Outdoor thermal comfort intelligent evaluation method and system taking psychological state as medium

Abstract

The invention discloses an outdoor thermal comfort intelligent evaluation method and system taking a psychological state as a medium. The method specifically comprises the following steps: acquiring psychological index data and multi-scale parameters, wherein the multi-scale parameters comprise environmental physical parameters and physiological response parameters; inputting the environmental physical parameters and the physiological response parameters into a dynamic feature extraction module, and performing cross-modal feature fusion through environmental physical parameter coding and physiological time sequence feature extraction to generate a fusion feature vector; constructing a psychological intermediary prediction module according to the psychological index data, inputting the fusion feature vector into the psychological intermediary prediction module, and generating a dynamic adjustment coefficient; and inputting the fusion feature vector and the dynamic adjustment coefficient into a thermal comfort evaluation module, and outputting a multi-scale prediction result. According to the method, the problem of space-time alignment of environmental parameter transient and physiological response hysteresis is solved, and the individual adaptability and prediction reliability of outdoor thermal comfort evaluation are remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of thermal environment assessment and artificial intelligence, and specifically relates to an intelligent assessment method and system for outdoor thermal comfort using psychological state as a medium. Background Art

[0002] Under the dual pressures of global climate change and urbanization, there is an urgent need to create high-quality, comfortable, and healthy outdoor thermal environments. Scientifically, accurately, and comprehensively assessing human thermal comfort in outdoor environments has become a highly challenging and cutting-edge research topic in the field of thermal environmental perception. While current human thermal comfort assessment technologies are continuously developing and improving, two key bottlenecks remain, severely limiting the accuracy and reliability of assessment results.

[0003] First, traditional thermal comfort assessment models often focus on using a single physical indicator or analyzing physiological signals. The widely used Predicted Mean Value (PMV) model primarily predicts the thermal sensation of the human body under steady-state conditions based on physical environmental parameters such as air temperature, humidity, wind speed, and mean radiant temperature. By comprehensively considering both physical and physiological indicators, a number of classic steady-state heat transfer mechanism models have emerged, such as PET and SET. However, the human body's perception of the outdoor thermal environment is a complex and dynamic process, and ignoring the mediating role of psychological state can lead to significant deviations in the assessment or prediction of actual human thermal comfort.

[0004] Secondly, while existing multimodal fusion methods for thermal comfort assessment aim to integrate multiple data sources to improve the comprehensiveness and accuracy of assessments, they face numerous practical challenges. Currently, many multimodal fusion methods often rely on static weight assignment strategies, pre-setting fixed weights for different data sources (such as physical environmental parameters, physiological signals, and psychological questionnaire data) and then performing a simple weighted summation. However, this static weight assignment approach ignores the varying importance of different data sources in different environments and individual states. In reality, a complex dynamic coupling exists between environmental stimuli, physiological responses, and psychological feedback. Static weight assignment fails to capture this dynamic coupling, resulting in fused assessment results that fail to accurately reflect the human body's true thermal comfort experience in complex outdoor environments. Furthermore, some multimodal fusion methods employ only shallow feature splicing, simply combining features from different data sources without deeply exploring the inherent connections and underlying patterns between them, which also limits improvements in assessment performance. Summary of the Invention

[0005] The purpose of the present invention is to provide an intelligent outdoor thermal comfort assessment method and system based on psychological state, which solves the problem of spatiotemporal alignment of the transient nature of environmental parameters and the lag of physiological responses, significantly improves the individual adaptability and prediction reliability of the thermal comfort assessment of humans in complex outdoor thermal environments, and solves at least one of the above-mentioned problems of the prior art.

