Method and device for assessing health of personnel in high-risk work environment, electronic equipment and storage medium

Abstract

The application discloses a personnel health assessment method and device in a high-risk operation environment, electronic equipment and a storage medium, and relates to the technical field of health assessment. The method comprises the following steps: collecting real-time physiological health data of an operator in a high-risk operation environment; extracting a feature vector corresponding to the real-time physiological health data, and inputting the feature vector into a preset health risk assessment model, so that the health risk assessment model estimates the probability of the occurrence of the health risk of the operator according to the feature vector, and outputs the probability of the occurrence of the health risk of the operator; comparing the probability with a preset health risk probability threshold, and then determining the health state of the operator according to the comparison result, and generating a corresponding health warning signal when the health state of the operator is a dangerous state. The application can improve the overall safety and work efficiency of the high-risk operation environment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, and in particular to a personnel health assessment method and device in a high-risk operation environment, an electronic device and a storage medium. BACKGROUND

[0002] Safety management in a high-risk operation environment has always been faced with the problems that pre-operation guardians cannot objectively grasp the physical health status of operation personnel, muscle fatigue during operation is difficult to prevent and reduce, and high-risk operation risks are difficult to reduce, which seriously affect the construction of safety enterprises.

[0003] A high-risk operation environment refers to an operation environment with high safety risks under specific conditions, limited by geographical environment, technology and other harmful factors. Such operation environment usually involves dangerous factors such as high altitude, high pressure, flammable, explosive, toxic, radioactive, etc. The characteristics of high-risk operation environment include high risk, high danger, need for special skills, special equipment requirements and multiple risk factors. Therefore, in a high-risk operation environment, the health status of operation personnel usually needs to be managed and controlled. However, the current traditional management and control method usually relies on regular health check of operation personnel or monitoring during work to ensure the safety of operation personnel. However, such method cannot timely understand the health status of operation personnel and cannot respond in time, thereby causing work failure and accident risk due to health reasons of operation personnel, affecting the overall safety and work efficiency of high-risk operation environment. SUMMARY

[0004] The present application provides a personnel health assessment method and device in a high-risk operation environment, an electronic device and a storage medium, to solve the technical problem that the method of regular health check of operation personnel or monitoring during work cannot timely understand the health status of operation personnel and cannot respond in time, affecting the overall safety and work efficiency of high-risk operation environment.

[0005] In order to solve the above technical problem, the present application embodiment provides a personnel health assessment method in a high-risk operation environment, comprising:

[0006] Collecting real-time physiological health data of operation personnel in a high-risk operation environment; wherein the real-time physiological health data includes electrocardiogram data, skin electricity data and electromyogram data;

[0007] Extracting a feature vector corresponding to the real-time physiological health data, and inputting the feature vector into a preset health risk assessment model, so that the health risk assessment model evaluates the probability of occurrence of health risk of operation personnel according to the feature vector, and outputs the probability of occurrence of health risk of operation personnel;

[0008] The probability is compared with a preset health risk probability threshold, and then the health state of the worker is determined according to the comparison result, and when the health state of the worker is a dangerous state, a corresponding health warning signal is generated.

[0009] The health risk assessment model is obtained by training a preset neural network model, wherein the health risk assessment model takes historical physiological health data of the worker as input and takes a probability of a health risk corresponding to the historical physiological health data as output.

[0010] As a preferred solution, the real-time physiological health data of the worker in the high-risk working environment is collected, comprising:

[0011] The real-time activity intensity index and the real-time environmental change index of the worker are obtained.

[0012] The corresponding data collection frequency is calculated according to the real-time activity intensity index and the real-time environmental change index, and then the real-time physiological health data of the worker in the high-risk working environment is collected according to the data collection frequency.

[0013] The calculation formula of the data collection frequency is:

[0014]

[0015] Wherein, f(t) is the data collection frequency at time t; f0 is the initial collection rate; g(A(t)) is a function of activity intensity; k1 is an adjustment coefficient for controlling the influence degree of activity intensity on the collection frequency; A(t) is the activity intensity index of the worker at time t; h(E(t)) is a function of environmental change; k2 is an adjustment coefficient for controlling the influence degree of environmental change on the collection frequency; E(t) is the environmental change index of the worker at time t.

[0016] As a preferred solution, before extracting the feature vector corresponding to the real-time physiological health data, further comprising:

[0017] The ECG data is wavelet transformed and filtered;

[0018] The baseline value of the skin electricity data of the human body in a calm state is obtained, and then the skin electricity data is baseline calibrated according to the baseline value;

[0019] The electromyography data is decomposed into a plurality of independent component signals, and each of the independent component signals is denoised to obtain denoised electromyography data;

[0020] The ECG data after wavelet transformation and filtering, the skin electricity data after baseline calibration, and the electromyography data after denoising are respectively subjected to data normalization processing.

[0021] As a preferred solution, the extracting the feature vector corresponding to the real-time physiological health data comprises:

[0022] respectively extracting electrocardio data features of the electrocardio data, skin electricity data features of the skin electricity data, and electromyography data features of the electromyography data;

[0023] splicing the electrocardio data features, the skin electricity data features, and the electromyography data features to obtain a corresponding joint feature vector, and extracting cross-modal features of each data feature in the joint feature vector;

[0024] integrating the cross-modal features of each data feature in the joint feature vector to obtain a corresponding comprehensive feature vector, and then taking the comprehensive feature vector as the feature vector corresponding to the real-time physiological health data.

[0025] As a preferred solution, the health risk assessment model is:

[0026]

[0027] wherein, UZZq(YCCq=1|FVUE) is the probability of occurrence of health risk given FVUE; w0 is the intercept term; w1, w2, …, w s are weight coefficients corresponding to each dimension of FVUE; FVUE1, FVUE2, …, FVUE s are each element of FVUE.

