A system and method for non-contact remote monitoring of electrostatic potential

Abstract

The present invention provides a non-contact long-distance electrostatic potential monitoring system and method. The system includes a plurality of electrostatic potential sensors arranged around a human body for real-time collection, amplification and transmission of electrostatic signals generated by a moving human body; a computer communicatively connected to the plurality of electrostatic potential sensors for digitally collecting, processing and storing digital signals transmitted by the plurality of electrostatic potential sensors, and inputting the processed data into a trained human body electrostatic potential prediction model to determine the real-time electrostatic potential of the human body; the human body electrostatic potential prediction model is obtained by training a symbolic regression algorithm with measured data. The present invention uses electrostatic potential sensors with high sensitivity, high precision and low noise, and combines a symbolic regression algorithm to realize online monitoring of the human body electrostatic potential, can flexibly adjust the test distance and frequency response characteristics, has strong interpretability of the model form, good simplicity, and low deployment cost.

Description

Technical Field

[0001] This document relates to the technical fields of electrostatic protection engineering and electrostatic potential monitoring technology, and particularly relates to a system and method for non-contact long-distance monitoring of electrostatic potential. Background Art

[0002] During movement, the rapid contact and separation between the human body and objects will cause static charges to continuously accumulate on the body, manifested as an increase in the electrostatic potential. When the electrostatic potential exceeds the breakdown voltage of the surrounding medium (such as air), an electrostatic discharge phenomenon will occur. The electrostatic energy is rapidly released within a few nanoseconds, forming a high voltage, transient large current and accompanied by strong electromagnetic radiation, which will cause serious harm to electrostatic-sensitive substances such as microelectronic devices, flammable and explosive dangerous goods, etc. Therefore, in the fields of medical epidemic prevention, food safety, electronics industry, and energy chemical industry, real-time monitoring of the electrostatic potential of moving human bodies has important value for risk prediction, accident location, attribution analysis, and protection design, etc.

[0003] Existing electrostatic potential measurement methods are divided into contact type and non-contact type. Among them, the contact type method means that the measured charged object is directly in contact with or electrically connected to the input electrode of the electrometer, and the electrometer needs to be electrically connected to the ground to obtain the reference ground (Ground Reference) potential. Therefore, the portability and off-line measurement ability of the contact type method cannot meet the requirements of moving human body testing. The non-contact type method is based on the principle of electrostatic induction or optics, and indirectly obtains the electrostatic potential of the charged object by measuring parameters such as electrostatic field strength. It mainly includes the field mill type and the direct induction type; the field mill type method uses the mechanical movement of the shielding electrode (such as rotating blades, vibrating capacitors or microelectromechanical systems (MEMS)) to modulate the electrostatic field between the sensor induction electrode and the measured charged object, realizing the AC modulation of the DC electrostatic field, and then using a mature weak AC signal amplification circuit (such as phase-sensitive detection) to realize the electrostatic potential test. The mechanical structure of the induction electrode of this type of method is complex and is easily affected by vibration and shock. Usually, the test distance is required to be fixed (within a few centimeters), and the response time is long (about 0.1 second), which cannot meet the requirements of long-distance real-time measurement of human body electrostatic potential.

[0004] The existing detection of human body static electricity potential mainly uses the direct induction method, which modulates the static electric field by the relative displacement between the charged body and the induction electrode, and uses a quasi-direct current amplification circuit with ultra-high input resistance and ultra-low input capacitance to test the static electricity potential of the charged body. For example, in the existing technology methods, such as the sparse low-power sensor network based on EPIC, breathing monitoring within 1.5 m is achieved by testing the human body static electricity potential. For example, an indoor positioning and personnel identification system called "Platypus" is implemented based on EPIC, and a mathematical model for long-distance non-contact testing of human body static electricity potential is proposed based on physical analysis. Six sensors are arranged within a range of 2 m * 2.5 m, and the normalized root mean square error range of testing the human body static electricity potential is 0.07 - 0.16. The testing range and accuracy of the above systems are mainly affected by the performance of the EPIC chip, and the testing range and frequency response characteristics of a single sensor cannot be flexibly adjusted. Only by increasing the number and density of sensors can the testing range and accuracy of the system be improved.

[0005] In view of this, in order to solve the above problems, there is an urgent need to provide a non-contact long-distance monitoring method for static electricity potential with a wider detection scenario and better versatility. Summary of the Invention

[0006] One or more embodiments of this specification provide a non-contact long-distance monitoring system for static electricity potential, including:

[0007] A plurality of static electricity potential sensors are arranged around the human body, and are used for real-time collection, amplification and transmission of the static electricity signals generated by the moving human body;

[0008] A computer is communicatively connected to the plurality of static electricity potential sensors, collects, processes and stores the digital signals transmitted by the plurality of static electricity potential sensors, and inputs the processed data into a trained human body static electricity potential prediction model to determine the real-time static electricity potential of the human body;

[0009] The human body static electricity potential prediction model is obtained by training the symbolic regression algorithm using the actual measured data of the human body static electricity obtained in real time by a plurality of static electricity potential sensors.

[0010] One or more embodiments of this specification provide a non-contact long-distance monitoring method for static electricity potential, including the steps of:

[0011] Real-time collection of the static electricity signals generated by the moving human body through a plurality of static electricity potential sensors arranged around the human body, amplification and processing, and then transmission to the computer;

[0012] The computer collects, processes and stores the input digital signals, and inputs the processed data into a trained human body static electricity potential prediction model to determine the real-time static electricity potential of the human body; wherein,

[0013] The human body static electricity potential prediction model is obtained by training a symbolic regression algorithm with the measured data of the human body static electricity obtained in real time by multiple static electricity potential sensors.

