Estimation method of maximum SAR value of human head based on support vector machine regression model

Abstract

The present invention relates to a method for estimating the maximum SAR value of a human head based on a support vector machine regression model, comprising the steps of arranging a polyhedral measuring body and a plurality of electric field sensors to collect electric field measurement data vectors of a human head with the human head as the center, simulating the polyhedral measuring body and the plurality of electric field sensors, and adjusting the simulated electric field test values ​​under different azimuth angles and elevation angles of the human head and different distances between the electromagnetic radiation source and the human head to obtain a simulated electric field measurement data vector set, and constructing a prediction model of the maximum specific absorption rate value of a human head based on support vector machine regression to achieve the prediction of the maximum specific absorption rate value of a human head. The model established by the method has high applicability to complex space environments and high prediction accuracy.

Description

Technical Field

[0001] The invention relates to the field of bioelectromagnetism, and in particular to a technology for predicting the maximum SAR value of a human head based on support vector machine regression. Background Art

[0002] With the popularity of various electronic products such as mobile phones and computers, people have gradually realized that electromagnetic waves can cause harm to the human body. In order to evaluate the value of electromagnetic radiation received by the human head, people usually use the specific absorption rate to calculate the electromagnetic power absorbed by unit mass of human tissue. This indicator helps to evaluate the potential impact of electromagnetic radiation on health. However, direct experiments to measure the specific absorption rate require a lot of time and resources, so using computer simulation methods to simulate the electromagnetic radiation of mobile phones to the human head has become a more economical and efficient option.

[0003] In order to maintain people's health and safety, all countries have formulated electromagnetic radiation exposure and safety assessment standards adapted to their national conditions to ensure that people's exposure to electromagnetic radiation environments remains within a safe range. my country's national standard "GB21288-2007 Biological Electromagnetic Radiation Protection Standard" issued in 2007 stipulates exposure limits, that is, the average specific absorption rate (SAR) value of any 10 grams of biological tissue in any continuous 6 minutes shall not exceed 2.0W / kg.

[0004] The impact of mobile phones on the human head has always been a hot topic of concern. For example, Yi Miao et al. proposed a calculation method for electromagnetic simulation in the paper "Uncertainty Analysis of Electromagnetic Radiation from Mobile Phones to the Human Head". They described in detail the basic principles and steps of the finite element method, as well as the functions and solution process of the electromagnetic simulation software used. The study also demonstrated the calculation process of the electromagnetic simulation model, including the establishment of the model, the radiation pattern of mobile phone radiation at different frequencies, and the SAR value acting on the human head. By comparing with the electromagnetic environment control limit values ​​in my country, the results show that the SAR values ​​are all below the safety limit values ​​of the human electromagnetic environment, thereby evaluating the safety of the human head being exposed to electromagnetic radiation from mobile phones.

[0005] However, the study did not use a more realistic and complex human head model to accurately reflect the impact of head tissue structure on radiation. At the same time, the study only considered the electromagnetic radiation exposure of the human head when answering the phone, but did not consider the distribution of the SAR value of the mobile phone or antenna at different tilt angles on the human head, as well as the electromagnetic radiation effects at different positions of the head. Summary of the invention

[0006] The purpose of the present invention is to avoid the shortcomings of the prior art and provide a method for estimating the maximum SAR value of the human head based on a support vector machine regression model, which takes into account the distribution of SAR values ​​of the human head at different tilt angles of the mobile phone or antenna, and the electromagnetic radiation effects of various tissues on the head, and can cope with real scenarios and has strong feasibility.

[0007] To achieve the above object, the technical solution adopted by the present invention is: a method for estimating the maximum SAR value of a human head based on a support vector machine regression model, comprising the following steps:

[0008] Step 1: With the human head as the center, a regular polyhedron measuring body is arranged which surrounds the human head and consists of n measuring planes, where n≥6, M electric field sensors are arranged on each measuring plane, and the electromagnetic radiation source to be measured is placed outside the regular polyhedron measuring body;

[0009] When the electromagnetic radiation source to be measured transmits a signal to the human head, the electric field test value is obtained through M electric field sensors to obtain the electric field measurement data vector of the human head. ;