[0028] As a preferred solution, the training of the health risk assessment model comprises:

[0029] obtaining historical physiological health data of the worker;

[0030] extracting a historical feature vector corresponding to the historical physiological health data, taking the historical feature vector as input and the probability of occurrence of health risk corresponding to the historical physiological health data as output according to a preset loss function, training a preset neural network model, and obtaining the health risk assessment model;

[0031] wherein, the loss function is:

[0032]

[0033] wherein, RT g is the sample number of the training data; ycq cap is the true health risk label of the capth training sample; FVUE capis the integrated feature vector corresponding to the cap-th training sample; PPI is a probability; In is a natural logarithm; LTNNQ(wujj) is a loss function; YIQ is a variable of the health risk assessment model prediction result.

[0034] As a preferred solution, the probability is compared with a preset health risk probability threshold, and then the health state of the worker is determined according to the comparison result, and when the health state of the worker is a dangerous state, a corresponding health warning signal is generated.

[0035] The probability is compared with a preset health risk probability threshold, and when the probability is equal to the health risk probability threshold, it is determined that the health state of the worker is a normal state, and when the probability is not equal to the health risk probability threshold, it is determined that the health state of the worker is a risk state, and a corresponding health warning signal is generated.

[0036] On the basis of the above-mentioned embodiments, another embodiment of the present application provides a personnel health assessment device in a high-risk working environment, comprising: a real-time physiological health data acquisition module, a health risk assessment module and a health warning module.

[0037] The real-time physiological health data acquisition module is used for collecting real-time physiological health data of workers in a high-risk working environment; wherein the real-time physiological health data includes electrocardiogram data, skin electricity data and electromyogram data.

[0038] The health risk assessment module is used for extracting a feature vector corresponding to the real-time physiological health data, and inputting the feature vector into a preset health risk assessment model, so that the health risk assessment model evaluates the probability of occurrence of health risk of the worker according to the feature vector, and outputs the probability of occurrence of health risk of the worker; wherein the health risk assessment model is obtained by training a preset neural network model with historical physiological health data of the worker as input and the probability of occurrence of health risk corresponding to the historical physiological health data as output.

[0039] The health warning module is used for comparing the probability with a preset health risk probability threshold, and then determining the health state of the worker according to the comparison result, and when the health state of the worker is a dangerous state, a corresponding health warning signal is generated.

[0040] On the basis of the above-mentioned embodiments, still another embodiment of the present application provides an electronic device, which comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, and the processor implements the personnel health assessment method in a high-risk working environment according to the above-mentioned embodiments when executing the computer program.

[0041] On the basis of the above-mentioned embodiments, a further embodiment of the present application provides a storage medium comprising a stored computer program, wherein the computer program, when executed, controls a device in which the storage medium is located to perform the personnel health assessment method in a high-risk work environment as described in the above-mentioned embodiments of the present application.

[0042] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0043] The present application provides a personnel health assessment method in a high-risk work environment, real-time physiological health data of workers in a high-risk work environment is collected; wherein the real-time physiological health data includes electrocardiogram data, skin electricity data and electromyography data; the feature vector corresponding to the real-time physiological health data is extracted, and the feature vector is input into a preset health risk assessment model, so that the health risk assessment model evaluates the probability of occurrence of health risk of workers according to the feature vector, and outputs the probability of occurrence of health risk of workers; the probability is compared with a preset health risk probability threshold, and then the health status of workers is determined according to the comparison result, and when the health status of workers is in a dangerous state, a corresponding health warning signal is generated; wherein the health risk assessment model is obtained by training a preset neural network model with historical physiological health data of workers as input and the probability of occurrence of health risk corresponding to the historical physiological health data as output.

[0044] The present application can analyze and evaluate the real-time physiological health data of workers in real time, and through the preset health risk assessment model, the probability of occurrence of health risk of workers can be identified, and the potential risk of the health status of workers can be obtained according to the probability. This real-time monitoring and evaluation capability enables the on-site workers to timely understand their own health status, so as to take appropriate measures before the risk occurs, avoid accidents, achieve the effect of timely response, and protect the life safety of workers in a high-risk work environment. When the health status of workers is in a dangerous state, a corresponding health warning signal can be generated in time, and timely health warning can effectively prevent health problems from further worsening, reduce work errors and accident risks caused by health reasons, and improve the overall safety and work efficiency of the high-risk work environment. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 is a flowchart of a personnel health assessment method in a high-risk work environment provided by an embodiment of the present application;

[0046] Figure 2 is a structural schematic diagram of a personnel health assessment device in a high-risk work environment provided by an embodiment of the present application. DETAILED DESCRIPTION

[0047] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions in the present application will be described clearly and completely below in conjunction with the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all the other embodiments obtained by a person of ordinary skill in the art without creative effort belong to the scope of protection of the present application.

[0048] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of this application; the terms "comprising" and "having," and any variations thereof, as used in the specification and claims and the aforementioned description of the drawings, are intended to cover not exclusive inclusions.

[0049] In the description of the embodiments of the present application, the technical terms "first", "second", etc. are only used to distinguish different objects, and cannot be understood as indicating or implying relative importance or implicitly indicating the number, specific order or primary and secondary relationship of the indicated technical features. In the description of the embodiments of the present application, the meaning of "a plurality of" is two or more, unless otherwise explicitly and specifically limited.

[0050] Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearance of the phrase in various places in the specification does not necessarily all refer to the same embodiment, or necessarily alternatives to other embodiments. It will be explicitly and implicitly appreciated by a person of ordinary skill in the art that the embodiments described herein can be combined with other embodiments.

