Model establishment and analysis method and system for multi-node associated human skin temperature

Abstract

The invention discloses a multi-node human skin temperature associated model establishment and analysis method and system, and the method comprises the steps: determining human skin temperature measurement points, and carrying out the skin temperature monitoring of each measurement point; decomposing the skin temperature sequence data of each measuring point into a plurality of hierarchies through wavelet transform; connecting the measuring points according to a human body structure and forming a plurality of multi-node graph data with connected edge features, and constructing a graph data set under a plurality of environmental conditions; using the graph data set to train and test the recognition performance of different graph neural network models on environmental conditions, and obtaining an optimal graph neural network model; and by extracting the weights of the edge features and the node features of the trained optimal graph neural network model, the change trend and relevance of the skin temperature of each measuring point of the human body under different environmental conditions are analyzed. According to the method, the relevance between the skin temperatures of all the nodes is introduced, the multi-node dynamic thermal response in different dynamic environments is represented in a refined mode, and deeper thermal response analysis can be achieved through the method.

Description

Technical Field

[0001] This invention relates to the field of human skin temperature thermal response analysis technology, specifically to a method and system for establishing and analyzing a multi-node correlated human skin temperature model. Background Technology

[0002] As a complex thermoregulatory system, the human body's thermal response and dynamic interaction with the environment directly impact work efficiency and health. With the increasing diversity of human activities and the growing trend of frequent extreme weather events globally, analyzing the human body's thermal response has become a key research direction for ensuring life safety and enhancing environmental adaptability. Skin temperature, as a core indicator of the human body's thermal response, plays a crucial role in reflecting the thermoregulatory response to external thermal stimuli, and its dynamic changes directly participate in the body's thermal balance. Non-invasive skin temperature analysis methods have become the mainstream approach for studying human thermal responses in fields such as healthcare, sports science, and architectural design. However, existing skin temperature analysis methods often target static environments, making it difficult to characterize the dynamic thermal response of the human body under different conditions, and they neglect the important role of the correlation between multiple nodes of skin temperature. Therefore, it is necessary to develop a new correlated multi-node human skin temperature analysis method. Summary of the Invention

[0003] This invention provides a model establishment and analysis method and system for multi-node correlated human skin temperature, in order to solve the technical problem that existing skin temperature analysis methods are often designed for static environments, making it difficult to characterize the dynamic thermal response of moving human bodies under different environments, and neglecting the important role of the correlation between multi-node skin temperatures.

[0004] According to the first aspect, one embodiment provides a method for establishing and analyzing a multi-node-related human skin temperature model, the method comprising:

[0005] Step S1: Determine the skin temperature measurement points and monitor the skin temperature at each point;

[0006] Step S2: Decompose the skin temperature sequence data of each measuring point into multiple levels using wavelet transform;

[0007] Step S3: Connect each measuring point according to the human body structure to form a multi-node graph data with multiple edge features connected. The node features include the wavelet energy of each level obtained by wavelet transform of the corresponding measuring point, and construct graph datasets under multiple environmental conditions.

[0008] Step S4: Use the graph dataset to train and test the recognition performance of different graph neural network models for environmental conditions, and obtain the best graph neural network model.

[0009] Step S5: By extracting the weights of the edge features and node features of the best trained graph neural network model, analyze the changing trends and correlations of skin temperature at various measurement points on the human body under different environmental conditions.

[0010] Further, step S1 specifically includes:

[0011] S11, the determined human skin temperature measurement points include the forehead, left back, left chest, left forearm, left wrist, back of left calf, right upper arm, back of right hand, front of right thigh, and front of right calf.

[0012] The S12 uses a wireless button-type data logger to continuously monitor the skin temperature at the measurement points, and the logger is fixed to each measurement point through a medical patch.

[0013] Further, step S2 specifically includes:

[0014] S21, using maximum overlap discrete wavelet transform, is decomposed into multiple levels;

[0015] S22, for each decomposition level, calculate the wavelet energy value and use it as a feature to characterize the corresponding level.

[0016] Further, step S4 specifically includes:

[0017] Step S41: Evaluate the environmental condition recognition performance of different graph neural network models based on multiple evaluation metrics to obtain the best model. The evaluation metrics include accuracy, precision, recall, and F1 score.

[0018] Further, step S4 specifically includes:

[0019] In step S42, after selecting the best model, Gaussian noise is introduced to enhance robustness, and a two-stage learning strategy is adopted to improve model performance.

[0020] Further, step S41 specifically includes:

[0021] The optimal graph neural network model was determined to be the graph attention network model. The graph attention network model includes three GAT layers, three activation function layers, three dropout layers, two batch normalization (BN) layers, an average pooling layer, and a fully connected layer. The three GAT layers are connected to the three activation function layers, the three activation function layers are connected to the two dropout layers and the average pooling layer, the two dropout layers and the average pooling layer are connected to the two BN layers and the dropout layer, respectively, and the recognition result is finally output through the fully connected layer.