[0010] Step 2: Using electromagnetic simulation software, simulate n measurement planes, nM electric field sensors, electromagnetic radiation sources and the human head, establish a human head electric field simulation model, obtain simulated electric field test values ​​under different azimuth angles and elevation angles of the human head and different distances between the electromagnetic radiation source and the human head by simulating nM electric field sensors, and then obtain a simulated electric field measurement data vector set including the test values ​​of nM electric field sensors and the maximum specific absorption rate test value of the human head;

[0011] Step 3: construct a prediction model for the maximum specific absorption rate value of the human head based on support vector machine regression according to the simulated electric field measurement data vector set;

[0012] Step 4: Transform the measured data vector E mea Substitute the maximum specific absorption rate value of the human head into the above-mentioned prediction model to obtain the predicted maximum specific absorption rate value of the human head.

[0013] Furthermore, the step 1 is specifically as follows:

[0014] Six measuring planes are arranged around the human head, and the six measuring planes form a regular hexahedral measuring body around the human head, and the side length of the regular hexahedral measuring body is 200-400 mm;

[0015] On the six measurement planes, M electric field sensors are evenly arranged, that is, there are 6M electric field sensors in total;

[0016] The electromagnetic radiation source to be measured is placed outside the regular polyhedron measurement body, and the measurement data vectors of 6M electric field sensors are read. .

[0017] Furthermore, the step 2 is specifically as follows:

[0018] Step 21: In electromagnetic simulation software, a cube environment space model with air as the medium is constructed, and a three-dimensional model of a human head is set in the cube environment space model;

[0019] Step 22, with the three-dimensional model of the human head as the center, a simulated regular polyhedron measurement volume is established which surrounds the three-dimensional model of the human head and is composed of n simulated measurement planes; then, M simulated electric field sensors are set on each simulated measurement plane, and finally, simulated electromagnetic radiation sources are arranged in the cubic environmental space model, and the simulated electromagnetic radiation sources are set at any position outside the simulated regular polyhedron measurement volume, so that the electric field simulation model of the human head is obtained;

[0020] Step 23, in the human head electric field simulation model, a coordinate system is established with the geometric center of the human head three-dimensional model as the origin, the central axis of the human head three-dimensional model as the z-axis, and the facial orientation as the x-axis, for adjusting the orientation of the human head three-dimensional model with an azimuth dimension of 0° to 360° and a pitch angle dimension of 0° to 180°;

[0021] The step sizes of the azimuth dimension and the elevation dimension are set, and the test values ​​of nM simulated electric field sensors and the maximum specific absorption rate value of the three-dimensional model of the human head under J different azimuth and elevation conditions are simulated and calculated in the electric field simulation model of the human head, and the simulated data vector samples of the test values ​​of nM simulated electric field sensors and the maximum specific absorption rate test value of the three-dimensional model of the human head are obtained;

[0022] Step 24, change the distance between the simulated electromagnetic radiation source and the three-dimensional model of the human head N times, and repeat step 23 for each distance value, to obtain JN simulated data vectors containing nM simulated electric field sensor test values ​​and human head maximum specific absorption rate test values, which is a simulated electric field measurement data vector set.

[0023] Furthermore, in step 21, the size of the cube environment space model is:

[0024] 1000mm~2000mm×1000mm~2000mm×1000mm~2000mm;

[0025] In the step 22, a simulated hexahedral measuring body consisting of 6 simulated measuring planes is established, and 9 simulated electric field sensors are arranged on each simulated measuring plane, wherein the side length of the simulated hexahedral measuring body is 200 mm to 400 mm;

[0026] In step 23, the orientation of the three-dimensional model of the human head is adjusted in the azimuth dimension and the pitch angle dimension respectively with a step length of 45°, and the simulated electric field sensor test value and the maximum specific absorption rate value of the three-dimensional model of the human head are calculated in sequence under 40 different azimuth and pitch angle conditions, and the data vector samples of 40 simulated electric field sensor test values ​​are obtained as follows:

[0027] ,

[0028] The data vector samples of the maximum specific absorption rate test values ​​of 40 human head 3D models are:

[0029] ,

[0030] Among them, the subscripts represent the azimuth dimension value and the elevation dimension value, E is a data vector composed of 6×9 simulated electric field sensor test values, and y is a data vector of the maximum specific absorption rate test value;

[0031] Then, step 24 specifically involves changing the distance between the simulated electromagnetic radiation source and the three-dimensional model of the human head N times, and repeating step 23 for each distance value, to obtain 40N sample data vectors, which are the simulated electric field measurement data vector set.