[0051] In the description of the embodiments of the present application, the term "and / or" is only a description of the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which can represent the three cases of A alone, A and B together, and B alone. In addition, the character " / " in the present application generally represents an "or" relationship between the associated objects.

[0052] In the description of the embodiments of the present application, the term "a plurality of" refers to two or more (including two), and similarly, "a plurality of groups" refers to two or more groups (including two groups), and "a plurality of pieces" refers to two or more pieces (including two pieces).

[0053] In the description of the embodiments of the present application, unless otherwise explicitly specified and limited, the technical terms "mounting", "connection", "linking", "fixing" and the like should be understood in a broad sense, for example, can be fixed connection, can also be detachable connection, or integral; can be mechanical connection, can also be electrical connection; can be direct connection, can also be indirect connection through an intermediate medium, can be internal communication of two elements or interaction relationship between two elements. For those skilled in the art, the specific meanings of the above terms in the embodiments of the present application can be understood according to the specific circumstances.

[0054] Embodiment one

[0055] Please refer to Figure 1 A flowchart of a personnel health assessment method in a high-risk operation environment provided by an embodiment of the present application, comprising the following specific steps:

[0056] S1, collecting real-time physiological health data of operating personnel in a high-risk operation environment; wherein the real-time physiological health data includes electrocardiogram data, skin electricity data and electromyogram data;

[0057] Preferably, the collecting of real-time physiological health data of operating personnel in a high-risk operation environment comprises: obtaining real-time activity intensity indicators and real-time environmental change indicators of operating personnel; calculating corresponding data collection frequency according to the real-time activity intensity indicators and the real-time environmental change indicators, and then collecting real-time physiological health data of operating personnel in a high-risk operation environment according to the data collection frequency;

[0058] The calculation formula of the data collection frequency is:

[0059] f(t)=f0×g(A(t))×h(E(t));

[0060] g(A(t))=1+k1A(t);

[0061] h(E(t))=1+k2E(t);

[0062] Wherein, f(t) is the data collection frequency at time t; f0 is the initial collection rate; g(A(t)) is the function of activity intensity; k1 is an adjustment coefficient for controlling the influence degree of activity intensity on the collection frequency; A(t) is the activity intensity indicator of operating personnel at time t; h(E(t)) is the function of environmental change; k2 is an adjustment coefficient for controlling the influence degree of environmental change on the collection frequency; E(t) is the environmental change indicator of operating personnel at time t.

[0063] Preferably, before extracting the feature vector corresponding to the real-time physiological health data, further comprising: performing wavelet transform and filtering processing on the electrocardiogram data; obtaining a baseline value of the electrodermal data of the human body in a calm state, and then performing baseline calibration on the electrodermal data according to the baseline value; decomposing the electromyography data into a plurality of independent component signals, and performing denoising processing on each of the independent component signals to obtain denoised electromyography data; and performing data normalization processing on the electrocardiogram data after wavelet transform and filtering processing, the electrodermal data after baseline calibration, and the electromyography data after denoising, respectively.

[0064] Specifically, the personnel health assessment method in a high-risk operation environment according to the present application comprises the following specific steps:

[0065] 1. Obtain real-time multi-source heterogeneous data (i.e., the real-time physiological health data) of a smart Internet of Things chest card chip worn by an operating personnel in a high-risk operation environment, wherein the real-time multi-source heterogeneous data comprises electrocardiogram data, electrodermal data, and electromyography data, and the real-time multi-source heterogeneous data is preprocessed. The adaptive sampling frequency technology is used to dynamically adjust the data acquisition frequency according to the activity intensity and environmental changes of the operating personnel, and the historical multi-source heterogeneous data of the smart Internet of Things chest card chip worn by the operating personnel in the high-risk operation environment is obtained.

[0066] Further, for the adaptive sampling frequency technology, the data acquisition frequency is dynamically adjusted according to the activity intensity and environmental changes of the operating personnel, and the adjustment process is as follows:

[0067] The influence of activity intensity and environmental changes on the acquisition frequency is calculated as follows:

[0068] The activity intensity influence is calculated according to the following formula:

[0069] g(A(t))=1+k1A(t);

[0070] In the formula, g(A(t)) represents a function of activity intensity, k1 is an adjustment coefficient for controlling the influence degree of activity intensity on the acquisition frequency, and A(t) represents an activity intensity index of the operating personnel at time t.

[0071] The environmental change influence is calculated according to the following formula:

[0072] h(E(t))=1+k2E(t);

[0073] In the formula, h(E(t)) represents a function of environmental changes, k2 is an adjustment coefficient for controlling the influence degree of environmental changes on the acquisition frequency, and E(t) represents an environmental change index of the operating personnel at time t.

[0074] The final acquisition frequency is:

[0075] f(t) = f0 x g(A(t)) x h(E(t));

[0076] In the formula, f(t) represents the data acquisition frequency at time t, in units of hertz, f0 represents the initial acquisition rate, that is, the default acquisition frequency when the worker just starts to wear the smart Internet of Things chest card chip and does not consider the activity intensity and environmental changes.

[0077] 2. The real-time multi-source heterogeneous data of the worker in the high-risk working environment is subjected to time series segmentation and windowing processing, the continuous real-time multi-source heterogeneous data stream is segmented into fixed-length time window data segments, so as to facilitate subsequent parallel processing and feature extraction. For electrocardiogram data, an adaptive filtering algorithm based on wavelet transform is used to remove noise interference of the electrocardiogram data. For skin electricity data, fractal dimension analysis is used to reveal the complexity of the nonlinear dynamic system, and baseline calibration is performed according to the physiological characteristics of the human body. For electromyography data, an electromyography data denoising technology based on blind source separation is used to decompose the electromyography data into multiple independent components, remove noise and interference components, and thus extract pure electromyography data.