[0014] One or more embodiments of the present specification provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the above-mentioned initial human body static electricity potential prediction model training method and the non-contact long-distance monitoring static electricity potential method are implemented.

[0015] One or more embodiments of the present specification provide a storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned initial human body static electricity potential prediction model training method and the non-contact long-distance monitoring static electricity potential method are implemented.

[0016] The present invention uses a static electricity potential sensor with high sensitivity, high precision, and low noise, and combines a symbolic regression algorithm to realize the online monitoring of the human body static electricity potential. It can flexibly adjust the test distance and frequency response characteristics, has strong interpretability of the model form, good simplicity, and low deployment cost, making the system have a wider application scenario and stronger versatility. Description of the Drawings

[0017] In order to more clearly illustrate the technical solutions in one or more embodiments of the present specification or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments recorded in the present specification. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0018] Figure 1 It is a schematic diagram of the system for non-contact long-distance monitoring of static electricity potential provided by the present invention;

[0019] Figure 2 It is a schematic diagram of the model training system in the system for non-contact long-distance monitoring of static electricity potential provided by the present invention;

[0020] Figure 3 It is a schematic diagram of the principle of the static electricity potential sensor circuit in the system for non-contact long-distance monitoring of static electricity potential provided by the present invention to measure the human body static electricity potential;

[0021] Figure 4 It is a schematic diagram of the data structure of the symbolic regression algorithm in the system for non-contact long-distance monitoring of static electricity potential provided by the present invention;

[0022] Figure 5 It is a comparison diagram before and after the computer provided by the present invention cleans the collected signals, where,

[0023] Figure (a) is the original curve diagram of 5-channel static electricity potential signals collected by a computer, and Figure (b) is the curve diagram after data cleaning of Figure (a);

[0024] Figure 6 Schematic diagram of the test scenario for remote measurement of human static electricity potential in the test provided by the present invention;

[0025] Figure 7 Schematic diagram of the test scenario for remote measurement of human static electricity potential with a positioning system set in the test provided by the present invention;

[0026] Figure 8 Fitting results of the model for training data in the test provided by the present invention. Among them, the upper figure is the fitting result of Model I for the training data, and the lower figure is the fitting result of Model II for the training data;

[0027] Figure 9 Fitting results of the model for test data in the test provided by the present invention. Among them, the upper figure is the fitting result of Model I for the test data, and the lower figure is the fitting result of Model II for the test data;

[0028] Figure 10 Error statistical chart of Model I-II for all data in the test provided by the present invention;

[0029] Figure 11 Distribution diagram of the absolute value of human static electricity potential in the test provided by the present invention;

[0030] Figure 12 Error statistical chart of Model III-IV for all data in the test provided by the present invention;

[0031] Figure 13 Error statistical chart of Model V-VI for all data in the test provided by the present invention;

[0032] Figure 14 Flowchart of the method for non-contact remote monitoring of static electricity potential provided by the present invention;

[0033] Figure 15 Flowchart of the training of the initial human static electricity potential prediction model provided by the present invention;

[0034] Figure 16 Schematic diagram of the framework of the computer device provided by the present invention. Detailed implementation manners

[0035] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the following will clearly and completely describe the technical solutions in one or more embodiments of this specification in conjunction with the accompanying drawings in one or more embodiments of this specification. Obviously, the described embodiments are only a part of the embodiments of this specification, rather than all the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of this document.

[0036] The following will make a detailed description of the present invention in combination with the specific implementation manners and the accompanying drawings of the specification.

[0037] System Embodiment

[0038] According to an embodiment of the present invention, a non-contact long-distance monitoring system for static electricity potential is provided. As Figure 1 shown, it is a schematic structural diagram of the non-contact long-distance monitoring system for static electricity potential provided in this embodiment. The non-contact long-distance monitoring system for static electricity potential according to an embodiment of the present invention includes:

[0039] A plurality of static electricity potential sensors 1 are arranged around the human body, and are used for collecting and amplifying the static electricity signals generated by the moving human body in real time and then transmitting them to the computer;

[0040] The computer is communicatively connected to the plurality of static electricity potential sensors, digitizes, processes, and stores the digital signals transmitted by the plurality of static electricity potential sensors, and inputs the processed data into a trained human body static electricity potential prediction model to determine the real-time static electricity potential of the human body.

[0041] The human body static electricity potential prediction model is obtained by training a symbolic regression algorithm with the actual measured human body static electricity data obtained in real time by a plurality of static electricity potential sensors.

[0042] It should be noted that the plurality of static electricity potential sensors 1 are the static electricity potential sensors disclosed in a non-contact dielectric surface potential detection device and method disclosed in Chinese invention patent CN 11384973 A (publication number). The static electricity potential sensor includes: a static electricity induction signal amplification circuit and an induction electrode. The output end of the static electricity induction signal amplification circuit can be connected to an oscilloscope. The induction electrode is used for inducing the static electricity signal generated on the surface of the moving human body; the static electricity induction signal amplification circuit is used for sampling and processing the induction signal. In this embodiment, the plurality of static electricity potential sensors 1 can be fixedly placed by a six-axis robotic arm in the above-mentioned existing document, or can be arranged on the wall through a mounting bracket.