[0032] Furthermore, the step three comprises:

[0033] Step 31: Based on the simulated electric field measurement data vector set, a prediction model for the maximum specific absorption rate value of the human head based on support vector machine regression is established. for:

[0034] ,

[0035] In the formula, is the Lagrange multiplier, is the Gaussian kernel function, b is the bias term of the model, is the simulated electric field sensor test value contained in the ith and pth simulated data vectors in the simulated electric field measurement data vector set, and N is the number of times the distance between the simulated electromagnetic radiation source and the three-dimensional model of the human head is adjusted;

[0036] ,

[0037] Where γ is the kernel function attenuation coefficient; is a variable variable, and its optimal value is used to make the prediction model of the maximum specific absorption rate of the human head The highest prediction accuracy;

[0038] Step 32: Solve the prediction model by solving the following minimization optimization problem: Parameters in :

[0039] ,

[0040] In the formula, The simulated maximum specific absorption rate measurement value of the human head contained in the pth simulated data vector in the simulated electric field measurement data vector set is then solved using the Bayesian optimization algorithm to obtain the prediction model Parameter optimization solution.

[0041] Furthermore, in step 31, a prediction model for the maximum specific absorption rate value of the human head based on support vector machine regression is established according to the 40N simulated electric field measurement data vector set. for:

[0042] ,

[0043] Where:

[0044] ,

[0045] Where γ is the kernel function attenuation coefficient;

[0046] In step 32, the following minimization optimization problem is used to solve the parameters in the prediction model: ,

[0047] ,

[0048] In the formula, E p is the simulated electric field sensor test value contained in the pth simulated data vector in the simulated electric field measurement data vector set, y p It is the maximum specific absorption rate measurement value of the human head contained in the pth simulated data vector in the simulated electric field measurement data vector set; that is, a prediction model of the maximum specific absorption rate value of the human head based on support vector machine regression is obtained for a simulated regular hexahedral measurement volume that surrounds the three-dimensional model of the human head and is composed of six simulated measurement planes.

[0049] Furthermore, the step 4 is specifically as follows:

[0050] The electric field measurement data vector E of the human head mea Substitute the prediction model of the maximum specific absorption rate of the human head into the predicted value of the maximum specific absorption rate of the human head to obtain the predicted value of the maximum specific absorption rate of the human head :

[0051] ,

[0052] In the formula, is the predicted maximum specific absorption rate value of the human head; is the Lagrange multiplier obtained by solving the above minimization optimization problem, Ei is the simulated electric field sensor test value contained in the ith simulated data vector in the simulated electric field measurement data vector set, is the bias term of the model obtained by solving the minimization optimization problem; J is the number of conditions for different azimuth and pitch angles of the human head;

[0053] in,

[0054] ,

[0055] In the formula, is the attenuation coefficient of the kernel function obtained by solving the minimization optimization problem.

[0056] The beneficial effects of the present invention are as follows: the present invention proposes a technology for estimating the maximum SAR value of a human head based on incident electric field sampling, adopts support vector machine regression (SVR) and Bayesian optimization method, constructs a machine learning model of the maximum SAR value of a human body based on support vector machine (SVR), and is used to predict the maximum specific absorption rate of a human head in space. The model has high applicability to complex spatial environments and high prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is a schematic diagram of the electric field simulation model of the human head of the present invention;

[0058] Figure 2 is the predicted result of the horn antenna radiation level;

[0059] Figure 3 is the predicted result of the radiation level of the dipole antenna;

[0060] Figure 4 It is the predicted result of the radiation level of the mobile phone antenna. DETAILED DESCRIPTION

[0061] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0062] In order to achieve the above object, the present invention provides the following specific implementation methods:

[0063] Example 1: Figure 1 As shown, a method for estimating the maximum SAR value of a human head based on a support vector machine regression model comprises the following steps:

[0064] S01. With the human head as the center, a regular polyhedron measuring body is arranged which surrounds the human head and consists of n measuring planes, where n≥6, M electric field sensors are arranged on each measuring plane, and the electromagnetic radiation source to be measured is placed outside the regular polyhedron measuring body;