[0078] Further, for the segmentation of the continuous real-time multi-source heterogeneous data stream into fixed-length time window data segments, the segmentation process is as follows:

[0079] The time series of the original real-time multi-source heterogeneous data is set as x(t);

[0080] The real-time multi-source heterogeneous data is segmented into window data segments with a fixed length L:

[0081]

[0082] In the formula, N represents the number of time windows, and T represents the total length of the original physiological data;

[0083] For the n window data segment x n (t), the corresponding time range is:

[0084] t∈[(n-1)L, nL];

[0085] In the formula, n represents a variable identifying the time window sequence number, n = 1, 2, …, N, which is used to identify different time windows, nL represents the time position at which the nth window ends, and (n-1)L represents the time position at which the nth window starts.

[0086] Further, for electrocardiogram data, an adaptive filtering algorithm based on wavelet transform is used to remove noise interference of the electrocardiogram data, and the removal process is as follows:

[0087] The electrocardiogram data in the nth window data segment is set as x n，EC(t), where EC represents electrocardiogram data; and n，EC Wavelet transform is performed on (t), and the transform formula is as follows:

[0088]

[0089] where W n，EC (a, b) represents the coefficient after wavelet transform, a represents a scale parameter, b represents a translation parameter, represents a selected wavelet function, and dt represents an integral operator;

[0090] The adaptive filter is updated based on a least mean square algorithm, and the process is as follows:

[0091] h n，k+1 = h n，k + 2μe n，k x n，EC (t-k);

[0092] where h n，k+1 represents h n，k , h n，k represents the next updated value, h n，k represents the coefficient of the electrocardiogram data in the nth window at the kth filter, μ is a step parameter, e n，EC represents an estimation error, represents the current filter output, x n，EC (t-k) represents the value of the electrocardiogram data in the nth window at time point t-k;

[0093] By continuously updating the parameters of the adaptive filter, the output of the adaptive filter is as close as possible to the original electrocardiogram data x n，EC (t), and is denoised, and finally the denoised electrocardiogram data

[0094] Further, for the skin conductance data, fractal dimension analysis is used to reveal the complexity of the nonlinear dynamic system, and baseline calibration is performed according to the physiological characteristics of the human body, and the analysis process is as follows:

[0095] For a discrete skin conductance data sequence x n，ED (i), where (i = 1, 2, …, M), M is the number of data points in the window, and the calculation steps of the box dimension method are as follows:

[0096] First, a box with a side length of ε is used to cover the space region where the skin conductance data sequence is located;

[0097] The number of boxes covering the skin conductance data sequence is set to M(ε);

[0098] Then, change the side length of the box ε, get different Q(ε), where Q(ε) represents the minimum number of boxes required to cover the sequence of skin conductance data at a specific ε;

[0099] According to the fractal theory, the fractal dimension D satisfies the following relationship:

[0100] Q(ε)~ε -D ;

[0101] By fitting Q(ε) at different ε, and taking the logarithm of Q(ε) and ε, we get logQ(ε) and Qlogε, where log represents the natural logarithm. Nonlinear dynamic systems often exhibit complex and unpredictable behavior. Through the analysis of Q(ε), the complexity can be quantified, and the change of Q(ε) can reveal the behavior of the system at different time scales, helping to understand the dynamic characteristics of the system. By fitting the relationship between Q(ε) and ε, the fractal dimension D can be obtained, which is a numerical indicator of the complexity of the system.

[0102] The linear equation logQ(ε)=-Qlogε+C is obtained by linear regression fitting; in the equation, C represents a constant;

[0103] According to the physiological characteristics of the human body, the skin conductance data is calibrated as a benchmark:

[0104] Set the baseline value of the skin conductance data of the human body in a calm state as B0, and the actual measured skin conductance data as x n，ED (t).

[0105] The calibration formula is:

[0106]

[0107] In the equation, represents the calibrated skin conductance data.

[0108] The skin conductance data (EDA) may have baseline drift due to physiological or environmental factors, that is, the baseline level of the data gradually rises or falls over time. Baseline calibration can remove this drift, making the data more stable and easier to analyze. By calibration, the data collected from different individuals or different time points can be standardized to the same baseline level, which helps comparison and analysis. Calibration can also reduce the impact of noise and improve signal quality, making meaningful physiological changes in skin conductance data more obvious, and making it easier to extract features from calibrated data.

[0109] Further, for electromyography data, an electromyography data denoising technology based on blind source separation is used to decompose the electromyography data into multiple independent components, remove the noise and interference components, and extract pure electromyography data. The extraction process is as follows:

[0110] Set the electromyography data x n，EM (t) is mixed by a plurality of independent source signals S n，v (t), (v = 1, 2, …, KU, ) and KU is the number of source signals;

[0111] The mixing process is represented by the following model:

[0112]

[0113] In the formula, a n，vj represents the element of the mixing matrix;

[0114] The electromyography data is processed by using a blind source separation method:

[0115] The electromyography data x n，EM (t) is centrally processed so that the mean value is zero;

[0116] The separation matrix W n is estimated by iterative calculation so that the output signal y n，v (t) = W n x n，EM (t) is as close as possible to the independent component (i.e. the source signal);

[0117] The iterative formula is:

[0118]

[0119] In the formula, is the updated value of W n , g(y n ) represents a nonlinear function, E{g ACM (y n )} represents an expectation operation, I represents a unit matrix, y n represents an output signal vector, represents the derivative of the output signal vector, represents the expectation value of the product of g(y n ) and ;

[0120] The separation matrix W n is updated by continuous iteration, and the separated independent component signal y n，v (t) is finally obtained; the signal corresponding to the noise and interference component is set as z n，v (t), and the clean electromyography data is represented as:

[0121]

[0122] S2, extract a feature vector corresponding to the real-time physiological health data, and input the feature vector into a preset health risk assessment model, so that the health risk assessment model evaluates a probability of occurrence of a health risk of the worker according to the feature vector, and outputs the probability of occurrence of the health risk of the worker;

[0123] The health risk assessment model is obtained by training a preset neural network model with historical physiological health data of the worker as input and a probability of occurrence of a health risk corresponding to the historical physiological health data as output.