[0022] Further, step S5 specifically includes:

[0023] S51, node feature weights reflect the changing trend of skin temperature at each measuring point, while edge feature weights reflect the dynamic correlation of skin temperature information between measuring points. By using GNNExplainer to generate feature masks, key node features and edge features are identified, and different feature weights are redistributed according to their contribution to the recognition performance of environmental conditions.

[0024] Furthermore, step S5 specifically includes:

[0025] S52, in the human body image formed by connecting the measurement points according to the human body structure, the edges with relatively high edge feature weights are marked as black solid lines, and the remaining edges are marked as gray dashed lines.

[0026] S53, through comparison and analysis of human images under different environmental conditions, analyzes the changing trends and correlations of skin temperature at various measuring points under different environments, and further reveals the relationship between the environment and the human body's thermoregulation mechanism.

[0027] According to the second aspect, one embodiment provides a model establishment and analysis system for multi-node correlation of human skin temperature, the system comprising:

[0028] The temperature monitoring module is used to determine the skin temperature measurement points on the human body and to monitor the skin temperature at each measurement point.

[0029] The wavelet transform module is used to decompose the skin temperature sequence data of each measurement point into multiple levels through wavelet transform.

[0030] The graph dataset construction module is used to connect the measurement points according to the human body structure and form a multi-node graph data with multiple edge features. The node features include the wavelet energy of each level obtained by wavelet transform of the corresponding measurement points, and to construct graph datasets under multiple environmental conditions.

[0031] The optimal model acquisition module is used to train and test the recognition performance of different graph neural network models on environmental conditions using graph datasets, and obtain the optimal graph neural network model.

[0032] The correlation analysis module is used to analyze the changing trends and correlations of skin temperature at various measurement points on the human body under different environmental conditions by extracting the weights of the edge features and node features of the best trained graph neural network model.

[0033] According to a third aspect, one embodiment provides an electronic device, the device comprising: a processor and a memory;

[0034] The memory is used to store one or more program instructions;

[0035] The processor is configured to run one or more program instructions to perform the steps of a multi-node correlated human skin temperature model establishment and analysis method as described in any of the preceding claims.

[0036] According to a fourth aspect, one embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of a multi-node correlated human skin temperature model establishment and analysis method as described in any of the preceding claims.

[0037] This invention provides a method and system for establishing and analyzing a multi-node correlated human skin temperature model, which has the following beneficial effects:

[0038] (1) This invention is applicable not only to thermal response analysis in a static state, but also to thermal response analysis of human skin temperature of moving individuals in complex scenarios, and has better versatility.

[0039] (2) By introducing the correlation between skin temperatures at each node, the dynamic thermal response of multiple nodes under different dynamic environments can be characterized in a more refined manner. This method can achieve a deeper level of thermal response analysis.

[0040] (3) This discovery opens up new avenues for the development of personalized temperature control devices in complex scenarios, the shape design of phase change temperature control materials, and the study of human-environment interaction. Attached Figure Description

[0041] Figure 1 A flowchart illustrating a method for establishing and analyzing a multi-node correlated human skin temperature model, as provided in one embodiment of the present invention;

[0042] Figure 2 This is a schematic diagram illustrating the specific implementation process of a multi-node correlated human skin temperature model establishment and analysis method provided in one embodiment of the present invention;

[0043] Figure 3 This is a schematic diagram showing the location of measurement points for the skin temperature of a moving human body in a multi-node correlated human skin temperature model establishment and analysis method provided in an embodiment of the present invention.

[0044] Figure 4 This is a schematic diagram of a sample in a plateau tunnel environment case where the skin temperature on the left back of a subject is decomposed into 5 layers using a multi-node correlation model establishment and analysis method for human skin temperature in an embodiment of the present invention.

[0045] Figure 5 This is a schematic diagram of the graph data structure of human skin temperature characteristics at multiple nodes in a plateau tunnel environment case, provided in a model establishment and analysis method for multi-node correlation of human skin temperature according to an embodiment of the present invention.

[0046] Figure 6 This is a schematic diagram of the GAT model structure used in a plateau tunnel environment case study of a multi-node correlation human skin temperature model establishment and analysis method provided in an embodiment of the present invention.

[0047] Figure 7 Loss plots of different graph neural network models in a plateau tunnel environment case study of a multi-node correlation model establishment and analysis method for human skin temperature provided in an embodiment of the present invention;

[0048] Figure 8 This is a schematic diagram illustrating the multi-node correlation in different environments in a plateau tunnel environment case study, which is part of a method for establishing and analyzing a multi-node correlation model of human skin temperature provided in an embodiment of the present invention.