[0065] When the electromagnetic radiation source to be measured transmits a signal to the human head, the electric field test value is obtained through M electric field sensors to obtain the electric field measurement data vector of the human head. ;

[0066] S02. In electromagnetic simulation software, a cubic environment space model with air as the medium is constructed, and a three-dimensional model of a human head is set in the cubic environment space model;

[0067] S03, such as Figure 1 As shown, with the three-dimensional model of the human head as the center, a simulated regular polyhedron measurement volume consisting of n simulated measurement planes surrounding the three-dimensional model of the human head is established; then, M simulated electric field sensors are set on each simulated measurement plane, and finally, simulated electromagnetic radiation sources are arranged in the cubic environmental space model, and the simulated electromagnetic radiation sources are set at any position outside the simulated regular polyhedron measurement volume, so that the electric field simulation model of the human head is obtained;

[0068] S04. In the human head electric field simulation model, a coordinate system is established with the geometric center of the human head three-dimensional model as the origin, the central axis of the human head three-dimensional model as the z-axis, and the facial orientation as the x-axis, so as to adjust the orientation of the human head three-dimensional model with an azimuth dimension of 0° to 360° and a pitch angle dimension of 0° to 180°;

[0069] The step sizes of the azimuth dimension and the elevation dimension are set, and the test values ​​of nM simulated electric field sensors and the maximum specific absorption rate value of the three-dimensional model of the human head under J different azimuth and elevation conditions are simulated and calculated in the electric field simulation model of the human head, and the simulated data vector samples of the test values ​​of nM simulated electric field sensors and the maximum specific absorption rate test value of the three-dimensional model of the human head are obtained;

[0070] S05. Change the distance between the simulated electromagnetic radiation source and the three-dimensional model of the human head N times. For each distance value, repeat step 23 to obtain JN simulated data vectors containing nM simulated electric field sensor test values ​​and human head maximum specific absorption rate test values, which is a simulated electric field measurement data vector set.

[0071] S06. Based on the simulated electric field measurement data vector set, a prediction model for the maximum specific absorption rate value of the human head based on support vector machine regression is established. for:

[0072] ,

[0073] In the formula, is the Lagrange multiplier, is the Gaussian kernel function, b is the bias term of the model, is the simulated electric field sensor test value contained in the ith and pth simulated data vectors in the simulated electric field measurement data vector set, and N is the number of times the distance between the simulated electromagnetic radiation source and the three-dimensional model of the human head is adjusted;

[0074] ,

[0075] Where γ is the kernel function attenuation coefficient; is a variable variable, and its optimal value is used to make the prediction model of the maximum specific absorption rate of the human head The highest prediction accuracy;

[0076] S07. Solve the prediction model by solving the following minimization optimization problem Parameters in :

[0077] ,

[0078] In the formula, The simulated maximum specific absorption rate measurement value of the human head contained in the pth simulated data vector in the simulated electric field measurement data vector set is then solved using the Bayesian optimization algorithm to obtain the prediction model Parameter optimization solution.

[0079] S08, the electric field measurement data vector E of the human head mea Substitute the prediction model of the maximum specific absorption rate of the human head into the predicted value of the maximum specific absorption rate of the human head to obtain the predicted value of the maximum specific absorption rate of the human head :

[0080] ,

[0081] In the formula, is the predicted maximum specific absorption rate value of the human head; is the Lagrange multiplier obtained by solving the above minimization optimization problem, E i is the simulated electric field sensor test value contained in the ith simulated data vector in the simulated electric field measurement data vector set, is the bias term of the model obtained by solving the minimization optimization problem; J is the number of conditions for different azimuth and pitch angles of the human head;

[0082] in,

[0083] ,

[0084] In the formula, is the attenuation coefficient of the kernel function obtained by solving the minimization optimization problem.

[0085] Embodiment 2: A method for estimating the maximum SAR value of a human head based on a support vector machine regression model, comprising the following steps:

[0086] S01. Arrange 6 measurement planes around the human head, the 6 measurement planes form a regular hexahedron measurement body around the human head, and the side length of the regular hexahedron measurement body is 200-400 mm;

[0087] On the six measurement planes, M electric field sensors are evenly arranged, that is, there are 6M electric field sensors in total;

[0088] The electromagnetic radiation source to be measured is placed outside the regular polyhedron measurement body, and the measurement data vectors of 6M electric field sensors are read. .