[0043] When using the electrostatic potential sensor of this patent, the induction electrode is electrically and mechanically non-contact with the human body to be measured, with a distance of more than 1 m. Therefore, the equivalent coupling capacitance between the induction electrode and the human body to be measured is extremely small. In addition, the frequency of the electrostatic potential signal of the human body to be measured is low and is easily interfered by the 50 Hz / 60 Hz power frequency signal. Therefore, the electrostatic potential sensor needs to have both an extremely small input capacitance and an extremely high input resistance to accurately detect the human electrostatic signal. Different from the existing commercial electrostatic potential sensor EPIC, this paper comprehensively uses three typical positive feedback electronic technologies of neutralization, bootstrap, and active protection, combined with shielding, power frequency notch, and programmable amplification technologies to design an electrostatic potential sensor. The principle is as Figure 3 shown, where C e represents the equivalent capacitance between the sensor induction electrode and the human body to be measured, and its size is determined by the size of the induction electrode and its distance from the measured part. C in , C x and R in represent the input capacitance, stray capacitance, and input resistance of the conditioning and amplification circuit respectively. C e and C in , C x form a capacitive voltage divider network, and C e and R in form a first-order high-pass filter, both of which will reduce the sensitivity of the circuit to amplify low-frequency electrostatic induction signals. The neutralization positive feedback circuit feeds the circuit output signal back to the circuit input terminal through the neutralization capacitor C n , thereby reducing the circuit equivalent input capacitance C in . The bootstrap and bias network not only provides a controllable DC leakage channel to prevent amplifier saturation, but also reduces the leakage current at the positive input terminal of the amplifier in the low-frequency signal segment determined by C b , improving the equivalent input circuit. Active protection uses a unity-gain amplifier to drive the shielding case of the induction electrode to reduce the stray capacitance C x , which can minimize the leakage current of the induction electrode and power frequency interference. Compared with the existing EPIC, the advantage of this sensor is that it can flexibly adjust the size of the induction electrode and the frequency response characteristics of the circuit, so as to flexibly adjust the test range and sensitivity of the sensor. And, according to different requirements, amplifiers with different performance parameters can be selected to meet the requirements of the actual application scenario for sensor performance, power consumption, price, and production cycle.

[0044] The system of this embodiment uses an electrostatic potential sensor with high sensitivity, high precision, and low noise, combines a symbolic regression algorithm, and uses a symbolic regression machine learning method to mine the transfer function between the human body's electrostatic potential and the sensor test results from experimental data, realizing the compensation of the test results of the long-distance non-contact human body's electrostatic potential, realizing the detection of the human body's electrostatic potential, and this system can flexibly adjust the test distance and frequency response characteristics, with strong interpretability and good simplicity of the model form, making the system have a wider range of usage scenarios and stronger versatility.

[0045] Preferably, in this embodiment, the initial human body electrostatic potential prediction model training system and process are as follows.

[0046] In this embodiment, the symbolic regression algorithm is used as the initial human body electrostatic potential prediction model for training. Among them, as Figure 2 shown, the model training system includes:

[0047] A plurality of electrostatic potential sensors 1 are arranged around the human body, used for collecting and amplifying the electrostatic signals generated by the moving human body in real time and then transmitting them to the computer;

[0048] The contact electrostatic potential measuring instrument is connected to the human body and the oscilloscope at both ends, used for measuring the electrostatic potential of the moving human body in real time and transmitting it to the oscilloscope;

[0049] The oscilloscope is connected to the computer, used for receiving the human body electrostatic potential measured by the electrostatic potential measuring instrument;

[0050] The computer is respectively communicatively connected to a plurality of electrostatic potential sensors and connected to the oscilloscope, collects, processes, and stores the digital signals output by the oscilloscope and the electrostatic potential sensors, and inputs the processed data into the initial human body electrostatic potential prediction model to train the model until the error between the predicted output response and the target output response of the model is minimized, then the trained human body electrostatic potential prediction model is obtained;

[0051] In this embodiment, the initial human body electrostatic potential prediction model uses a fitness function to measure the output response and the target output response, and uses the root mean square error (RMSE) for calculation;

[0052] Among them, the initial human body electrostatic potential prediction model uses a fitness function to measure the output response and the target output response, and uses the root mean square error for calculation, and the signals of the electrostatic potential sensor path obtained are used as the input of the prediction model, and the signals of the contact electrostatic potential meter path are used as the target output response.

[0053] In this embodiment, the symbolic regression algorithm uses an evolutionary algorithm to search the mathematical expression space, aiming to minimize the error between the measured data and the predicted data of the expression, and automatically updates the mathematical expression optimized by the measured data. Different from other linear or nonlinear regression methods, symbolic regression does not require pre-defining the form of the target expression, can search for both the form and parameters of the mathematical expression simultaneously, greatly reduces the dependence on artificial prior knowledge, expands its applicable range, and provides a flexible and simple method for nonlinear prediction modeling.