[0006] In a first aspect, the present invention provides an intelligent outdoor thermal comfort assessment method using psychological state as a medium, the method specifically comprising: Acquiring psychological index data and multi-scale parameters, wherein the psychological index data includes multiple emotional factor data that can be converted into overall emotional disorder degree data, and the multi-scale parameters include environmental physical parameters and physiological response parameters; The environmental physical parameters and physiological response parameters are input into the dynamic feature extraction module, and the fused feature vector is generated by encoding the environmental physical parameters and extracting the physiological time series features, and performing cross-modal feature fusion. Construct a psychological mediation prediction module based on psychological indicator data, input the fused feature vector into the psychological mediation prediction module, and generate a dynamic adjustment coefficient by integrating the emotional state prediction; The fused feature vector and dynamic adjustment coefficient are input into the thermal comfort assessment module, and the multi-scale prediction results are output by integrating the spatiotemporal feature interaction pattern and loss function optimization.

[0007] In a second aspect, the present invention provides an intelligent outdoor thermal comfort assessment system using psychological state as a medium, the system specifically comprising: a data acquisition module for acquiring psychological index data and multi-scale parameters, wherein the psychological index data includes a plurality of emotional factor data that can be converted into overall emotional disorder degree data, and the multi-scale parameters include environmental physical parameters and physiological response parameters; The feature fusion module is used to input the environmental physical parameters and physiological response parameters into the dynamic feature extraction module, and generate a fused feature vector by encoding the environmental physical parameters and extracting the physiological time series features, and performing cross-modal feature fusion; The dynamic adjustment module is used to construct a psychological mediation prediction module based on psychological indicator data, input the fused feature vector into the psychological mediation prediction module, and generate a dynamic adjustment coefficient by predicting the degree of emotional disorder; The prediction output module is used to input the fused feature vector and dynamic adjustment coefficient into the thermal comfort assessment module, and output multi-scale prediction results by integrating the spatiotemporal feature interaction pattern and loss function optimization.

[0008] In a third aspect, the present invention provides a computer device comprising: a memory and a processor and a computer program stored in the memory. When the computer program is executed on the processor, it implements the intelligent assessment method for outdoor thermal comfort using psychological state as a medium as described in the above method.

[0009] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for intelligently assessing outdoor thermal comfort using psychological state as a medium as described in the above method is implemented.

[0010] Compared with the prior art, the present invention has at least one of the following technical effects: 1. This invention innovatively proposes a new assessment paradigm driven by psychological mediation. By designing a three-stage deep model for dynamic extraction of environmental and physiological characteristics, joint reasoning of psychological states, and probabilistic prediction of thermal comfort, it establishes a nonlinear mapping mechanism between multimodal parameters. Compared with traditional methods, this technology introduces psychological states as dynamic modulatory variables, employing a cross-modal attention mechanism and gated fusion technology to address the challenge of spatial and temporal alignment between the transient nature of environmental parameters and the lag of physiological responses. This significantly improves the individual adaptability and predictive reliability of thermal comfort assessment in complex outdoor scenarios, providing a scientific basis for decision-making in the field of thermal environment monitoring.

[0011] 2. This invention addresses the technical bottleneck of thermal comfort assessment in outdoor environments by innovatively introducing psychological state as a key mediating variable, constructing a multi-layered deep learning model that integrates physical environmental parameters, physiological indicators, and psychological state. This model not only enables intelligent and accurate prediction of individual thermal comfort perception but also systematically reveals the role of psychological state in regulating thermal perception, breaking through the limitations of traditional assessments that rely solely on physical or physiological indicators. It has significant scientific value and broad engineering application prospects.

[0012] 3. This invention utilizes a real-time weight allocation mechanism based on a dual-channel neural network, building a dynamic feature fusion framework through collaborative analysis of environmental physical parameters and human physiological indicators. This technology breaks through the traditional static weighting model and utilizes a combination of environmental perception confidence and physiological feedback sensitivity.

[0013] 4. This invention employs a three-level feature processing approach. First, an environmental parameter encoder extracts nonlinear thermal stress features. A bidirectional LSTM captures physiological temporal dependencies. Finally, a gated attention network enables dynamic cross-modal interaction. By decoupling the spatial distribution of the physical environment from the temporal lag of physiological responses in stages, this eliminates scale conflicts in multi-source data and improves feature fusion efficiency.