[0124] Preferably, the extracting the feature vector corresponding to the real-time physiological health data comprises: respectively extracting electrocardiogram data features of the electrocardiogram data, skin electricity data features of the skin electricity data, and electromyography data features of the electromyography data; splicing the electrocardiogram data features, the skin electricity data features, and the electromyography data features to obtain a corresponding joint feature vector, and extracting cross-modal features of each data feature in the joint feature vector; integrating the cross-modal features of each data feature in the joint feature vector to obtain a corresponding comprehensive feature vector, and then taking the comprehensive feature vector as the feature vector corresponding to the real-time physiological health data.

[0125] 3. A multi-modal learning framework is introduced to integrate different types of physiological signals to extract cross-modal features.

[0126] Further, for introducing the multi-modal learning framework to integrate different types of physiological signals to extract cross-modal features, the extraction process is as follows:

[0127] Normalizing the real-time data:

[0128]

[0129] In the formula, Xnq Xnq represents the original data, and Xnq min Xnq represents the minimum value in the data set, and Xnq max Xnq represents the maximum value in the data set.

[0130] A convolutional neural network is constructed to extract features of the electrocardiogram data:

[0131] Convolution layer 1:

[0132] The number of input channels is set to 1, the number of output channels is c1, the size of the convolution kernel is EJ1x EJ1, and the step size is Scx1.

[0133] The calculation formula of the convolution operation is:

[0134]

[0135] In the formula, represents the output feature map of the convolutional layer 1, represents the normalized electrocardiogram data matrix as the input of the convolutional layer 1, and Conv1 represents the first convolutional layer in the electrocardiogram data feature extraction network.

[0136] Activation layer 1:

[0137] ReLU activation operation is performed on the output of the convolutional layer 1 to introduce nonlinear features:

[0138]

[0139] In the formula, represents the output feature map after the ReLU activation function processing, represents the output feature map of the convolutional layer 1;

[0140] Pooling layer 1:

[0141] The maximum pooling operation is adopted, and the pooling kernel size is Ej1x Ej1, and the step is Scx2.

[0142]

[0143] In the formula, represents the output feature map of the pooling layer 1, and MaxPool represents the maximum pooling layer, represents the output feature map of the activation layer 1;

[0144] The output of the pooling layer 1 is flattened and connected to a fully connected layer, and the number of neurons of the fully connected layer is fvb1.

[0145]

[0146] In the formula, FE EC represents the output of the fully connected layer, and FC represents the fully connected layer itself, represents the output feature map of the pooling layer 1, and Flatten represents the flattening operation.

[0147] Among them, the extracted electrocardiogram data features are local features, nonlinear features, global features, etc. These features are used for electrocardiogram data classification, anomaly detection or other analysis tasks, and useful features can be learned from the original electrocardiogram data through CNN.

[0148] The electrodermal data feature extraction process is as follows:

[0149] A multi-layer perceptron structure is constructed to extract the features of the electrodermal data:

[0150] Input layer:

[0151] The input dimension is m2, that is, the feature dimension of the preprocessed skin-electricity data;

[0152] Hidden layer 1:

[0153] The number of neurons is mr1, and a ReLU activation function is used:

[0154]

[0155] In the formula, represents the normalized skin-electricity data matrix, represents the output after processing by hidden layer 1 and passing through the ReLU activation function;

[0156] Hidden layer 2:

[0157] The number of neurons is mr2, and a ReLU activation function is also used:

[0158]

[0159] In the formula, APT2 ED represents the output of hidden layer 2;

[0160] Output layer:

[0161] The number of neurons is fvb2;

[0162]

[0163] In the formula, FPF ED represents the output of the output layer, and OL represents the output layer;

[0164] Among them, the time domain features, frequency domain features, time-frequency domain features and nonlinear features of the DEA signal can be extracted by using a multi-layer perception.

[0165] The process of extracting features from electromyographic data is as follows:

[0166] A long short-term memory network is constructed to extract features from electromyographic data:

[0167] LSTM layer:

[0168] The output dimension is mmjz3, and the number of hidden units is lerc1;

[0169]

[0170] In the formula, represents the output of the LSTM layer, represents the normalized electromyographic data matrix;

[0171] Fully connected layer (FC):

[0172] The output of the last time step of the LSTM layer is connected to a fully connected layer, and the number of neurons of the fully connected layer is fvb3:

[0173]

[0174] In the formula, FDD EM is the output of the fully connected layer;

[0175] Wherein, the extracted myoelectric data features are time series features, long-term dependence features and nonlinear features and context features, which can be used for myoelectric signal classification, identification, anomaly detection and other tasks, and through LSTM, useful features can be learned from the original myoelectric data.