[0054] In this embodiment, a multi-gene symbolic regression method is used, which combines the population search ability of the multi-gene genetic algorithm (Multi-Gene Genetic programming, MGGP) and the parameter estimation ability of the linear least squares method to find the optimal linear combination of the mathematical expressions corresponding to all gene individuals in the population, so as to minimize the error between the predicted output response of the mathematical model and the target output response. Among them, the data structure of the multi-gene symbolic regression algorithm can refer to Figure 4 , and the expressions of the initial gene individuals in the genetic population are formed by randomly combining basic mathematical units. Furthermore, new expressions are generated through selection and crossover genetic operations. The fitness function is used to evaluate the fitting ability of all offspring expressions to the experimental data, and the optimal ones are preferentially retained. Repeat the above process until the expression meets the required accuracy or reaches the optimization time, and the algorithm terminates, returning a set of expressions that are most likely to test the internal mechanism of the data. The fitness function is used to measure the response prediction The goodness of fit to the target output variable y is calculated using the root mean square error (RMSE) in this embodiment.

[0055] Preferably in this embodiment, since the electrostatic signals collected during the training process of the model will contain noise due to environmental factors, the computer needs to perform data cleaning operations on the collected electrostatic potential signals before inputting them into the model for training. Specifically, it includes operations such as data filtering, downsampling, and de-offset. The following specifically takes the system with 4 electrostatic potential sensors and an electrostatic potential measuring instrument as an example to illustrate the process of this operation, which can be specifically as Figure 5 shown.

[0056] From Figure 5(a) It can be seen that the static electricity signal during human movement is a low-frequency signal of about 1 Hz. There is 50 Hz power frequency noise and DC offset in the test results, which are due to the low-frequency electromagnetic radiation of the power supply lines in the test chamber, the residual static electricity potential of the human body under test, and the influence of static electricity-carrying objects around. Therefore, it is necessary to perform data cleaning operations such as filtering, downsampling, and offset removal on the collected static electricity data. The filtering operation uses the lowpass digital filtering function lowpass() in MATLAB to perform noise reduction on the recorded data. The cut-off frequency parameter is set to 1 Hz, the steepness parameter is set to 0.9999, and the stopband attenuation parameter is set to 1000 dB. The downsampling operation selects the resample() function in MATLAB, and the target sampling rate parameter is set to 100 Hz, and the sampling method is set to linear resampling. Finally, the DC offset amount is subtracted from the resampled data. The DC offset amount is selected as the mean value of the first 100 sample points of each data time series, that is, the mean value of the body's static electricity potential when the human body under test stands still for 1 s before movement. The data cleaning result is as Figure 5 (b) shown.

[0057] In this embodiment, it should be noted that the used static electricity potential sensor has an analog output interface and a Bluetooth transmission digital output port; when using the digital output port, the analog output signal is directly digitally collected and wirelessly transmitted to the computer by using its own analog-to-digital conversion circuit and Bluetooth digital transmission circuit;

[0058] when using the analog output port, it can be connected to an oscilloscope through a coaxial cable for analog-to-digital conversion, signal acquisition and transmission, and then the oscilloscope is connected to the computer.

[0059] The above technical solutions of the embodiments of the present invention are verified below through specific experimental cases in combination with the accompanying drawings.

[0060] 1. Model training site

[0061] As Figure 6 shown, it is the training scene diagram of the initial symbolic regression model (initial human body static electricity potential prediction model) provided for this case. Among them, in this experiment, a 5m*5m square site is selected on the first floor of a laboratory with a width of 10m and a length of 12m. The floor is made of tiles, there is no grounding body in the test site, and the distance from the surrounding walls or off-site grounding bodies is not less than 1.5m.

[0062] Four electrostatic potential sensors are respectively arranged at the four endpoints of the square test site to measure the electrostatic potential of the moving human body in the test site. The electrostatic sensors are connected to a multi-channel digital storage oscilloscope (PicoScope 4824 from Pico) through coaxial cables. The oscilloscope collects and stores the output signals of the sensors and transmits the data to a computer via USB for data recording. In addition, in order to obtain the accurate value of the human body electrostatic potential, a commercial high-resistance electrometer (6157B from Tek) and a custom spherical capacitor (equivalent capacitance of about 3 pF) are used in this experiment. The electrostatic potential of the moving human body is measured using the capacitance voltage division principle. The custom spherical capacitor is connected to a handheld ultra-wideband positioning tag sensor (UWB_Tag) and the high-resistance electrometer respectively through coaxial cables. Using the analog output function of the electrometer, its analog output signal is also connected to the digital oscilloscope to achieve multi-channel synchronous acquisition of the human body electrostatic potential signal. During the test, the air temperature in the site is 21 degrees Celsius and the relative humidity is 45%.

[0063] For the training of the model, four electrostatic potential sensors and a contact electrostatic potential measuring instrument are used to obtain four-channel signals of the human body moving in the Figure 6 site in real time and input them into the model until the error between the predicted output response and the target output response of the model is minimized, then the trained human body electrostatic potential prediction model is obtained.

[0064] 2. Model Detection Site

[0065] In order to detect the factors affecting the accuracy of the model, this experiment tried two methods to achieve long-distance and non-contact measurement of human body electrostatic potential. They are using ultra-wideband positioning data and not using ultra-wideband positioning data as independent variables, and the measured value of the contact human body electrostatic potential measuring instrument as the target. The settings of the test sites are shown in Figure 6 and Figure 7 respectively. Among them,

[0066] Figure 7 is a test site built with the condition that ultra-wideband positioning data and the electrostatic potential data measured by electrostatic sensors are used as independent variables and the measured value of the contact human body electrostatic potential measuring instrument as the target. On the basis of Figure 6 , one endpoint of the square test site ( Figure 7 the lower left endpoint in Figure 7 ) is defined as the coordinate origin, and the two adjacent endpoints are the X-axis endpoint ( Figure 7At the upper left corner endpoints (in the upper left corner), base station sensors of the ultra-wideband positioning system (UWB Mini 3sPlus of YCHIOT Company) are arranged on these three endpoints. They can be placed at the top of a 1.5-meter-high tripod and are named UWB_Anchor_O, UWB_Anchor_X, and UWB_Anchor_Y respectively. When the test subject holds the ultra-wideband positioning tag sensor (UWB_Tag) and moves, the positioning system outputs the positioning results through UWB_Anchor_O, and the refresh rate is 100Hz. The positioning data is transmitted to the computer through a USB data cable to complete data recording. Among them, the spatial positioning accuracy of the ultra-wideband positioning system is 10cm.