[0014] 5. This invention innovatively designs a dual-path loss function for psychological and thermal comfort. For the thermal comfort path, a psychological calibration truth-value constraint is introduced for the adjustment coefficient to ensure the physical interpretability of the dynamic weights. For the psychological path, variable sparsification is achieved through KL divergence and L1 regularization. While maintaining prediction accuracy, this system outputs three interpretable indicators: thermal comfort score, contribution of psychological influencing factors, and prediction confidence interval.

[0015] 6. This invention achieves a technological breakthrough in the field of intelligent environmental perception by constructing a thermal comfort assessment system driven by psychological mediation. It innovatively introduces psychological state as a dynamic adjustment variable, using the psychological mediation module to correct the contribution weights of environmental and physiological characteristics in real time, addressing the assessment bias caused by traditional models that ignore psychological fluctuations. A bidirectional gated fusion network is designed to simultaneously model the transient characteristics of environmental parameters and the lag effects of physiological responses, overcoming the challenge of spatiotemporal asynchrony in multi-source data. A bidirectional feedback mechanism is constructed for thermal comfort regression and psychological modeling, achieving multidimensional traceability of assessment results through variable sparsity constraints and true value supervision of dynamic adjustment coefficients. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 1 is a flow chart of an intelligent outdoor thermal comfort assessment method using psychological state as a medium, provided by one embodiment of the present invention; Figure 2 1 is a schematic structural diagram of a multi-level deep learning model of an intelligent outdoor thermal comfort assessment method using psychological state as a medium, provided by an embodiment of the present invention; Figure 3 1 is a schematic diagram of the structure of an outdoor thermal comfort intelligent assessment system using psychological state as a medium provided by an embodiment of the present invention; Figure 4 It is a structural diagram of a computer device provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0017] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail.

[0018] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0019] It will also be understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0020] As used in this specification and the appended claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0021] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0022] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present application. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0023] In the embodiments of the present application, the execution subject of the process includes a terminal device, which includes but is not limited to: a server, a computer, a smart phone, a tablet computer, and other devices capable of executing the method disclosed in the present application. Figure 1 A flow chart of an intelligent outdoor thermal comfort assessment method using psychological state as a medium disclosed in one embodiment of the present invention is shown, and is described in detail as follows: S101, obtaining psychological index data and multi-scale parameters, wherein the psychological index data includes a plurality of emotional factor data that can be converted into overall emotional disorder degree data, and the multi-scale parameters include environmental physical parameters and physiological response parameters.

[0024] In this embodiment, to ensure scientific quantification of psychological indicators and their high correlation with thermal comfort, this solution uses the BPOMS scale to characterize psychological states. This index system ensures a refined expression of psychological states and provides a solid foundation for the scientific construction of the model's intermediary layer. While collecting psychological indicators, multi-dimensional parameter information is also collected simultaneously, providing rich input information for the model and improving the dynamic response capabilities of the prediction.

[0025] Specifically, the collected data includes: (1) Psychological index data: The six emotional factors of the BPOMS scale, namely tension (T), anger (A), fatigue (F), confusion (C), depression (D) and vitality (V), were evaluated using a 5-point Likert scale (the BPOMS scale is used to measure an individual's subjective psychological state), revealing the moderating effect of emotional tendencies on thermal comfort evaluation.

[0026] (2) Environmental physical parameters: Four key variables, namely air temperature, relative humidity, radiation temperature and wind speed, fully reflect the physical characteristics of the thermal environment.

[0027] (3) Physiological response parameters: Synchronously collect skin temperature, heart rate variability (HRV) and local sweat rate to accurately quantify individual physiological adaptability and heat stress response.