[0176] The features extracted from the three different modal data are spliced in a splicing fusion manner to form a joint feature vector:

[0177] FGG fusion = Contee (FE EC , FPF ED , FDD EM );

[0178] In the formula, FGG fusion represents the joint feature vector after splicing, and Contee represents the splicing operation;

[0179] A new fully connected neural network is constructed to further extract cross-modal features:

[0180] Input layer:

[0181] The input dimension is fvb1+ fvb2+ fvb3, that is, the dimension of the joint feature vector after splicing;

[0182] Hidden layer 1 (HL1):

[0183] The number of neurons is cso3, and the ReLU activation function is used:

[0184]

[0185] In the formula, FQQ fusion represents the output of the hidden layer 1, and FGG

[0186] Hidden layer 2 (HL2):

[0187] The number of neurons is cso4, and the ReLU activation function is used:

[0188]

[0189] In the formula, represents the output of the hidden layer 2;

[0190] Output layer (OL):

[0191] The number of neurons is kxc:

[0192]

[0193] In the formula, FQyy caross represents the output of the output layer.

[0194] where the cross-modal features refer to features extracted and fused from multiple different types of physiological signals, which can provide more comprehensive information than a single modality. Cross-modal features include: electrocardiogram features, skin electricity features, electromyography features, time series features, nonlinear features, context features, and fusion features. These cross-modal features can be used for monitoring and analyzing various physiological states, such as emotion recognition, stress level assessment, and exercise monitoring. By fusing data from different modalities, the accuracy and robustness of the model can be improved, as different modalities may exhibit different sensitivities under different physiological states.

[0195] Preferably, the health risk assessment model is:

[0196]

[0197] where UZZq(YCCq=1|FVUE) is the probability of health risk occurrence given FVUE; w0 is the intercept term; w1, w2, …, w s are the weight coefficients corresponding to each dimension of FVUE; FVUE1, FVUE2, …, FVUE s are the elements of FVUE.

[0198] Preferably, the training of the health risk assessment model includes: obtaining historical physiological health data of workers; extracting a historical feature vector corresponding to the historical physiological health data; according to a pre-set loss function, taking the historical feature vector as input and the probability of health risk occurrence corresponding to the historical physiological health data as output, training a pre-set neural network model to obtain the health risk assessment model; wherein the loss function is:

[0199]

[0200] where RT g is the number of samples of the training data; ycq cap is the true health risk label of the cap-th training sample; FVUE capFVUE is the integrated feature vector corresponding to the cap training sample; PPI is probability; In is natural logarithm; LTNNQ(wujj) is a loss function; YIQ is a variable of the prediction result of the health risk assessment model.

[0201] 4. A health risk assessment model is constructed by using big data and machine learning algorithm to evaluate the probability of health risk occurrence of the workers.

[0202] Further, for the health risk assessment model constructed by using big data and machine learning algorithm, the construction process is as follows:

[0203] The joint feature vector after splicing is integrated into a comprehensive feature vector FVUE:

[0204] FVUE = [FQyy caross ];

[0205] A health risk assessment model is constructed by using a logistic regression algorithm:

[0206]

[0207] In the formula, UZZq(YCCq = 1 | FVUE) represents the probability of health risk occurrence given FVUE, w0 represents the intercept term, w1, w2, …, w s represent the weight coefficients corresponding to each dimension of FVUE, FVUE1, FVUE2, …, FVUE s represent each element of FVUE;

[0208] The health risk assessment model is trained by historical multi-source heterogeneous data:

[0209]

[0210] In the formula, RT g represents the number of samples of the training data, ycq cap represents the true health risk label of the cap training sample, FVUE cap represents the integrated feature vector corresponding to the cap training sample, PPI represents probability, In represents natural logarithm, LTNNQ(wujj) represents a loss function, and YIQ represents a variable of the prediction result of the health risk assessment model.

[0211] The gradient descent method is used to update the weight parameters so that the loss function LTNNQ(wujj) reaches a minimum value, and the formula is as follows:

[0212]

[0213] In the formula, wujj jsxis the jth weight parameter of the jth weight vector, represents a learning rate for controlling the step size of each update, represents the partial derivative of the loss function LTNNQ(wujj) with respect to the weight parameter wujj jsx .

[0214] After the health risk assessment model is trained, the real-time multi-source heterogeneous data is extracted through the above steps to obtain a feature vector FVUE new .

[0215] The FVUE new is brought into the trained health risk assessment model, and the probability of the occurrence of the health risk of the worker is calculated as UZZq(YCCq=1|FVUE new ).

[0216] S3, compare the probability with a preset health risk probability threshold, and then determine the health status of the worker according to the comparison result, and generate a corresponding health warning signal when the health status of the worker is a dangerous state.

[0217] Preferably, the probability is compared with a preset health risk probability threshold, and then the health status of the worker is determined according to the comparison result, and a corresponding health warning signal is generated when the health status of the worker is a dangerous state. The method comprises: comparing the probability with a preset health risk probability threshold, and determining the health status of the worker as normal when the probability is equal to the health risk probability threshold, and determining the health status of the worker as a risk state when the probability is not equal to the health risk probability threshold, and generating a corresponding health warning signal.

[0218] 5. When the health risk of the worker reaches a certain threshold, a health warning is triggered.

[0219] The health risk probability threshold is set to

[0220] If the health of the worker is determined to be a high-risk state, a first-level warning is issued;

[0221] If the health of the worker is determined to be normal, no warning is issued;

[0222] If the health of the worker is determined to be a low-risk state, a second-level warning is issued;

[0223] When the first-level warning is issued, the intelligent Internet of Things chest card chip emits an alarm sound, and the alarm sound is long and reminds the worker;

[0224] When the secondary early warning is given, the intelligent Internet of Things chest card chip sends out an alarm sound, and the alarm sound intermittently reminds the operating personnel.