[0067] 3. Test parameters

[0068] Three male participants were invited to participate in the experiment wearing sports shoes (rubber combined with EVA soles). Each time during the test, the participants randomly selected a starting point within the test site and walked slowly (defined as walk) and ran fast (defined as run) ten times each. Whether there is a phenomenon of both feet leaving the ground at the same time was used as the basis for distinguishing walking and running. A total of 60 groups of data were obtained. Each group of data includes the positioning trajectory data (time series signals output by 3 base station sensors) of the participants during the movement and the human body static electricity potential test data (contact measurement time series of a commercial electrometer and non-contact measurement time series of 4-way sensors). The naming method of the test data is that P1W represents the ten groups of data recorded by the first test subject for "walk" ten times, P2R represents the ten groups of data recorded by the second test subject for "run" ten times, and so on.

[0069] 4. Static electricity potential test

[0070] When using the positioning data, the measured value of the static electricity potential sensor and the positioning data are used as the independent variables of the symbolic regression algorithm at the same time, and the measured value of the electrometer is used as the target, hoping to obtain a compensation function from the sensor output to the human body static electricity potential, so as to realize the compensation of the static electricity potential sensor. When not using the positioning data, only the measured value of the static electricity potential sensor is used as the independent variable of the symbolic regression algorithm, and other parameter settings remain unchanged.

[0071] When selecting the test data of "single person single group", "single person multiple groups" or "three people multiple groups" as the training samples respectively, the remaining data is used as the verification sample set.

[0072] "Single person with multiple groups" means selecting three sets of test data of a certain person, and "three persons with multiple groups" means selecting three sets of test data for each of the three persons; and for example, if there are ten sets of "single person with single group" data, when one set is selected as the test data, the other nine sets of data are used as the validation sample set; for each experiment, according to the above method, randomly train the data, set the initial value of the random evolution algorithm while keeping other parameters unchanged, run the symbolic regression algorithm once to obtain a set of optimal models. Repeat the above process ten times.

[0073] The fitness function of the symbolic regression algorithm is the root mean square error. When verifying the effectiveness of the model, the normalized root mean square error (NRMSE) is selected.

[0074]

[0075] The optimal models obtained from training with the "single person with single group" data ten times are as follows:

[0076]

[0077]

[0078] Among them, Ve represents the model estimated value of the human body static electric potential Human Body Electric voltage, and E s1 -E s4 respectively represent the output values of four static electric potential sensors (Electric voltage Sensorouput), and D s1 -D s4 represents the distance from the human body to the four static electric sensors. Model ModelI uses the static electric sensors and positioning data to predict the true static electric potential of the human body, and Model ModelII only uses the test data of the static electric sensors. The fitting results of these two models for the training data are as Figure 8 shown.

[0079] Figure 8 The training data for [the figure] is P1W4, that is, the fourth "walk" test of the first subject. The upper figure shows the compensation result of Model I for the training data, and the NRMSE is equal to 0.0231. The lower figure shows the compensation result of Model II, and the NRMSE is equal to 0.0396. It can be seen that the symbolic regression algorithm has sufficiently good modeling and prediction capabilities to meet the compensation requirements of non-contact static electric potential sensors. The factors affecting the model accuracy lie in the quality of the training data set and the parameter selection of the evolution algorithm. Furthermore, the generalization ability of this model was verified using the remaining data. Figure 9 shows the performance of the model on the P3W5 group of data. The NRMSE of Model I is 0.2357, and the NRMSE of Model II is 0.1622. The statistical results of all data are asFigure 10 as shown

[0080] Figure 10 In it, the horizontal axis scale is the name of the data group, and the vertical axis is the NRMSE distribution diagram of the model on these ten groups of data. It can be seen from the figure that the median of NRMSE of the prediction result of ModelI is between 0.2 and 0.4, and the error of predicting the human body static electricity potential is 20%-40%. Although the training data of Model 1 comes from P1W4, the median error of Model 1 in the P1W group is 0.3, while the error on other groups of data is even smaller, even reaching 0.2. This shows that this model has good generalization ability and is not affected by the personal characteristics (clothing, shoes, stride, step frequency, arm swing, etc.) of Subject 1 and does not enter the overfitting state. Similarly, the training data of Model 2 comes from P3W7. The median of the NRMSE error of Model 2 on the other groups of data except the P1W group is between 0.2 and 0.3, and the error is smaller than that of Model 1. This shows that when monitoring the human body static electricity potential in a long-distance non-contact manner, the distance information from the sensor to the human body does not contribute significantly to the accuracy of the result, but instead increases the complexity of the algorithm and the error source. This is because the relative position information of the four static electricity sensors is already reflected in their output signals, and this information has been implicitly utilized during the symbolic regression modeling.