[0049] Figure 9 This is a schematic diagram illustrating the difference in multi-node correlation under normal and low pressure in a plateau tunnel environment case study of a multi-node correlation model establishment and analysis method for human skin temperature provided in an embodiment of the present invention.

[0050] Figure 10 The variation trend of average human body temperature under various working conditions in a plateau tunnel environment case study provided by an embodiment of the present invention is illustrated in the model establishment and analysis method for multi-node correlation of human skin temperature.

[0051] Figure 11 This is a schematic diagram of the logical structure of a multi-node correlated human skin temperature model establishment and analysis system provided in one embodiment of the present invention. Detailed Implementation

[0052] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings. Similar elements in different embodiments are referred to by associated similar element reference numerals. In the following embodiments, many details are described to facilitate a better understanding of the invention. However, those skilled in the art will readily recognize that some features may be omitted in different situations, or may be replaced by other elements, materials, or methods. In some cases, certain operations related to the present invention are not shown or described in the specification. This is to avoid obscuring the core parts of the invention with excessive description. For those skilled in the art, detailed description of these related operations is not necessary; they can fully understand the related operations based on the description in the specification and general technical knowledge in the art.

[0053] Furthermore, the features, operations, or characteristics described in the specification can be combined in any suitable manner to form various embodiments. At the same time, the steps or actions in the method description can be rearranged or adjusted in a manner obvious to those skilled in the art. Therefore, the various orders in the specification and drawings are only for the clear description of a particular embodiment and do not imply a necessary order, unless otherwise stated that a particular order must be followed.

[0054] The first embodiment of this invention provides a method for establishing and analyzing a multi-node correlated human skin temperature model. The following is in conjunction with... Figure 1 and Figure 2 Please provide a detailed explanation.

[0055] like Figure 1 As shown, in step S1, the human skin temperature measurement points are determined, and the skin temperature at each measurement point is monitored.

[0056] The above steps specifically include:

[0057] S11, Test Scenario Selection: Select the dynamic scenario you want to study.

[0058] S12, Tester selection: Select the most widely distributed group of people in this dynamic scenario to match the analysis needs of this scenario.

[0059] S13, Test Exercise Selection: Select common or equivalent exercises in this scenario and arrange for the test group to perform such exercises.

[0060] S14, Measurement Point Selection: Select skin temperature measurement points for the human body based on the scenario and movement requirements. Determine whether the number and location of measurement points will significantly affect the subject's comfort. If so, reselect measurement points. If not, proceed to the next step, model input.

[0061] Specifically, this embodiment takes a plateau tunnel environment as an example, and the selected environmental parameters include: (1) Air temperature (Ta): a gradient sequence of 31℃, 34℃ and 37℃ is set. When the air temperature exceeds 37℃, the risk of heat stress in the human body will increase significantly. To ensure the safety of the subjects, the highest experimental temperature in this experiment is set at 37℃; (2) Atmospheric pressure (PB): two altitude levels are simulated - 96.5kPa and 65.7kPa (altitude 3500 meters, representing a typical plateau tunnel environment); (3) Relative humidity (RH): based on the measured humidity range of the plateau tunnel environment, it is maintained at a constant 70%. The environmental parameter control is shown in Table 1.

[0062] Table 1 Environmental Parameters

[0063]

[0064] Prior power analysis was performed using G*Power software to determine the minimum sample size as 10 healthy male participants, whose gender composition matched the demographic characteristics of high-altitude tunnel workers. The study actually recruited 14 male graduate students. Inclusion criteria included: no history of long-term high-altitude (>1500 meters) residence in the past year, at least 60 days since the last high-altitude exposure, and no thermoregulatory or cardiovascular diseases. Participant information is shown in Table 2.

[0065] Table 2 Subject Information

[0066]

[0067] a Dubois surface area,A=0.202 weight 0.425 height 0.725

[0068] b BMI = weight / height 2

[0069] In tunnel construction projects, tasks of medium labor intensity constitute the largest proportion of the total workload. Based on ISO 8996 standard, a 165W / m... 2 The metabolic equivalent was used as a reference value for moderate labor intensity. A controlled exercise-rest alternation pattern was employed using a 4km / h treadmill interval training program (20 minutes of walking / 5 minutes of rest cycle) to simulate the work rhythm of tunnel construction. The experiment consisted of 5 consecutive interval training sessions, with a total duration of 125 minutes.

[0070] A wireless button-type data logger was used to continuously monitor skin temperature, and the device was fixed to each measurement point using medical-grade adhesive patches. All patch surfaces were covered with aluminum foil tape to reduce interference from environmental radiant heat. To deeply analyze the dynamic response of skin temperature at multiple points, and to ensure effective heat dissipation by controlling the aluminum foil coverage, a symmetrical ten-point measurement scheme was designed, such as... Figure 3 As shown. Human skin temperature measuring points include the forehead, left back, left chest, left forearm, left wrist, back of left calf, right upper arm, back of right hand, front of right thigh, and front of right calf.