[0028] The collected data is preprocessed, feature extracted, and normalized to form a data set. The data set includes: (1) Environmental physical parameter matrix: A four-dimensional environmental physical characteristic vector consisting of air temperature (°C), relative humidity (%), radiation temperature (°C), and wind speed (m / s). This matrix constitutes the basic data for dynamic environmental changes.

[0029] (2) Multi-channel time series matrix of physiological indicators: including three types of physiological characteristics: skin temperature (℃), heart rate variability (low frequency / high frequency ratio calculated based on RR interval) and local sweat rate. This matrix constitutes the physiological response data of the human body to thermal stimulation.

[0030] (3) Psychological state quantitative matrix: Based on the six emotional factors, the structured scoring data of the degree of total mood disturbance (TMD) can be calculated. The result is stored in the form of a numerical vector.

[0031] (4) Thermal comfort rating vector: Based on the subject's thermal comfort rating in the test environment, it uses a 1-10 scale to reflect the individual's subjective thermal comfort perception intensity. Data is collected through the subject's subjective evaluation to support supervised learning training.

[0032] S102: Input the environmental physical parameters and physiological response parameters into the dynamic feature extraction module, encode the environmental physical parameters and extract the physiological time series features, and perform cross-modal feature fusion to generate a fused feature vector.

[0033] In this embodiment, the dynamic feature extraction module is designed to extract effective features from environmental physical parameters and physiological response parameters respectively, and fuse these features to generate a feature vector that can fully reflect the state of the human body in an outdoor thermal environment.

[0034] Data cleaning is performed on collected environmental physical parameters (such as air temperature, humidity, wind speed, and mean radiant temperature) to remove outliers and noise. For example, for data collected by an air temperature sensor, if the temperature value at a certain moment is outside the acceptable range (e.g., below -50°C or above 80°C), the data is considered an outlier and can be removed or replaced with the average of the data from adjacent moments. The cleaned data is then normalized to map the different dimensions of environmental physical parameters to a unified numerical range, facilitating subsequent feature extraction and fusion. A typical normalization method is maximum-minimum normalization, which subtracts the minimum value from each parameter value and then divides the result by the difference between the maximum and minimum values. Next, a hierarchical coding approach is used to encode the environmental physical parameters. The environmental physical parameters are grouped according to their physical meaning and relevance, for example, air temperature and mean radiant temperature are grouped together, while humidity and wind speed are grouped together. For each parameter group, statistical features (such as mean, variance, maximum, minimum, etc.) and temporal features (such as rate of change, trend, etc.) are extracted. Statistical features reflect the overall state of a parameter over a specific period of time, while temporal features reveal how the parameter changes over time. The extracted statistical and temporal features are combined to form a feature vector for each set of environmental physical parameters. These feature vectors are concatenated to produce a complete encoding vector for the environmental physical parameters. This encoding vector comprehensively characterizes the physical state of the outdoor thermal environment.

[0035] The collected physiological response parameters (such as heart rate and skin conductivity) are filtered to remove high-frequency noise and baseline drift. A digital filter (such as a Butterworth filter) can be used to filter the physiological signals, with an appropriate cutoff frequency set based on the frequency characteristics of the signals. The filtered physiological signals are then segmented, dividing the continuous signals into several time windows. The length of each time window can be set based on actual needs, for example, 1 minute or 5 minutes. This facilitates the extraction of temporal features of the physiological signals and reflects physiological changes in the human body over different time periods. Next, for each time window, the time domain features (such as mean, standard deviation, and peak value) and frequency domain features (such as power spectral density and dominant frequency) are extracted. Time domain features reflect the temporal fluctuations of physiological signals, while frequency domain features reveal the frequency components and energy distribution of physiological signals. A sliding window approach is used to extract features from physiological signals. After extracting a set of features from each time window, the window slides forward at a predetermined step size to extract features for the next time window. This approach yields a time series of physiological signal features. The extracted feature sequences are further processed, such as using principal component analysis (PCA) to reduce their dimensionality, remove redundant features, and retain the most representative features, thereby obtaining a physiological time series feature vector. This feature vector can reflect the dynamic changes in the human body's physiological response to outdoor thermal environments.