[0089] S02. In the electromagnetic simulation software, a cube environment space model with air as the medium is constructed, and a three-dimensional model of the human head is set in the cube environment space model. The size of the cube environment space model is:

[0090] 1000mm~2000mm×1000mm~2000mm×1000mm~2000mm;

[0091] S03. With the three-dimensional model of the human head as the center, a simulated regular hexahedron measuring body is established which surrounds the three-dimensional model of the human head and is composed of 6 simulated measurement planes. At the same time, 9 simulated electric field sensors are arranged on each simulated measurement plane, wherein the side length of the simulated regular hexahedron measuring body is 200 mm to 400 mm; simulated electromagnetic radiation sources are arranged in the cubic environmental space model, and the simulated electromagnetic radiation sources are arranged at any position outside the simulated regular multi-faceted measuring body, thus obtaining a human head electric field simulation model;

[0092] S04. In the human head electric field simulation model, a coordinate system is established with the geometric center of the human head three-dimensional model as the origin, the central axis of the human head three-dimensional model as the z-axis, and the facial orientation as the x-axis, so as to adjust the orientation of the human head three-dimensional model with an azimuth dimension of 0° to 360° and a pitch angle dimension of 0° to 180°;

[0093] With a step length of 45°, the orientation of the 3D model of the human head is adjusted in the azimuth dimension and the pitch angle dimension respectively. The simulated electric field sensor test values ​​and the maximum specific absorption rate value of the 3D model of the human head are calculated in turn under 40 different azimuth and pitch angle conditions. The data vector samples of the 40 simulated electric field sensor test values ​​are obtained as follows:

[0094] ,

[0095] The data vector samples of the maximum specific absorption rate test values ​​of 40 human head 3D models are:

[0096] ,

[0097] Among them, the subscripts represent the azimuth dimension value and the elevation dimension value, E is a data vector composed of 6×9 simulated electric field sensor test values, and y is a data vector of the maximum specific absorption rate test value;

[0098] S05. Change the distance between the simulated electromagnetic radiation source and the three-dimensional model of the human head N times. For each distance value, repeat step 23 to obtain 40N sample data vectors containing nM simulated electric field sensor test values ​​and human head maximum specific absorption rate test values, which is a simulated electric field measurement data vector set.

[0099] S06. Based on 40N simulated electric field measurement data vector sets, a prediction model for the maximum specific absorption rate value of the human head based on support vector machine regression is established. for:

[0100] ,

[0101] Where:

[0102] ,

[0103] Where γ is the kernel function attenuation coefficient;

[0104] S07, solve the prediction model by minimizing the following optimization problem Parameters in ,

[0105] ,

[0106] In the formula, E p is the simulated electric field sensor test value contained in the pth simulated data vector in the simulated electric field measurement data vector set, y p It is the maximum specific absorption rate measurement value of the human head contained in the pth simulated data vector in the simulated electric field measurement data vector set; that is, a prediction model of the maximum specific absorption rate value of the human head based on support vector machine regression is obtained for a simulated regular hexahedral measurement volume that surrounds the three-dimensional model of the human head and is composed of six simulated measurement planes.

[0107] S08, the electric field measurement data vector E of the human head mea Substitute the prediction model of the maximum specific absorption rate value of the human head into the predicted value of the maximum specific absorption rate of the human head to obtain the predicted value of the maximum specific absorption rate of the human head:

[0108] ,

[0109] In the formula, is the predicted maximum specific absorption rate value of the human head; is the Lagrange multiplier obtained by solving the above minimization optimization problem, E iis the simulated electric field sensor test value contained in the ith simulated data vector in the simulated electric field measurement data vector set, is the bias term of the model obtained by solving the minimization optimization problem; J is the number of conditions for different azimuth and pitch angles of the human head;

[0110] in,

[0111] ,

[0112] In the formula, is the attenuation coefficient of the kernel function obtained by solving the minimization optimization problem.

[0113] like Figure 1 As shown, the present invention utilizes the specific sampling values ​​described in Example 2, and the electromagnetic simulation software simulates the propagation of electromagnetic radiation, adopts multiple electric field sensors to obtain the electric field data around the human head, and obtains multiple groups of electric field sensor position electric field values ​​and corresponding human head maximum specific absorption rate sample data by changing the relative distance and angle between the electric field sensor and the human body.