[0071] like Figure 1 As shown, in step S2, the skin temperature sequence data of each measuring point is decomposed into multiple levels through wavelet transform.

[0072] The above steps specifically include:

[0073] S21, Wavelet Transform: The skin temperature at each measurement point is decomposed into multiple levels using wavelet transform. Level 0 (coarse scale) represents the low-frequency baseline trend and the slow response of skin temperature, while subsequent levels progressively capture higher-frequency transient fluctuations and detailed changes. For each decomposition level, the wavelet energy value is calculated to characterize the features of that level. In each environmental condition, there are multiple subjects, and each subject has multiple measurement points. For each measurement point, there is only one skin temperature sequence, which is decomposed into multiple levels using wavelet transform.

[0074] Specifically, in this embodiment, the maximum overlap discrete wavelet transform is used to decompose the data into 5 levels. The calculation formula for the maximum overlap discrete wavelet transform is shown below:

[0075]

[0076] Wavelet energy:

[0077]

[0078] Where N represents the signal length, J represents the maximum number of decomposition levels, j represents the current level, l represents the l-th level of the wavelet decomposition, and h j Represents a high-pass filter. G represents the high-pass filter of the l-th layer. j Represents a low-pass filter. This represents the low-pass filter of the l-th layer, where L represents the filter length, t represents the time index, and W... j V represents the wavelet coefficients. j E represents the scaling factor. j This represents wavelet energy, and modN represents circular convolution to handle boundary effects. The decomposition example is shown below. Figure 4 As shown.

[0079] like Figure 1 As shown, in step S3, the measurement points are connected according to the human body structure to form a multi-node graph data with multiple edge features connected. The node features include the wavelet energy of each level obtained by wavelet transform of the corresponding measurement points, and a graph dataset under multiple environmental conditions is constructed.

[0080] The above steps specifically include:

[0081] S31, Graph Data Establishment: Based on the selected measurement points, these points are connected according to human anatomy to form a multi-node graph data structure with multiple interconnected edge features. Each node is input with multi-layer wavelet energy obtained from wavelet transform, and the sampling interval is input as an auxiliary parameter to jointly form the node features. Multiple graph data sets are generated for each subject across multiple scenarios. Finally, all graph data are aggregated to form the dataset for the dynamic scenarios of this study.

[0082] Specifically, the impact of sampling intervals of 0.5, 1, 1.5, 2, 2.5, 3, 3.5, 4, and 4.5 min on wavelet energy was tested, as shown in Table 3. This table specifically characterizes whether there is a significant difference in wavelet energy at different sampling frequencies under this environment. If there is, it is written as Pn; otherwise, it is represented by a slash. The sampling interval and the five wavelet energy levels mentioned in S2 are used as nodes for input. Based on the human anatomy, the measurement points in S1 are connected, and data from subjects in six high-altitude tunnel environments are collected, ultimately yielding 702 valid graph data points, forming the final graph dataset. The graph data structure is as follows... Figure 5 As shown.

[0083] Table 3 shows the influence of the test sampling interval on wavelet energy.

[0084]

[0085]

[0086] P1 represents the forehead; P2 represents the front of the right thigh; P3 represents the back of the right hand; P7 represents the left forearm; P8 represents the left wrist; P9 represents the back of the left calf; PMST represents the average temperature wavelet energy of the whole body.

[0087] like Figure 1 As shown, in step S4, the graph dataset is used to train and test the recognition performance of different graph neural network models for environmental conditions, and the best graph neural network model is obtained.

[0088] The above steps specifically include:

[0089] S41. Select a suitable graph neural network based on the graph dataset and test the recognition performance of the graph neural network in different scenarios.

[0090] Specifically, as an advanced deep learning method, graph neural networks (GNNs) exhibit excellent performance in processing graph-structured data through node neighborhood aggregation operations, and can effectively learn the topological features of the data.

[0091] To obtain the optimal model, this embodiment evaluates the environmental condition recognition performance of different graph neural network models based on multiple evaluation metrics. The evaluation metrics include accuracy, precision, recall, and F1 score. The formulas for each evaluation metric are shown below:

[0092]

[0093] Where TP represents the number of correctly predicted positive samples, FP represents the number of incorrectly predicted positive samples, TN represents the number of correctly predicted negative samples, and FN represents the number of incorrectly predicted negative samples.

[0094] The recognition performance of different graph neural network models is shown in Table 4. The loss of different models is as follows: Figure 7 As shown.

