteries. The facial arteries extend upwards from both sides of the chin, pass through both sides of the nose, cross the eyebrows, and reach the forehead. Using the facial arteries, the sides of the chin, i.e., the origin of the facial arteries, can be selected as the starting area, and the relevant facial feature areas along the path of the facial arteries are considered as important regions.

[0048] Furthermore, a Cartesian coordinate system is established, with the area on both sides of the chin as the starting area, and the vertical position of this area as the origin of the horizontal axis. The vertical distance between other areas and the starting area is used as the coordinate of the X-axis.

[0049] like Figure 3 As shown, step 2 specifically involves: along the facial arteries, a series of characteristic regions can be obtained, and pulse signals with the same frequency can be sampled from each region: X1, X2, ..., X... n Where X1 is the initial region signal, then the distances between X2, ..., Xn and X1 are L1, ..., L n Plotted in a Cartesian coordinate system, where the X-axis represents the distance between the pulse signal and the initial pulse signal, and the Y-axis represents the amplitude, or intensity, of the pulse signal.

[0050] Step 3 specifically involves superimposing the signals from other regions and the starting region to obtain the superimposed signal for each key region. The phase of the signal in each region represents the phase difference with the signal from the starting region. Assuming the starting signal is X1 and the signal from another region is X2, the superimposed signal is X... out A schematic diagram of their interactions is shown below. Figure 4 As shown:

[0051] Figure 4 The signal X is obtained by superimposing signals X1 and X2. out Its expression is as follows:

[0052]

[0053]

[0054]

[0055] Step 4 specifically involves: assuming blood pressure is y and pulse conduction time is x. p Their relationship is expressed by the following formula:

[0056]

[0057] Where a and b are constants, which are related to the elasticity of blood vessel walls in the human body.

[0058] Assuming the distance between X2 and X1 is L, that the same pulse signal travels through X2 and X1, that the time to travel through X1 is t1, and the time to travel through X2 is t2, and that the pulse wave propagation velocity is v, then we have: =L / v

[0059]

[0060]

[0061]

[0062] If the distance between two points is short, then the Taylor first-order expansion can be used to obtain...

[0063]

[0064] As can be seen from the above formula, Blood pressure y is directly proportional to distance L and inversely proportional to pulse wave velocity v. Since conduction time is inversely proportional to velocity, blood pressure y can be expressed in terms of v.

[0065]

[0066] The slope w / v is obtained by regressing the linear regression system. The frequency w can be converted from the heart rate obtained by the IPPG method. Therefore, dividing the slope by w gives 1 / v. Substituting this into the blood pressure calculation formula above will give the blood pressure.

[0067] Obviously, the above embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the present invention.

Claims

1. A long-distance real-time blood pressure monitoring linear regression system, characterized in that: Includes the following steps: Step 1: Capture a face using an RGB camera, and detect multiple facial feature regions in the acquired face image; Step 2: Perform IPPG pulse signal detection on each facial feature region, and superimpose the detected signals to obtain multiple superimposed signals; Step 3: Extract the phase of each superimposed signal, calculate the time and distance by using the phase and relative position relationship of each superimposed signal, and then calculate the pulse wave conduction velocity; Step 3 specifically involves superimposing the signals from the key regions and the starting region to obtain a superimposed signal for each key region. The phase of the superimposed signal for each key region represents the phase difference with the starting region signal. Assuming the starting signal is X1 and the other region signal is X2, the superimposed signal is X... out The expression for their interaction is as follows: ； ； ； in Represents the frequency of a sinusoidal signal. Indicates phase; Step 4: Calculate the real-time blood pressure based on the relationship between blood flow velocity and blood pressure; Step 4 specifically involves: Assuming blood pressure is y and pulse conduction time is xp, their relationship is expressed by the following formula: ； Where a and b are constants, which are related to the elasticity of blood vessel walls in the human body; Assuming the distance between X2 and X1 is L, that the same pulse signal travels through X2 and X1, that the time to travel through X1 is t1, and the time to travel through X2 is t2, and that the pulse wave propagation velocity is v, then we have: =L / v ； ； ； If the distance between two points is short, it can be obtained using the first-order Taylor expansion: ； As can be seen from the above formula, Blood pressure y is directly proportional to distance L and inversely proportional to pulse wave velocity v. Since conduction time is inversely proportional to velocity, blood pressure y can be expressed in terms of v as follows: ； The slope w / v is obtained by regressing the linear regression system. The frequency w can be converted from the heart rate obtained by the IPPG method. Therefore, dividing the slope by w gives 1 / v. Substituting this into the blood pressure calculation formula above, the real-time blood pressure can be obtained.

2. The long-distance real-time blood pressure monitoring linear regression system according to claim 1, characterized in that: Step 1 specifically involves: detecting the acquired face image using a face feature point detection algorithm to obtain 468 feature points, and taking the surrounding area of ​​each feature point to obtain 468 feature regions. The surrounding area is a square with a side length of 30*30 pixels centered on the feature point.

3. The long-distance real-time blood pressure monitoring linear regression system according to claim 2, characterized in that: The facial feature point detection algorithm selects hotspot regions (ROIs) based on the direction of facial arteries. The facial arteries extend upwards from both sides of the chin, pass through both sides of the nose, cross the eyebrows, and reach the forehead. The algorithm selects the two sides of the chin as the starting area and the areas where the facial arteries intersect with the facial feature regions along their path as key areas.

4. The long-distance real-time blood pressure monitoring linear regression system according to claim 3, characterized in that: Step 2 specifically involves: along the facial arteries, multiple feature regions can be obtained, and each feature region can be sampled to obtain a pulse signal with the same frequency: X1, X2, ..., X... n Where X1 is the initial region signal, then X2, ..., X n The distances to X1 are L1, ..., L n Establish a Cartesian coordinate system with the vertical position of the starting region as the origin of the horizontal axis. The X-axis represents the vertical distance between the pulse signal and the starting pulse signal, and the Y-axis represents the amplitude, i.e., intensity, of the pulse signal.

Images