[0225] Therefore, the personnel health assessment method in a high-risk operation environment is provided, and the following beneficial effects can be achieved by the method:

[0226] Through the adaptive sampling frequency technology, the data acquisition frequency can be flexibly adjusted according to the actual activity intensity of the operating personnel and the environmental changes, the frequency is automatically increased when the operating personnel performs high-intensity physical labor or the environmental parameters change sharply, more accurate real-time data is ensured, a reliable basis is provided for subsequent analysis and evaluation, which helps to timely find the changes of the physiological indicators of the operating personnel in special conditions, improves the monitoring ability of potential risks, appropriately reduces the sampling frequency in a relatively stable state, saves energy, reasonably utilizes the data storage space, and prolongs the service life of the intelligent Internet of Things chest card chip.

[0227] Through targeted processing of different types of real-time multi-source heterogeneous data, such as removing noise interference by using a wavelet transform-based adaptive filtering algorithm for electrocardio data, the accuracy and readability of the electrocardio data can be improved, purer data is provided for subsequent heart health analysis, and the fractal dimension analysis and baseline calibration of the skin electricity data help to reveal the complexity of the nonlinear dynamic system and make it more consistent with the actual human physiological conditions, better reflect the physiological state changes of the operating personnel, and the electromyography data denoising technology based on blind source separation can extract pure electromyography data, provide accurate basis for muscle fatigue and injury evaluation, and these processing methods make the physiological data more valuable, and lay a solid foundation for subsequent comprehensive analysis and risk assessment.

[0228] The multi-modal learning framework is introduced, different types of physiological signals such as electrocardio, skin electricity and electromyography are integrated, and cross-modal features are extracted, the integration can comprehensively and comprehensively reflect the body state of the operating personnel, and overcome the limitations of single physiological signal analysis, through extraction of cross-modal features, the correlation and synergistic effect between different physiological signals can be captured, the body adaptation and potential health problems of the operating personnel in a high-risk operation environment can be better understood, and more rich and accurate information is provided for health risk assessment.

[0229] The health risk assessment model is constructed by using big data and a machine learning algorithm, and can analyze and evaluate the physiological data of the workers in real time. Through learning and analyzing a large amount of multi-source heterogeneous data, the model can accurately identify the change trend and potential risk of the health state of the workers. The real-time monitoring and evaluation capability enables the workers on site to know their health state in time, so that corresponding measures can be taken before the risk occurs, accidents are avoided, the effect of timely response is achieved, and the life safety of the workers in the high-risk operation environment is ensured. When the health risk of the workers reaches a certain threshold, a health warning can be triggered in time. The timely health warning can effectively prevent the health problem from further deterioration, reduce work mistakes and accident risks caused by health reasons, and improve the overall safety and work efficiency of the high-risk operation environment.

[0230] Embodiment Two

[0231] Please refer to Figure 2 A personnel health assessment device in a high-risk operation environment provided by an embodiment of the present application, which comprises a real-time physiological health data acquisition module, a health risk assessment module and a health warning module.

[0232] The real-time physiological health data acquisition module is used to collect real-time physiological health data of workers in a high-risk operation environment. The real-time physiological health data includes electrocardiogram data, skin electricity data and electromyogram data.

[0233] The health risk assessment module is used to extract a feature vector corresponding to the real-time physiological health data, and input the feature vector into a preset health risk assessment model, so that the health risk assessment model evaluates the probability of occurrence of the health risk of the workers according to the feature vector, and outputs the probability of occurrence of the health risk of the workers. The health risk assessment model is obtained by training a preset neural network model with historical physiological health data of the workers as input and the probability of occurrence of the health risk corresponding to the historical physiological health data as output.

[0234] The health warning module is used to compare the probability with a preset health risk probability threshold, and then determine the health state of the workers according to the comparison result. When the health state of the workers is a dangerous state, a corresponding health warning signal is generated.

[0235] It should be noted that the apparatus embodiments described above are only illustrative, and the units described as separate components can or can not be physically separated, and the components displayed as units can or can not be physical units, i.e. can be located in one place or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment according to actual needs. In addition, the connection between the modules in the apparatus embodiment provided by the present application indicates that there is a communication connection between them, which can be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement it without creative labor.

[0236] Those skilled in the art can clearly understand that, for the convenience and brevity, the specific working process of the above-described apparatus can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0237] Embodiment three

[0238] Correspondingly, the embodiment of the present application provides an electronic device, the device includes a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, when the processor executes the computer program, the personnel health assessment method in high-risk working environment described in the above application embodiment is realized.

[0239] The electronic device can be a desktop computer, a notebook computer, a palm computer, a cloud server and other computing devices. The device can include but is not limited to a processor and a memory.

[0240] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc. The processor is the control center of the device, which connects all parts of the device through various interfaces and lines.

[0241] Embodiment four

[0242] Correspondingly, the embodiment of the present application provides a storage medium, the storage medium comprising a stored computer program, wherein the computer program controls a device where the storage medium is located to perform the personnel health assessment method in a high-risk work environment when the computer program is running.

[0243] The memory can be used to store the computer program, and the processor realizes various functions of the device by running or executing the computer program stored in the memory and calling data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required by a function, and the like; and the data storage area can store data created according to the use of the mobile phone and the like. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, for example, a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state memory devices.

[0244] The storage medium is a computer readable storage medium, and the computer program is stored in the computer readable storage medium. The computer program can realize the steps of each method embodiment when executed by the processor. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or some intermediate forms, etc. The computer readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the contents included in the computer readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction, for example, in some jurisdictions, according to legislation and patent practice, the computer readable medium does not include electrical carrier signals and telecommunication signals.

[0245] The above is the preferred embodiment of the present application. It should be noted that for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which are also considered within the scope of protection of the present application.