[0095] Table 4. Recognition performance of different graph neural network models

[0096]

[0097] After comparing and analyzing the recognition performance of different models, the best model is modeled using a Graph Attention Network (GAT) architecture, and the neighborhood aggregation operation is defined by the following formula:

[0098] e ij =LeakyReLU(a T [Wh i ||Wh j ])

[0099]

[0100] The characteristics of node i in layer l; is the feature vector of node j in the previous layer; W is the learnable weight matrix; a is the trainable parameter vector of the attention mechanism; || is the concatenation operation; σ is the Sigmoid function.

[0101] The GAT model structure is as follows: Figure 6 As shown, the model mainly consists of three GAT layers, three activation function layers, three Dropout layers, two BN layers, an average pooling layer, and a fully connected layer. The three GAT layers are connected to the three activation function layers, which in turn are connected to the two Dropout layers and one average pooling layer. The two Dropout layers and one average pooling layer are then connected to the two BN layers and one Dropout layer, respectively, and the recognition result is finally output through the fully connected layer. The GAT layers capture the correlation features between different nodes through an attention mechanism. The activation function layers introduce non-linear mappings, allowing the neural network to learn complex relationships. The Dropout layers randomly discard some neuron outputs, reducing inter-neuron dependencies and mitigating the risk of overfitting. The BN layers standardize the input, stabilize the training process, accelerate convergence, and improve the model's generalization ability.

[0102] S42, after selecting the best model, introduces Gaussian noise to enhance its robustness, and adopts a two-stage learning strategy to improve model performance.

[0103] The formula for the two-stage learning strategy is shown below:

[0104]

[0105] Where Lr is the current learning rate, Lr min and Lrmax For the minimum and maximum learning rates, T cur T represents the current training round. max This indicates the maximum number of training rounds.

[0106] Based on the above results, the model can accurately determine the subject's working condition based on the skin temperature at each node. Conversely, the model learns the commonalities in the skin temperature at each node among different subjects under a certain working condition.

[0107] like Figure 1 As shown, in step S5, by extracting the weights of the edge features and node features of the best trained graph neural network model, the changing trends and correlations of skin temperature at various measuring points on the human body under different environmental conditions are analyzed.

[0108] The above steps specifically include:

[0109] S42, Correlation Analysis: The model characterizes the correlation of skin temperature features at each node through feature weights. By extracting the weights of edge features and node features, the model analyzes the changing trends and correlation characteristics of skin temperature at each node of the human body under different environments, revealing a deeper understanding of the relationship between dynamic scenes and the human body's thermoregulation mechanism.

[0110] Specifically, during training, the model's edge and node features are assigned different feature weights, which can be directly output by the model. The differences in node and edge feature weights under different environmental conditions stem from the differences in skin temperature response exhibited by the human body in different scenarios.

[0111] GNNExplainer is used to generate feature masks, obscuring some node and edge features. Different feature weights are then redistributed based on their contribution to the environmental condition recognition performance, as shown in the following formula:

[0112] P edge [i,j]=σ(M edge [i,j])

[0113] P feat [i] = σ(M) feat [i])

[0114] Among them, P edge [i,j] represents the importance probability of the edge between nodes i and j, M edge [i,j] represents the edge feature weights output by the model mask, σ is the Sigmoid function, and P feat [i] represents the importance probability of the edge between nodes i, M featThis represents the node feature weight values ​​output by the model mask. A larger weight value indicates a more important node or edge feature. Here, node feature weights reflect the changing trends of skin temperature at each node, while edge features reflect the dynamic correlation of skin temperature information between measurement points. By analyzing the graph data of correctly identified subjects, the average values ​​of each node and edge feature for each subject under each working condition can be obtained.

[0115] To achieve better visualization, this embodiment presents edge features as human body images, marking the top 5 edge feature weights with solid black lines and the remaining edges with dashed gray lines. The largest node features are labeled with text. Feature analysis revealed that at 31℃ and 34℃ under normal pressure, nodes at greater distances exhibit more significant correlations, while at low pressure, nodes at closer distances exhibit more significant correlations. The node correlation characteristics at 37℃ under normal pressure and low pressure are similar, such as... Figure 8 As shown.

[0116] Furthermore, at the same temperature, the corresponding edge features are subtracted to obtain the differences in edge features between normal and low pressure conditions. The invention marks the top 5 edges with the largest differences in edge feature weights as solid black lines, and the remaining edges as dashed gray lines. At 31°C, the differences between normal and low pressure mainly focus on the node correlations related to the chest; at 34°C, the differences mainly focus on the node correlations related to the back; and at 37°C, the differences mainly focus on the node correlations related to the chest and back. Figure 9 As shown.

[0117] Depend on Figure 8 and Figure 9 It is evident that the human body exhibits significant differences at 37°C compared to 31°C and 34°C. This embodiment further discusses the relationship between this phenomenon and the human body's thermoregulation mechanism.