[0036] Because the acquisition frequency and sampling time of environmental physical parameters and physiological response parameters may differ, feature alignment is required. Interpolation or resampling can be used to align the two feature sequences to the same time scale, ensuring that each time point has corresponding environmental physical parameter features and physiological time series features. First, the importance weights of each feature in the environmental physical parameter feature vector and the physiological time series feature vector are calculated. The importance score of each feature is obtained by calculating the correlation between features or using a self-attention mechanism to learn the intrinsic connections between features. Then, a weighted summation of the environmental physical parameter feature vector and the physiological time series feature vector is performed based on the importance weights to obtain a fused feature vector. In this weighted summation, features with higher importance weights occupy a larger proportion in the fused feature vector, thereby better reflecting the correlation and complementary relationships between features of different modalities. To further improve the quality of the fused features, the fused feature vector is normalized to ensure that its numerical range meets the requirements of subsequent modules. The resulting fused feature vector comprehensively and comprehensively reflects the outdoor thermal environment and the physiological and psychological state of the human body within it, providing an effective feature representation for subsequent thermal comfort assessment.

[0037] Through the above implementation steps of environmental physical parameter encoding, physiological time series feature extraction, and cross-modal feature fusion, the dynamic feature extraction module can generate high-quality fused feature vectors, laying a solid foundation for subsequent thermal comfort assessment.

[0038] S103: construct a psychological mediation prediction module based on the psychological indicator data, input the fused feature vector into the psychological mediation prediction module, and generate a dynamic adjustment coefficient through emotional disorder prediction.

[0039] In this embodiment, the psychological mediation prediction module innovatively establishes a thermal comfort assessment mechanism based on dynamic compensation of mood disturbance. By introducing the quantitative indicator of the Brief Mood Inventory (BPOMS) - Total Mood Disturbance (TMD), a cross-modal causal reasoning chain of "environment-physiological characteristics → psychological mediation → thermal comfort" is constructed.

[0040] First, a prediction mechanism for the degree of emotional disturbance is constructed. Using a hierarchical analysis approach, a cluster analysis is performed on the different levels of emotional disturbance based on the severity of the emotional disturbance. Typical characteristic patterns for each level are identified. These characteristic patterns are then used to establish a prediction model for the degree of emotional disturbance. This branch employs a rule-based reasoning approach to match the relevant information in the input fused feature vector (such as environmental stimuli and physiological states reflected by environmental physical parameters and physiological response parameters) with the typical characteristic patterns to predict the degree of emotional disturbance in the current situation. For example, a high ambient temperature and an increased heart rate may indicate a tendency for significant emotional fluctuations. This branch uses this information to infer and produce a predicted value for the degree of emotional disturbance.

[0041] Then, based on the potential relationship between psychological state assessment values ​​and thermal comfort perception, a rule for generating a dynamic adjustment coefficient is established. For example, when the psychological state assessment value indicates a stable psychological state, potentially improving thermal tolerance to a certain extent, the dynamic adjustment coefficient can be appropriately reduced to adjust the output of the subsequent thermal comfort assessment module. Conversely, when the psychological state assessment value indicates a highly fluctuating psychological state, the dynamic adjustment coefficient can be appropriately increased, making the thermal comfort assessment results more likely to reflect the person's discomfort. In this way, an adjustment coefficient is generated that dynamically reflects the impact of the person's psychological state on thermal comfort perception.

[0042] Through the above-mentioned architectural design of the psychological mediation prediction module and the implementation steps such as the prediction of the degree of emotional disorder, an effective psychological mediation prediction module can be constructed and an accurate dynamic adjustment coefficient can be generated, providing key support for the subsequent thermal comfort assessment.

[0043] In step S104, the fused feature vector and the dynamic adjustment coefficient are input into the thermal comfort assessment module, and the multi-scale prediction results are output by integrating the spatiotemporal feature interaction pattern and loss function optimization.