[0081] The prediction errors of Model 1 (Model I) and Model 2 (Model II) on the P1W group of data are relatively large, which may be related to the absolute value of the human body static electricity potential in this group of data. As Figure 11 shown, the median of the absolute value of the human body static electricity potential in the P1W group is only about 300V, while the static electricity potential values of other groups of data are generally higher than this value.

[0082] The human body potential directly affects the amplitude of the induction signal of the static electricity potential sensor. When the absolute value of the static electricity potential is small, the signal-to-noise ratio of the sensor's induced static electricity signal will inevitably decrease, resulting in an increase in the measurement result error.

[0083] Generally speaking, training data is an important factor affecting the machine learning modeling ability. Next, the influence of different training data samples on the modeling results is further analyzed.

[0084] (1) The optimal symbolic regression model for the ten "single-person multi-group" experiments is:

[0085]

[0086]

[0087] (2) The optimal symbolic regression model for the ten "three-person multi-group" experiments is:

[0088]

[0089]

[0090] Models III and V utilize electrostatic sensor data and distance data, while Models IV and VI only utilize electrostatic sensor data. The verification results of these models on all data are as Figure 12 shown.

[0091] As can be seen from the figure, the median range of NRMSE of Model III on different data sets is 0.15 - 0.5, and the median range of NRMSE of Model IV on different data sets is 0.15 - 0.3; while the NRMSE range of Model I is 0.2 - 0.4, and the NRMSE range of Model II is 0.2 - 0.3; indicating that the prediction errors of Models III and IV decrease on some data, but increase on some data. Moreover, Models III and IV are trained on the P3W group data, but the errors on the P2W, P3W, P1R, P2R, and P3R group data all decrease; indicating that in the "single person, multiple groups" experiment, by increasing the training data, the accuracy of the symbolic regression model is improved. However, it is also found that the median of NRMSE of Models III and IV increases on the P1W group data, which may be due to the single data source causing the "overfitting" phenomenon, enabling the model to learn the motion characteristics of individual moving personnel in the training data, thereby reducing the generalization ability of the model.

[0092] The results of the "multiple people, multiple groups" experiment verify the above analysis, as Figure 13 shown. The median range of NRMSE of Model V is 0.1 - 0.3, and the median range of NRMSE of Model VI is 0.1 - 0.22. The mean and variance of NRMSE of Models V and VI are both lower than those of the previous models. Therefore, the "multiple people, multiple groups" experiment significantly improves the accuracy and generalization ability of the model by introducing differentiated training data. On some samples, the error of the model in predicting the human body's electrostatic potential has been reduced to about 10%, which meets the requirement of the IEC61340 - 4 - 5:2018 standard for the system accuracy of ±10% of the body voltage measuring system, indicating that this method and the prototype test system have good application prospects.

[0093] Preferably in this embodiment, 4 electrostatic potential sensors are arranged in a square at a distance of 5m * 5m to collect the induction signals generated by the human body, so that the overlap of the coverage range of each sensor is relatively small, and non - contact real - time monitoring of the human body's electrostatic potential can be preferably achieved.

[0094] It should be noted that the system and prediction method of this embodiment can be used not only for monitoring the static electricity potential of the human body as described above, but also for predicting the static electricity potential on objects such as animals and equipment according to the monitoring requirements.

[0095] The method provided by the present invention is based on the symbolic regression machine learning method, and proposes a non-contact static electricity potential sensor compensation method. A static electricity potential sensor with high sensitivity, high precision and low noise is designed, and a prototype test system is built to realize non-contact, long-distance monitoring and positioning inversion of the static electricity potential of a moving human body. This method does not need to use the physical information of the actual application site for mathematical modeling, but uses the on-site measured data to train the symbolic regression model, excavates the transfer function from the human body static electricity potential to the sensor signal in the actual scenario, and realizes the compensation of the sensor and the positioning of the personnel. The prototype test system is simple and practical, highly flexible and easy to integrate, so it is convenient to be applied to a large number of application scenarios that require real-time monitoring of the static electricity potential of a moving human body mentioned above. The next step is to study the method to improve the test accuracy of the human body static electricity potential, and expect to meet the requirements of the EC61340-4-5:2018 standard, so as to enhance the practical value of this method.

[0096] Method Embodiment

[0097] According to an embodiment of the present invention, there is provided a non-contact long-distance monitoring static electricity potential method implemented based on the above non-contact long-distance monitoring static electricity potential system, as Figure 14 shown, the non-contact long-distance monitoring static electricity potential method according to an embodiment of the present invention includes the steps:

[0098] S1. Real-time collect the static electricity signals generated by the moving human body through a plurality of static electricity potential sensors arranged around the human body, amplify and process them, and then transmit them to the computer;

[0099] S2. The computer collects, processes, and stores the input digital signals, and inputs the processed data into the trained human body static electricity potential prediction model to determine the real-time static electricity potential of the human body, where

[0100] the human body static electricity potential prediction model is obtained by training the symbolic regression algorithm with the human body static electricity measured data obtained in real time through a plurality of static electricity potential sensors.

[0101] Preferably in this embodiment, as Figure 15 shown, the specific process of training the initial human body static electricity potential prediction model is as follows:

[0102] A101. Real-time collect the static electricity signals generated by the moving human body through a plurality of static electricity potential sensors arranged around the human body, amplify and process them, and then transmit them; measure the static electricity potential of the moving human body in real time through a contact type static electricity potential measuring instrument;

[0103] A102. The oscilloscope receives the human body static electricity potential, performs analog-to-digital conversion and digital transmission on it;

[0104] A103. The computer collects, processes, and stores the digital signals output by the oscilloscope and the static electricity potential sensor, and inputs the processed data into the initial human body static electricity potential prediction model to train the model until the error between the predicted output response of the model and the target output response is minimized, then the trained human body static electricity potential prediction model is obtained.