[0118] The heat transfer process in the human body mainly occurs through convection, radiation, evaporation, and conduction. The formula for low pressure is shown below:

[0119] Convective heat transfer:

[0120]

[0121] Where C represents convective heat transfer, f cl h represents the thermal resistance coefficient of clothing. c,0 P represents the convective heat transfer coefficient at normal pressure. B P0 represents normal pressure, and P0 represents low pressure. sk -t a R represents the temperature difference between the body surface and the environment. cl This represents the thermal resistance of the clothing. When n = 2m, it is natural convection; when n = m, it is forced convection.

[0122] Radiative heat transfer:

[0123] R = α × f d ×[(t p +273) 4 -(t e +273) 4 ]

[0124] Where R represents the radiated power, α represents the Stefan-Boltzmann constant, and f d Representing a geometric factor, describing the geometric relationship between the two, t p t represents human body temperature e Represents ambient temperature

[0125] Evaporative cooling:

[0126]

[0127] Among them, E sk,p representing evaporative heat loss, w represents the skin moisture coefficient, and f represents the skin's humidity coefficient. cl h represents the thermal resistance coefficient of clothing. c,0 R represents the convective heat transfer coefficient. d P represents the thermal resistance of clothing. sk P represents the partial pressure of water vapor on the skin surface. a i represents the partial pressure of water vapor in ambient air cl This represents the moisture permeability coefficient of clothing.

[0128] Conductive heat dissipation:

[0129]

[0130] Where k represents the thermal conductivity and A represents the contact area. This represents the temperature gradient.

[0131] In low-pressure environments, the human body's radiative and conductive heat dissipation is largely unaffected, depending mainly on the ambient temperature. For convective heat transfer, the reduced air density in low-pressure environments weakens the fluid's convective disturbance ability, ultimately leading to a decrease in the convective heat transfer coefficient and thus reducing the body's convective heat dissipation capacity. However, in terms of evaporative heat dissipation, the decrease in atmospheric pressure lowers the partial pressure of water vapor, promoting water evaporation and thereby enhancing the body's evaporative heat dissipation capacity.

[0132] like Figure 10As shown in the figure, this embodiment observed that at 31°C, the mean whole-body temperature (MST) first increased and then gradually decreased throughout the experimental period; while at 34°C, the MST tended to stabilize after an initial increase. At both 31°C and 34°C, the MST in the low-pressure environment was higher than that in the normal-pressure environment. Conversely, at 37°C, the MST showed a continuous upward trend, and the MST value under normal-pressure conditions was higher than that under low-pressure conditions. At 31°C and 34°C, the initial MST was higher than the ambient temperature, resulting in heat dissipation through convection, radiation, and evaporation; however, at 37°C, the MST hardly exceeded the ambient temperature, causing the body to continuously absorb heat through convection and radiation, relying solely on evaporation for heat dissipation, thus disrupting the body's thermoregulatory balance and leading to heat accumulation.

[0133] In high-temperature, high-humidity, and low-pressure environments, heat dissipation between the human body and the environment is significantly hindered. The body's internal systems need to intervene through thermoregulation mechanisms to balance heat distribution, prevent localized heat accumulation, and avoid a series of heat-related risks. The heat transfer process within the human body is primarily described by the Pennes equation, which mainly includes terms for heat conduction, blood convection, and metabolic heat production, as shown in the following formula:

[0134]

[0135] Where T represents the temperature in the biological tissue, t represents time, and k represents the thermal conductivity of the biological tissue. The Laplace operator representing temperature, ρ and ρ b Representing the density of tissues and blood, c and c b ω represents the specific heat capacity of tissues and blood. b T represents blood perfusion rate. b It represents the temperature of the blood. Represents heat capacity. Represents the heat conduction term, ρ b c b ω b (T b -T) represents the blood convection term, Q m It represents metabolic heat.

[0136] If heat transfer within the body relies solely on conduction, a significant temperature gradient is required to meet the heat transfer demands. Under the high-temperature conditions of this experiment, the temperature difference between body and ambient temperatures was relatively small, resulting in low heat transfer efficiency. In low-pressure environments, the metabolic rate increases, leading to increased metabolic heat production; simultaneously, blood viscosity increases, and blood perfusion rate decreases, resulting in reduced skin blood flow. Furthermore, in the absence of blood perfusion, the thermal conductivity of human tissues is extremely low. To prevent localized heat accumulation, the human body primarily relies on blood convection to distribute heat throughout the body, thereby maintaining body temperature balance.