[0044] In this embodiment, the thermal comfort assessment module aims to comprehensively integrate the feature vector and the dynamic adjustment coefficient, and output accurate and multi-scale thermal comfort prediction results by integrating the spatiotemporal feature interaction pattern and loss function optimization, providing strong support for outdoor thermal environment research and application.

[0045] In some embodiments, reference Figure 2 The dynamic feature extraction module includes an environmental physical parameter encoder, a physiological time series feature extractor, and a cross-modal feature fusion unit. In the above step S102, the environmental physical parameters and physiological response parameters are input into the dynamic feature extraction module, and a fused feature vector is generated through environmental physical parameter encoding and physiological time series feature extraction, and cross-modal feature fusion is performed, which specifically includes: The environmental physical parameters are input into the environmental physical parameter encoder, and the air temperature parameters, relative humidity parameters, radiation temperature parameters and wind speed parameters are processed through a four-layer fully connected neural network to generate an environmental feature vector; Inputting the physiological response parameters into a physiological time series feature extractor, processing the physiological time series signals in multiple time windows through a bidirectional long short-term memory network to obtain a physiological feature vector, wherein the physiological time series signals include skin temperature, heart rate variability (HRV), and local sweat rate; The environmental feature vector and the physiological feature vector are input into the cross-modal feature fusion unit, and the environmental feature vector and the physiological feature vector are independently projected in two paths. The importance of each feature dimension is determined by similarity calculation to generate attention weights. Based on the attention weight, the environmental feature vector and the physiological feature vector are element-wise multiplied and added to obtain the fused feature vector.

[0046] In this embodiment, the environmental physical parameter encoder processes the input 4-dimensional environmental parameter vector through a four-layer fully connected neural network. , including air temperature ( ), relative humidity (RH), radiant temperature ( ) and wind speed ( ). Gaussian error linear unit (GeLU) is used as the activation function between network layers:

[0047] in is the error function, Represents the input value of a single neuron in a neural network before activation. GeLU has smoother gradient characteristics than ReLU, which is beneficial for modeling complex interactions between environmental parameters.

[0048] The network structure is designed as a bottleneck architecture that expands and compresses layer by layer. The first layer maps the 4-dimensional input to a 64-dimensional space (weight matrix ), the second layer is expanded to 128 dimensions ( ) to capture high-order nonlinear relationships, and the third layer is compressed back to 64 dimensions ( ) eliminates redundancy and finally outputs a 32-dimensional environment feature vector ( ), satisfying the following expression:

[0049] in , , , Corresponding to the bias vectors of the four fully connected layers, , , , The weight matrices corresponding to the four fully connected layers.

[0050] This network structure design breaks through the limitations of traditional linear combinations and captures the nonlinear coupling relationships between temperature, humidity, radiation, and wind speed, such as the synergistic effects in high-temperature and high-humidity environments. The output feature vector can represent the equivalent thermal stress level.

[0051] The physiological time series feature extractor uses a bidirectional long short-term memory network (BiLSTM) to process physiological time series signals within a 30-second time window. , the input dimension is 3 (skin temperature, heart rate variability HRV, local sweat rate), and the time step is T=30. First, forward propagation is performed to calculate the time step in chronological order The hidden state under , capturing the hysteresis effect of physiological response. Then reverse propagation is performed and calculated in reverse order , a prospective feature modeling physiological regulation, is calculated as follows:

[0052]

[0053] in, , are the weight matrices for forward and backward propagation, respectively. The former learns the dynamic association pattern between physiological signals and historical states, while the latter allows the reverse LSTM to learn feature patterns different from the forward propagation. , are the bias terms for forward and backward propagation respectively; The moment The physiological signal vector under [] represents the vector splicing operation, which converts the current input and historical status or future state Joint coding.