Claims

1. A method for assessing the health of personnel in a high-risk work environment, characterized in that, The method comprises the following steps: obtaining real-time activity intensity indicators and real-time environmental change indicators of workers; calculating corresponding data collection frequencies according to the real-time activity intensity indicators and the real-time environmental change indicators, and then collecting real-time physiological health data of workers in a high-risk work environment according to the data collection frequencies; wherein the real-time physiological health data includes electrocardiogram data, skin electricity data, and electromyography data; extracting electrocardiogram data features of the electrocardiogram data, skin electricity data features of the skin electricity data, and electromyography data features of the electromyography data, respectively; splicing the electrocardiogram data features, the skin electricity data features, and the electromyography data features to obtain a corresponding joint feature vector, and extracting cross-modal features of each data feature in the joint feature vector; integrating the cross-modal features of each data feature in the joint feature vector to obtain a corresponding comprehensive feature vector FVUE, then taking the comprehensive feature vector FVUE as a feature vector corresponding to the real-time physiological health data, and inputting the feature vector into a preset health risk assessment model to enable the health risk assessment model to evaluate the probability of occurrence of a health risk of workers according to the feature vector, and output the probability of occurrence of a health risk of workers; comparing the probability with a preset health risk probability threshold, and then determining the health status of workers according to the comparison result, and generating a corresponding health warning signal when the health status of workers is in a dangerous state; wherein the health risk assessment model is obtained by training a preset neural network model with historical physiological health data of workers as input and the probability of occurrence of a health risk corresponding to the historical physiological health data as output; the calculation formula of the data collection frequency is: ； ； ； wherein, is the data acquisition frequency at time t; is the initial acquisition frequency; is a function of the activity intensity; is an adjustment coefficient for controlling the degree of influence of the activity intensity on the acquisition frequency; is the activity intensity indicator of the worker at time t; is a function of the environmental change; is an adjustment coefficient for controlling the degree of influence of the environmental change on the acquisition frequency; is the environmental change indicator of the worker at time t.

2. The method of assessing the health of a person in a high-risk work environment of claim 1, wherein, before extracting the feature vector corresponding to the real-time physiological health data, the method further comprises the following steps: wavelet transforming and filtering the electrocardiogram data; obtaining a baseline value of skin electricity data of a human body in a calm state, and then baseline calibrating the skin electricity data according to the baseline value; decomposing the electromyography data into a plurality of independent component signals, and denoising each of the independent component signals to obtain denoised electromyography data; respectively performing data normalization on the wavelet-transformed and filtered electrocardiogram data, the baseline-calibrated skin electricity data, and the denoised electromyography data.

3. The method of assessing the health of a person in a high-risk work environment of claim 1, wherein, the health risk assessment model is: U ； wherein U is the probability of the occurrence of a health risk given the synthetic feature vector FVUE; is the intercept term; is the synthetic feature vector is the weight coefficient corresponding to each dimension; is the synthetic feature vector is each element of the synthetic feature vector 4. The method of assessing the health of a person in a high-risk work environment of claim 3, wherein, the training of the health risk assessment model comprises the following steps: obtaining historical physiological health data of workers; extracting a historical comprehensive feature vector corresponding to the historical physiological health data, and training a preset neural network model according to a preset loss function, with the historical comprehensive feature vector as input and the probability of occurrence of a health risk corresponding to the historical physiological health data as output, to obtain the health risk assessment model; wherein the loss function is: wherein, is the number of samples of the training data; is the true health risk label of the th training sample; is the comprehensive feature vector corresponding to the th training sample; is the probability; is the natural logarithm; is the loss function; is a variable of the health risk assessment model prediction result.

5. The method of assessing the health of a person in a high-risk work environment of claim 1, wherein, the comparison of the probability with a preset health risk probability threshold, and then determining the health status of workers according to the comparison result, and generating a corresponding health warning signal when the health status of workers is in a dangerous state, comprises: The probability is compared with a preset health risk probability threshold value, when the probability is equal to the health risk probability threshold value, the health state of the worker is determined as a normal state, when the probability is not equal to the health risk probability threshold value, the health state of the worker is determined as a risk state, and a corresponding health warning signal is generated.

6. A device for implementing the method for assessing the health of a person in a high-risk work environment as claimed in any one of claims 1 to 5, characterized in that, Comprise: Real-time physiological health data acquisition module, health risk assessment module and health warning module; The real-time physiological health data acquisition module is used for collecting real-time physiological health data of workers in a high-risk working environment; wherein the real-time physiological health data includes electrocardiogram data, skin electricity data and electromyography data; The health risk assessment module is used for extracting a feature vector corresponding to the real-time physiological health data, and inputting the feature vector into a preset health risk assessment model, so that the health risk assessment model estimates the probability of occurrence of health risk of workers according to the feature vector, and outputs the probability of occurrence of health risk of workers; wherein the health risk assessment model is obtained by training a preset neural network model with historical physiological health data of workers as input and the probability of occurrence of health risk corresponding to the historical physiological health data as output; The health warning module is used for comparing the probability with a preset health risk probability threshold value, and then determining the health state of the worker according to the comparison result, and generating a corresponding health warning signal when the health state of the worker is a dangerous state.

7. An electronic device, comprising: The storage medium stores a computer program, wherein the computer program controls the device where the storage medium is located to execute the personnel health assessment method in a high-risk working environment as claimed in any one of claims 1 to 5 when running.

8. A storage medium, characterized by The storage medium stores a computer program, wherein the computer program controls the device where the storage medium is located to execute the personnel health assessment method in a high-risk working environment as claimed in any one of claims 1 to 5 when running.

Images