[0137] like Figure 8As shown, under normal pressure, the human body's blood circulation efficiency is relatively high, which makes heat transfer more rapid and temperature changes in distant areas can be better synchronized. The body can more quickly regulate and balance skin temperature, resulting in relatively less heat accumulation. Therefore, the overall trend of skin temperature change is relatively gentle, dominated by slow wavelet energy characteristics. Under low pressure, due to the obstruction of blood heat transfer and the limitation of long-distance heat transfer, temperature changes in distant areas are asynchronous, and skin temperature regulation lags. Meanwhile, in nearby areas, due to relatively slowed blood circulation, the temperature dynamic correlation is stronger, and more heat accumulates, leading to a more rapid overall trend of skin temperature change, dominated by fast wavelet energy characteristics. However, when the ambient temperature reaches 37°C, the human body relies almost entirely on evaporative cooling, and heat dissipation is extremely difficult even under normal pressure. Therefore, the edge characteristics under normal and low pressure conditions show similarities.

[0138] like Figure 9 As shown, with increasing ambient temperature, the area of ​​greatest difference between normal and low-pressure conditions transitions from chest-related peripheral features to the back region, and finally to chest-back related peripheral features, with the main differences concentrated in the core region. This is because heat transfer in the core region relies more on blood convection, while the number and density of sweat glands on the back are greater than on the chest. As temperature rises, the body's need for heat dissipation becomes more urgent, and the back plays a more important role in evaporative cooling. At 37°C, both the chest and back play important roles in heat dissipation. Skin temperature exhibits diverse dynamic characteristics under different environmental conditions. Despite individual differences, the common fluctuations in skin temperature dynamics still reflect the response of the body's thermoregulatory mechanisms to different environmental conditions.

[0139] Corresponding to the aforementioned method for establishing and analyzing a multi-node correlated human skin temperature model, this invention also discloses a system for establishing and analyzing a multi-node correlated human skin temperature model, such as... Figure 11 As shown, it specifically includes:

[0140] The temperature monitoring module is used to determine the skin temperature measurement points on the human body and to monitor the skin temperature at each measurement point.

[0141] The wavelet transform module is used to decompose the skin temperature sequence data of each measurement point into multiple levels through wavelet transform.

[0142] The graph dataset construction module is used to connect the measurement points according to the human body structure and form a multi-node graph data with multiple edge features. The node features include the wavelet energy of each level obtained by wavelet transform of the corresponding measurement points, and to construct graph datasets under multiple environmental conditions.

[0143] The optimal model acquisition module is used to train and test the recognition performance of different graph neural network models on environmental conditions using graph datasets, and obtain the optimal graph neural network model.

[0144] The correlation analysis module is used to analyze the changing trends and correlations of skin temperature at various measurement points on the human body under different environmental conditions by extracting the weights of the edge features and node features of the best trained graph neural network model.

[0145] It should be noted that for a detailed description of the multi-node correlated human skin temperature model establishment and analysis system provided in the embodiments of the present invention, please refer to the relevant description of the multi-node correlated human skin temperature model establishment and analysis method provided in the embodiments of the present invention, which will not be repeated here.

[0146] In addition, embodiments of the present invention also provide an electronic device, the device comprising: a processor and a memory; the memory being used to store one or more program instructions; the processor being used to execute one or more program instructions to perform the steps of a multi-node correlated human skin temperature model establishment and analysis method as described in any of the preceding embodiments.

[0147] It should be noted that for a detailed description of an electronic device provided in the embodiments of the present invention, please refer to the relevant description of a multi-node associated human skin temperature model establishment and analysis method provided in the embodiments of the present invention, which will not be repeated here.

[0148] In addition, embodiments of the present invention also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the multi-node associated human skin temperature model establishment and analysis method as described in any of the preceding claims.

[0149] It should be noted that for a detailed description of a computer-readable storage medium provided in the embodiments of the present invention, please refer to the relevant description of a multi-node associated human skin temperature model establishment and analysis method provided in the embodiments of the present invention, which will not be repeated here.

[0150] Those skilled in the art will understand that all or part of the functions of the various methods in the above embodiments can be implemented by hardware or by computer programs. When all or part of the functions in the above embodiments are implemented by computer programs, the program can be stored in a computer-readable storage medium, which may include: read-only memory, random access memory, disk, optical disk, hard disk, etc., and the program is executed by a computer to achieve the above functions. For example, the program can be stored in the memory of a device, and when the program in the memory is executed by the processor, all or part of the above functions can be achieved. In addition, when all or part of the functions in the above embodiments are implemented by computer programs, the program can also be stored in a server, another computer, disk, optical disk, flash drive, or external hard drive, etc., and can be downloaded or copied to the memory of a local device, or the system of the local device can be updated. When the program in the memory is executed by the processor, all or part of the functions in the above embodiments can be achieved.

[0151] The above examples illustrate the present invention only to aid in understanding it and are not intended to limit the scope of the invention. Those skilled in the art can make various simple deductions, modifications, or substitutions based on the principles of this invention.