[0054] Based on the above, feature fusion is performed to convert the forward state at the final moment into Reverse to the initial state Splicing, output 128-dimensional feature vector through layer normalization (LayerNorm) The formula is as follows:

[0055] in, For the forward LSTM at time step The hidden state of arrive The temporal dependence characteristics of physiological signals. For the backward LSTM at time step The hidden state of arrive The reverse timing characteristics. , They are the trainable weight matrix and bias term, respectively. The former is used to map the concatenated 256-dimensional vector to the 128-dimensional feature space, and the latter is used to adjust the mean of the feature distribution.

[0056] This content captures the temporal dependence of dynamic changes in skin temperature, HRV spectral shifts (reflecting autonomic nervous system regulation), and the cumulative effect of sweat rate. The bidirectional architecture simultaneously models both the lagging and forward-looking characteristics of physiological responses, such as sweating delay and heart rate recovery. Finally, layer normalization eliminates feature scale differences, outputting high-quality physiological representations suitable for downstream tasks.

[0057] The cross-modal feature fusion unit designs an interpretable attention mechanism to achieve dynamic fusion of environmental and physiological features. First, feature interaction modeling is performed to transform the 32-dimensional environmental features into and 128-dimensional physiological characteristics Projecting to a 64-dimensional common space, the two paths are independently projected, allowing the environment and physiological features to retain their modality specificity and avoid information loss caused by early fusion:

[0058] in represents a linear projection transformation; is the input feature vector ( or ); is a trainable matrix, dim corresponds to the input feature dimension (when hour ,when hour ); is the bias vector used to shift the feature distribution.

[0059] On this basis, attention weight is performed Generate and determine the importance of each feature dimension through similarity calculation. The calculation method is as follows:

[0060] in It is a normalized exponential function that converts the original score into a probability distribution to ensure weight normalization; is the learnable attention vector, Transpose it; is the hyperbolic tangent function, which is used to compress the eigenvalue to interval, enhancing nonlinear expression capabilities; and They are respectively the environmental feature projection results and the physiological feature projection results after the above linear projection transformation.

[0061] After the attention weights are generated, feature fusion is performed to finely integrate the environmental and physiological features according to the attention weights, and perform element-by-element multiplication and addition:

[0062] in Represents the Hadamard product, that is, the corresponding elements are multiplied; finally, the eigenvector is fused , passed to downstream modules.

[0063] By dynamically adjusting the contribution weights of multimodal features, this mechanism enables the model to adapt to the dominant influencing factors under different environmental-physiological states, overcoming the limitations of traditional methods with fixed weight coefficients.

[0064] In some embodiments, reference Figure 2 The psychological mediation prediction module includes a feature decoupling compression network and a dynamic adjustment coefficient generation unit. In the above step S103, the fusion feature vector is input into the psychological mediation prediction module, and the dynamic adjustment coefficient is generated by predicting the overall emotional disorder level, which specifically includes: The fused feature vector is input into the feature decoupling and compression network, and the overall emotional disorder degree data is obtained through feature compression and emotion sensitivity factor mapping; The overall emotional disorder degree data is input into the dynamic adjustment coefficient generation unit, and the dynamic adjustment coefficient is generated by constraining the range through the hyperbolic tangent function.

[0065] Furthermore, the fused feature vector is input into the feature decoupling and compression network, and the TMD value is obtained through feature compression and emotion sensitivity factor mapping, which specifically includes: The fusion feature vector is subjected to feature dimension reduction and noise suppression through the bottleneck fully connected layer of the feature decoupling and compression network to generate primary compression features. Based on the layer normalization-fully connected joint distillation layer of the feature decoupling compression network, the emotion sensitivity factor that is highly correlated with the overall degree of emotional disorder is extracted from the primary compressed features; Based on the final layer of the feature decoupling compression network, the emotion sensitivity factor is mapped to the normalized overall emotion disorder scalar through affine transformation and Tanh activation to obtain the original overall emotion disorder prediction value; The target overall mood disorder degree prediction value is obtained by normalizing the original overall mood disorder degree prediction value.