[0114] A support vector machine regression model is trained based on the sample data obtained in Example 2. The trained model is deployed in a real-time monitoring system, and the electric field data around the human head is measured by an electric field sensor, and the support vector machine regression model predicts the maximum specific absorption rate value of the human head in real time.

[0115] In order to further illustrate the effect of the present invention, the following experimental examples are provided:

[0116] Experimental example: Figure 2-4 As shown, the advantages of the present invention can be further illustrated by the following simulation experiments:

[0117] 1. Simulation process

[0118] A simulation environment was built in the electromagnetic simulation software. In order to reduce the amount of calculation of the simulation software, a 2m×2m×2m cube was set up, the space was filled with vacuum materials, and a three-dimensional model of the human head was arranged in the room.

[0119] With the three-dimensional model of the human head as the center, a hexahedron surrounding the three-dimensional model of the human head is established. The side length of the hexahedron is 300 mm. Nine electric field measurement sensors are evenly arranged on each surface, that is, a total of 54 electric field measurement sensors. Antennas are arranged indoors as electromagnetic radiation sources, and the antenna operating frequency is 4.9 GHz.

[0120] 2000 sets of training sample data were obtained: a microstrip antenna was used as the electromagnetic radiation source, the distance range between the microstrip antenna and the three-dimensional human body model was (213, 2000) mm, and 50 values ​​were randomly selected within the range as distance sampling values ​​using the Latin hypercube sampling method; for each distance sampling value, the orientation of the three-dimensional human head model was adjusted, the azimuth angle of the three-dimensional human head model was within (0, 360°), and the pitch angle was within (0, 180°), and the three-dimensional human body model was rotated in the azimuth dimension and the pitch angle dimension with a step size of 45°, and the electric field sensor test value and the maximum specific absorption rate value of the human head were calculated in turn.

[0121] Get test sample data: Use horn antenna, dipole antenna and mobile phone antenna as electromagnetic radiation sources, the operating frequency is 4.9 GHz, the distance range between the test antenna and the three-dimensional model of the human body is (213, 2000) mm, and use the Latin hypercube sampling method to randomly select Q values ​​within this range as distance sampling values ​​to obtain Q groups of test sample data.

[0122] Among them, the Q corresponding to the horn antenna is 300, the Q corresponding to the dipole antenna is 200, the azimuth angle of the human head 3D model varies from (0, 360°) to (0, 180°), and the azimuth angle of the human head 3D model is rotated in azimuth and elevation angles with a step length of 45°, and the electric field sensor test value and the maximum specific absorption rate value of the human head are calculated in turn. When the test antenna is a mobile phone antenna, Q=16, and the azimuth and elevation angles of the human head 3D model are fixed at (0°, 0°).

[0123] 2. Simulation results:

[0124] The prediction results of the maximum specific absorption rate value of the human head based on the support vector machine regression model obtained from the training data are shown in Figure 2-Figure 4 , the feasibility, accuracy and generalization ability of using support vector machine regression model to predict maximum specific absorption rate were verified.

[0125] The R² parameters of the support vector machine regression model are as follows:

[0126]

[0127] R² is the coefficient of determination, which measures the model's ability to explain the variability of the dependent variable, that is, the ability of the independent variable to capture the changes in the dependent variable.

[0128] Figure 2 This is the prediction result of the horn antenna radiation level. The predicted value has a good linear relationship with the true value, and most of the points are distributed near the diagonal, indicating that the model's prediction level is high. However, in the area of ​​higher SAR values, the scattered points slightly deviate from the diagonal, indicating that the model still has some errors in predicting high SAR values, and further optimization is needed to improve accuracy.

[0129] Figure 3 This is the prediction result of the dipole antenna radiation level. The predicted value is highly correlated with the true value, and the scatter distribution is more concentrated and close to the diagonal, indicating that the overall prediction effect of the model is excellent. Especially in the high SAR value area, the error is small, indicating that the dipole antenna model can maintain stable prediction capabilities under different radiation levels.

[0130] Figure 4 This is the prediction result of the mobile phone antenna radiation level. In the low SAR value area, there is a good consistency between the predicted value and the true value, but there is a significant deviation in the high SAR value area. Overall, the model's prediction accuracy for the mobile phone antenna radiation level is low, especially in the high value area, which needs further optimization.