Claims

1. A method for establishing and analyzing a multi-node correlation model of human skin temperature, characterized in that, The method includes: Step S1: Determine the skin temperature measurement points and monitor the skin temperature at each point; Step S2: Decompose the skin temperature sequence data of each measurement point into multiple levels using wavelet transform; Step S3: Connect each measuring point according to the human body structure to form a multi-node graph data with multiple edge features connected. The node features include the wavelet energy of each level obtained by wavelet transform of the corresponding measuring point, and construct graph datasets under multiple environmental conditions. Step S4: Use the graph dataset to train and test the recognition performance of different graph neural network models for environmental conditions, and obtain the best graph neural network model. Step S5: By extracting the weights of the edge features and node features of the best trained graph neural network model, analyze the changing trends and correlations of skin temperature at various measurement points on the human body under different environmental conditions.

2. The method for establishing and analyzing a multi-node correlated human skin temperature model as described in claim 1, characterized in that, Step S1 specifically includes: S11, the determined human skin temperature measurement points include the forehead, left back, left chest, left forearm, left wrist, back of left calf, right upper arm, back of right hand, front of right thigh, and front of right calf. The S12 uses a wireless button-type data logger to continuously monitor the skin temperature at the measurement points, and the logger is fixed to each measurement point through a medical patch.

3. The method for establishing and analyzing a multi-node correlated human skin temperature model as described in claim 1, characterized in that, Step S2 specifically includes: S21, using maximum overlap discrete wavelet transform, is decomposed into multiple levels; S22, for each decomposition level, calculate the wavelet energy value and use it as a feature to characterize the corresponding level.

4. The method for establishing and analyzing a multi-node correlated human skin temperature model as described in claim 1, characterized in that, Step S4 specifically includes: Step S41: Evaluate the environmental condition recognition performance of different graph neural network models based on multiple evaluation metrics to obtain the best model. The evaluation metrics include accuracy, precision, recall, and F1 score.

5. The method for establishing and analyzing a multi-node correlated human skin temperature model as described in claim 4, characterized in that, Step S4 specifically includes: In step S42, after selecting the best model, Gaussian noise is introduced to enhance robustness, and a two-stage learning strategy is adopted to improve model performance.

6. The method for establishing and analyzing a multi-node correlated human skin temperature model as described in claim 4, characterized in that, Step S41 specifically includes: The optimal graph neural network model was determined to be the graph attention network model. The graph attention network model includes three GAT layers, three activation function layers, three dropout layers, two batch normalization (BN) layers, an average pooling layer, and a fully connected layer. The three GAT layers are connected to the three activation function layers, the three activation function layers are connected to the two dropout layers and the average pooling layer, the two dropout layers and the average pooling layer are connected to the two BN layers and the dropout layer, respectively, and the recognition result is finally output through the fully connected layer.

7. The method for establishing and analyzing a multi-node correlated human skin temperature model as described in claim 1, characterized in that, Step S5 specifically includes: S51, node feature weights reflect the changing trend of skin temperature at each measuring point, while edge feature weights reflect the dynamic correlation of skin temperature information between measuring points. By using GNNExplainer to generate feature masks, key node features and edge features are identified, and different feature weights are redistributed according to their contribution to the recognition performance of environmental conditions.

8. The method for establishing and analyzing a multi-node correlated human skin temperature model as described in claim 7, characterized in that, Step S5 further includes: S52, in the human body image formed by connecting the measurement points according to the human body structure, the edges with relatively high edge feature weights are marked as black solid lines, and the remaining edges are marked as gray dashed lines. S53, through comparison and analysis of human images under different environmental conditions, analyzes the changing trends and correlations of skin temperature at various measuring points under different environments, and further reveals the relationship between the environment and the human body's thermoregulation mechanism.

9. A system for establishing and analyzing a multi-node correlation model of human skin temperature, characterized in that, The system includes: The temperature monitoring module is used to determine the skin temperature measurement points on the human body and to monitor the skin temperature at each measurement point. The wavelet transform module is used to decompose the skin temperature sequence data of each measurement point into multiple levels through wavelet transform. The graph dataset construction module is used to connect the measurement points according to the human body structure and form a multi-node graph data with multiple edge features. The node features include the wavelet energy of each level obtained by wavelet transform of the corresponding measurement points, and to construct graph datasets under multiple environmental conditions. The optimal model acquisition module is used to train and test the recognition performance of different graph neural network models on environmental conditions using graph datasets, and obtain the optimal graph neural network model. The correlation analysis module is used to analyze the changing trends and correlations of skin temperature at various measurement points on the human body under different environmental conditions by extracting the weights of the edge features and node features of the best trained graph neural network model.

10. An electronic device, characterized in that, The device includes: a processor and a memory; The memory is used to store one or more program instructions; The processor is configured to run one or more program instructions to perform the steps of the multi-node correlated human skin temperature model establishment and analysis method as described in any one of claims 1 to 8.

Images