[0105] In this embodiment, the symbolic regression algorithm uses the evolutionary algorithm to search the mathematical expression space, aiming at minimizing the error between the measured data and the predicted data of the expression, and automatically finds the mathematical expression hidden behind the measured data. Different from other linear or non-linear regression methods, symbolic regression does not need to pre-define the form of the target expression, can search the form and parameters of the mathematical expression simultaneously, greatly reduces the dependence on artificial prior knowledge, expands its application scope, and provides a flexible and simple method for non-linear prediction modeling.

[0106] In this embodiment, the multi-gene symbolic regression method is used, combining the population search ability of the multi-gene genetic algorithm (Multi-Gene Genetic programming, MGGP) and the parameter estimation ability of the linear least squares method to find the optimal linear combination of the mathematical expressions corresponding to all gene individuals in the population, so that the error between the predicted output response of the mathematical model and the target output response is minimized. Among them, the data structure of the multi-gene symbolic regression algorithm can refer to Figure 3 , and the expressions of the initial gene individuals in the genetic population are formed by randomly combining basic mathematical units. Furthermore, new expressions are generated through selection and crossover genetic operations. The fitness function is used to evaluate the fitting ability of all offspring expressions to the experimental data, and the better ones are preferentially retained. Repeat the above process until the expression meets the required accuracy or reaches the optimization time, and the algorithm terminates, returning a set of expressions that are most likely to test the internal mechanism of the data. The fitness function is used to measure the response prediction The goodness of fit to the target output variable y, and in this embodiment, the root mean square error (RMSE) is used for calculation.

[0107] In this embodiment, the specific process of the computer digitizing and processing the induction signal or the human body static electricity potential of the induction signal in step S2 and step A103 is as follows:

[0108] Preferably, in this embodiment, since the electrostatic signals collected during the training of the model may contain noise due to environmental factors, the computer needs to perform data cleaning operations on the obtained electrostatic potential signals before inputting them into the model for training. Specifically, it includes operations such as data filtering, downsampling, and de-offset. Among them, the filtering operation uses the lowpass digital filtering function lowpass() in MATLAB to perform noise reduction on the recorded data. The cut-off frequency parameter is set to 1 Hz, the steepness parameter is set to 0.9999, and the stopband attenuation parameter is set to 1000 dB. The downsampling operation selects the resample() function in MATLAB, and the target sampling rate parameter is set to 100 Hz, and the sampling method is set to linear resampling. Finally, the resampled data is subtracted from the DC offset. The DC offset is selected as the mean of the first 100 sample points of each data time series, that is, the mean body electrostatic potential of the measured human body standing still for 1 s before exercise.

[0109] Preferably, in this embodiment, 4 electrostatic potential sensors are arranged in a square at a distance of 5 m * 5 m to collect the induction signals generated by the human body.

[0110] The embodiment of the present invention is a method embodiment corresponding to the above system embodiment. The specific operations of each processing step can be understood with reference to the description of the method embodiment, and will not be repeated here.

[0111] As Figure 16 shown, the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the initial human body electrostatic potential prediction model training method and the non-contact long-distance monitoring electrostatic potential method in the above embodiment, or when the computer program is executed by the processor, it implements the non-contact long-distance monitoring electrostatic potential method in the above embodiment. When the computer program is executed by the processor, it implements the following method steps:

[0112] S1. Real-time collect the electrostatic signals generated by the moving human body through a plurality of electrostatic potential sensors arranged around the human body, amplify and process them, and then transmit them to the computer;

[0113] S2. The computer collects, processes, and stores the input digital signals, and inputs the processed data into the trained human body electrostatic potential prediction model to determine the real-time electrostatic potential of the human body. Among them,

[0114] The human body electrostatic potential prediction model is obtained by training the symbolic regression algorithm with measured data.

[0115] The human body electrostatic potential prediction model is for the training process:

[0116] A101. Collect the static electricity signals generated by a moving human body in real time through multiple static electricity potential sensors arranged around the human body, amplify and process them, and then transmit them; measure the static electricity potential of the moving human body in real time through a contact-type static electricity potential measuring instrument.

[0117] A102. The oscilloscope receives the static electricity potential of the human body measured in real time by the contact-type static electricity potential measuring instrument, and performs analog-to-digital conversion and digital transmission on it.

[0118] A103. The computer collects, processes, and stores the digital signals output by the oscilloscope and the static electricity potential sensors, and inputs the processed data into the initial human static electricity potential prediction model to train the model until the error between the predicted output response of the model and the target output response is minimized, then the trained human static electricity potential prediction model is obtained.