[0131] Specifically, it indicates what proportion of the total variability of the dependent variable is caused by changes in the independent variable. During the training phase, the model showed a high degree of correlation, because 96% of the true value variation can be explained by the independent variable; and in the test phase, this correlation can reach 92%, which means that the model not only performs well on the training data, but also maintains a high prediction accuracy on the independent test data, indicating that the model has the ability to accurately predict the maximum specific absorption rate value of the human head and has a strong generalization ability.

[0132] The above numerical simulation experimental results show that the human head maximum SAR value estimation technology based on incident electric field sampling proposed in this patent is feasible and has practical application capabilities.

[0133] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for estimating the maximum SAR value of a human head based on a support vector machine regression model, characterized in that: The steps include: Step 1: With the human head as the center, a regular polyhedron measuring body is arranged which surrounds the human head and consists of n measuring planes, where n≥6, M electric field sensors are arranged on each measuring plane, and the electromagnetic radiation source to be measured is placed outside the regular polyhedron measuring body; When the electromagnetic radiation source to be measured transmits a signal to the human head, the electric field test value is obtained through M electric field sensors to obtain the electric field measurement data vector of the human head. ; Step 2: Using electromagnetic simulation software, simulate n measurement planes, nM electric field sensors, electromagnetic radiation sources and the human head, establish a human head electric field simulation model, obtain simulated electric field test values ​​of the human head at different azimuths and elevations and at different distances between the electromagnetic radiation source and the human head by simulating nM electric field sensors, and obtain a simulated electric field measurement data vector set including the test values ​​of nM electric field sensors and the maximum specific absorption rate test value of the human head; Step 3: Based on the simulated electric field measurement data vector set, a prediction model for the maximum specific absorption rate value of the human head based on support vector machine regression is constructed, specifically: Step 31: Based on the simulated electric field measurement data vector set, a prediction model for the maximum specific absorption rate value of the human head based on support vector machine regression is established. for: ， In the formula, is the Lagrange multiplier, is the Gaussian kernel function, b is the bias term of the model, is the simulated electric field sensor test value contained in the ith and pth simulated data vectors in the simulated electric field measurement data vector set, and N is the number of times the distance between the simulated electromagnetic radiation source and the three-dimensional model of the human head is adjusted; J is the number of conditions for different azimuth and elevation angles of the human head; ， Where γ is the kernel function attenuation coefficient; is a variable variable, and its optimal value is used to make the prediction model of the maximum specific absorption rate of the human head Highest prediction accuracy; Step 32: Solve the prediction model by solving the following minimization optimization problem: Parameters in : ， In the formula, The simulated maximum specific absorption rate measurement value of the human head contained in the pth simulated data vector in the simulated electric field measurement data vector set is then solved using the Bayesian optimization algorithm to obtain the prediction model Parameter optimization solution of ; Step 4: Transform the measured data vector E mea Substitute the maximum specific absorption rate value of the human head into the above-mentioned prediction model to obtain the predicted maximum specific absorption rate value of the human head.

2. The method for estimating the maximum SAR value of a human head based on a support vector machine regression model as claimed in claim 1, characterized in that: The step 1 is specifically as follows: Six measuring planes are arranged around the human head, and the six measuring planes form a regular hexahedron measuring body around the human head, and the side length of the regular hexahedron measuring body is 200-400 mm; On the six measurement planes, M electric field sensors are evenly arranged, that is, there are 6M electric field sensors in total; The electromagnetic radiation source to be measured is placed outside the regular polyhedron measurement body, and the measurement data vectors of 6M electric field sensors are read. .

3. The method for estimating the maximum SAR value of a human head based on a support vector machine regression model as claimed in claim 1, characterized in that: The step 2 is specifically as follows: Step 21: In electromagnetic simulation software, a cube environment space model with air as the medium is constructed, and a three-dimensional model of a human head is set in the cube environment space model; Step 22, with the three-dimensional model of the human head as the center, a simulated regular polyhedron measurement volume is established which surrounds the three-dimensional model of the human head and is composed of n simulated measurement planes; then, M simulated electric field sensors are set on each simulated measurement plane, and finally, simulated electromagnetic radiation sources are arranged in the cubic environmental space model, and the simulated electromagnetic radiation sources are set at any position outside the simulated regular polyhedron measurement volume, so that the electric field simulation model of the human head is obtained; Step 23, in the human head electric field simulation model, a coordinate system is established with the geometric center of the human head three-dimensional model as the origin, the central axis of the human head three-dimensional model as the z-axis, and the facial orientation as the x-axis, for adjusting the orientation of the human head three-dimensional model with an azimuth dimension of 0° to 360° and a pitch angle dimension of 0° to 180°; The step sizes of the azimuth dimension and the elevation dimension are set, and the test values ​​of nM simulated electric field sensors and the maximum specific absorption rate value of the three-dimensional model of the human head under J different azimuth and elevation conditions are simulated and calculated in the electric field simulation model of the human head, and the simulated data vector samples of the test values ​​of nM simulated electric field sensors and the maximum specific absorption rate test value of the three-dimensional model of the human head are obtained; Step 24, change the distance between the simulated electromagnetic radiation source and the three-dimensional model of the human head N times, and repeat step 23 for each distance value, to obtain JN simulated data vectors containing nM simulated electric field sensor test values ​​and human head maximum specific absorption rate test values, which is a simulated electric field measurement data vector set.

4. The method for estimating the maximum SAR value of a human head based on a support vector machine regression model as claimed in claim 3, characterized in that: In step 21, the size of the cube environment space model is: 1000mm~2000mm×1000mm~2000mm×1000mm~2000mm; In the step 22, a simulated hexahedral measuring body consisting of 6 simulated measuring planes is established, and 9 simulated electric field sensors are arranged on each simulated measuring plane, wherein the side length of the simulated hexahedral measuring body is 200 mm to 400 mm; In step 23, the orientation of the three-dimensional model of the human head is adjusted in the azimuth dimension and the pitch angle dimension respectively with a step length of 45°, and the simulated electric field sensor test value and the maximum specific absorption rate value of the three-dimensional model of the human head are calculated in sequence under 40 different azimuth and pitch angle conditions, and the data vector samples of 40 simulated electric field sensor test values ​​are obtained as follows: ， The data vector samples of the maximum specific absorption rate test values ​​of 40 human head 3D models are: ， Among them, the subscripts represent the azimuth dimension value and the elevation dimension value, E is a data vector composed of 6×9 simulated electric field sensor test values, and y is a data vector of the maximum specific absorption rate test value; Specifically, step 24 is to change the distance between the simulated electromagnetic radiation source and the three-dimensional model of the human head N times, and for each distance value, repeat step 23 to obtain 40N sample data vectors, which are the simulated electric field measurement data vector set.

5. The method for estimating the maximum SAR value of a human head based on a support vector machine regression model as claimed in claim 1, characterized in that: In step 31, a prediction model for the maximum specific absorption rate value of the human head based on support vector machine regression is established according to the 40N simulated electric field measurement data vector set. for: ， Where: ， Where γ is the kernel function attenuation coefficient; In step 32, the following minimization optimization problem is used to solve the parameters in the prediction model: , ， In the formula, E p is the simulated electric field sensor test value contained in the pth simulated data vector in the simulated electric field measurement data vector set, y p It is the maximum specific absorption rate measurement value of the human head contained in the pth simulated data vector in the simulated electric field measurement data vector set; that is, a prediction model of the maximum specific absorption rate value of the human head based on support vector machine regression is obtained for a simulated regular hexahedral measurement volume that surrounds the three-dimensional model of the human head and is composed of six simulated measurement planes.

6. The method for estimating the maximum SAR value of a human head based on a support vector machine regression model according to any one of claims 1 to 5, characterized in that: The step 4 is specifically as follows: The electric field measurement data vector E of the human head mea Substitute the prediction model of the maximum specific absorption rate of the human head into the predicted value of the maximum specific absorption rate of the human head to obtain the predicted value of the maximum specific absorption rate of the human head : ， In the formula, is the predicted maximum specific absorption rate value of the human head; is the Lagrange multiplier obtained by solving the above minimization optimization problem, E i is the simulated electric field sensor test value contained in the ith simulated data vector in the simulated electric field measurement data vector set, is the bias term of the model obtained by solving the minimization optimization problem; in, ， In the formula, is the attenuation coefficient of the kernel function obtained by solving the minimization optimization problem.

Images