[0119] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the various embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0120] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the apparatus or system embodiments, since they are basically similar to the method embodiments, the description is relatively simple. For the relevant parts, reference can be made to the description of the method embodiments. The apparatus and system embodiments described above are only illustrative. The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0121] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A non-contact long-distance electrostatic potential monitoring system, characterized in that, Comprising: A plurality of static potential sensors are arranged around the human body for real-time collection, amplification and transmission of static signals generated by a moving human body; The static electricity potential sensor includes a neutralization positive feedback circuit, a bootstrap and bias circuit, and an active protection circuit. The neutralization positive feedback circuit consists of a second operational amplifier, a third resistor R n1 , a fourth resistor R n2 , and a second capacitor C n . The output terminal of the second operational amplifier and the third resistor R n1 are both connected to the positive input terminal of the first operational amplifier through the second capacitor C n . The other end of the third resistor R n1 is connected to the positive input terminal of the second operational amplifier with the fourth resistor R n2 . The other end of the fourth resistor R n2 is grounded. The positive input terminal of the second operational amplifier is connected to the output terminal of the first operational amplifier. The bootstrap and bias circuit includes a series-connected first resistor R b1 and a second resistor R b2 . The other ends of the first resistor R b1 and the second resistor R b2 are respectively connected to the positive input terminal of the first operational amplifier and grounded. The common terminal of the first resistor R b1 and the second resistor R b2 is connected to the negative input terminal and the output terminal of the first operational amplifier through the first capacitor C b . An active protection circuit is provided between the positive input terminal and the output terminal of the first operational amplifier; A computer is communicatively connected to the plurality of static potential sensors to collect, process and store the digital signals transmitted by the plurality of static potential sensors, and input the processed data into a trained human body static potential prediction model to determine the real-time static potential of the human body; The human body static potential prediction model is obtained by training a symbolic regression algorithm using the actual measured data of the human body static electricity obtained in real time by a plurality of static potential sensors.

2. The non-contact long-distance electrostatic potential monitoring system according to claim 1, characterized in that Including a model training system for training an initial human body static potential prediction model, the system comprising: A plurality of static potential sensors are arranged around the human body for real-time collection, amplification and transmission of static signals generated by a moving human body; A contact type static potential measuring instrument, with both ends connected to the human body and an oscilloscope respectively, for real-time measurement of the static potential of a moving human body and transmission to the oscilloscope; An oscilloscope is connected to the computer for receiving the static potential of the human body measured by the static potential measuring instrument, and performing analog-to-digital conversion and digital transmission on it; A computer is communicatively connected to the plurality of static potential sensors and connected to the oscilloscope respectively, for collecting, processing and storing the digital signals output by the oscilloscope and the static potential sensors, and inputting the processed data into the initial human body static potential prediction model to train the model until the error between the predicted output response and the target output response of the model is minimized, then a trained human body static potential prediction model is obtained; Wherein, the initial human body static potential prediction model uses a fitness function to measure the output response and the target output response, and calculates using the root mean square error, the signal of the static potential sensor path obtained is used as the input of the prediction model, and the signal of the contact type static potential meter path is used as the target output response.

3. The non-contact long-distance electrostatic potential monitoring system according to claim 1 or 2, characterized in that, The static potential sensor includes a static induction signal amplification circuit and an induction electrode; The induction electrode is used for inducing the static signal of the moving human body; The static induction signal amplification circuit is used for sampling and processing the induction signal.

4. The non-contact long-distance electrostatic potential monitoring system according to claim 1, characterized in that The computer performs data cleaning on the collected static potential signals, including operations such as data filtering, downsampling and de-offset.

5. A non-contact long-distance method for monitoring electrostatic potential, characterized in that, For the system for non-contact long-distance monitoring of human body static potential according to any one of claims 1 to 4, the method includes: Real-time collecting the static signals generated by the moving human body through a plurality of static potential sensors arranged around the human body, amplifying and processing them and then transmitting them to the computer; The computer collects, processes and stores the input digital signals, and inputs the processed data into a trained human body static potential prediction model to determine the real-time static potential of the human body; wherein, The human body static potential prediction model is obtained by training a symbolic regression algorithm using the actual measured data of the human body static electricity obtained in real time by a plurality of static potential sensors.

6. The non-contact long-distance electrostatic potential monitoring method according to claim 5, characterized in that The specific process of training the initial human body static potential prediction model is as follows: Real-time collecting the static signals generated by the moving human body through a plurality of static potential sensors arranged around the human body, amplifying and processing them and then transmitting; real-time measuring the static potential of the moving human body through a contact type static potential measuring instrument; The oscilloscope receives the human body static electricity potential and performs analog-to-digital conversion and digital transmission on it; Collect, process, and store the digital signals output by the oscilloscope and the static electricity potential sensor, and input the processed data into the initial human body static electricity potential prediction model to train the model until the error between the predicted output response and the target output response of the model is minimized, then the trained human body static electricity potential prediction model is obtained; Among them, the initial human body static electricity potential prediction model uses a fitness function to measure the output response and the target output response, and uses the root mean square error calculation. The signal of the static electricity potential sensor path obtained is used as the input of the prediction model, and the signal of the contact type static electricity potentiometer path is used as the target output response.

7. The non-contact long-distance electrostatic potential monitoring method according to claim 5 or 6, characterized in that The static electricity potential sensor includes a static electricity induction signal amplification circuit and an induction electrode; The induction electrode is used to sense the static electricity signal of the moving human body; The static electricity induction signal amplification circuit is used to sample and process the induction signal.

8. The non-contact long-distance electrostatic potential monitoring method according to claim 5, characterized in that The computer performs data cleaning on the collected static electricity potential signals, including operations such as data filtering, downsampling, and de-offset.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the non-contact long-distance monitoring static electricity potential method according to any one of claims 5 to 8.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the non-contact long-distance monitoring static electricity potential method according to any one of claims 5 to 8.

Images