[0066] In this embodiment, a two-stage feature decoupling and compression algorithm is proposed to eliminate the collinear interference between environmental physical parameters and physiological signals. The input layer receives the 64-dimensional mixed feature vector from the cross-modal fusion module. First, the bottleneck fully connected layer is used to achieve feature space dimensionality reduction and noise suppression:

[0067] in is a trainable weight matrix responsible for projecting the 64-dimensional input features into a 32-dimensional latent space. is the corresponding bias term; It is a Gaussian error linear activation function. Its smooth gradient characteristics are more suitable for modeling the continuous distribution of emotion-related features than ReLU. The core function of this layer is to generate feature decoupling masks and cross-modal noise filtering. The sparse training automatically learns the inhibition mask of the steady-state component in the environmental parameters, while enhancing the activation intensity of the physiological transient features; the gating effect of the GELU function can effectively filter the coupling interference of environmental sensor noise and physiological signal drift.

[0068] In the primary compression feature On this basis, we introduce layer normalization-fully connected joint distillation layer to extract emotion-sensitive factors that are highly correlated with TMD from mixed features. :

[0069] in is the feature vector after layer normalization and full connection layer processing, which represents the abstract features highly correlated with emotional disorders extracted from the mixed features. is the trainable weight matrix, is the corresponding bias term; It is a layer normalization operation. In the specific operation, normalization is performed along the feature dimension:

[0070] in, yes No. Component, representing the first Normalized values ​​of dimensions. are the mean and variance, normalization parameters The feature scale can be learned and adjusted, and this layer automatically strengthens the weight coefficients of features related to mood disorders through back propagation.

[0071] The final layer uses an affine transformation-Tanh activation joint structure to map the emotion sensitivity factor to a normalized TMD scalar:

[0072] in is the trainable weight matrix, is the corresponding bias term; Constrain the output to interval, where The specific definition is: , and Respectively represent and The exponential function of Normalization processing to obtain the normalized value ,:

[0073] in, is the actual measure of the total mood disturbance score of the BPOMS, and All samples in the dataset are The minimum and maximum values ​​of interval.

[0074] Compared to the Sigmoid function, the zero-centered nature of Tanh activation makes the back-propagation gradient The region still maintains effective update capability, the bias term Dynamically compensate for individual psychological trait differences and set prior distributions through Bayesian optimization , making the model universal to groups of subjects with different emotional baselines.

[0075] Furthermore, a dynamic adjustment coefficient is generated :

[0076] Output By constraining the range with the hyperbolic tangent function, this coefficient controls the contribution weight of environmental characteristics and physiological characteristics. When the environmental characteristics contribution is enhanced (applicable to stable physiological state scenarios), The physiological characteristics weight is enhanced (suitable for individual abnormal state detection), psychological state compensation is further realized, and thermal comfort assessment deviation is analyzed and corrected.

[0077] The psychological mediation prediction module uses a feature decoupling compression network to quantify the discretization level of overall emotional disorder based on a two-stage feature decoupling compression algorithm, and ultimately generates a dynamic regulation coefficient with physiological and physical significance. , dynamically balance the contribution weights of environmental parameters and physiological signals.

[0078] In some embodiments, reference Figure 2 The thermal comfort assessment module includes a spatiotemporal feature integration unit, a multi-scale prediction unit, and a cross-module joint loss function. In step S104, the fused feature vector and the dynamic adjustment coefficient are input into the thermal comfort assessment module, and the multi-scale prediction result is output by integrating the spatiotemporal feature interaction mode and the loss function optimization, specifically including: Based on the spatiotemporal feature integration unit, the contribution weights of environmental features and physiological features in the fusion feature vector are adjusted in real time through the dynamic adjustment coefficient, and the spatiotemporal feature interaction pattern is integrated to obtain the final hidden state; Based on the multi-scale prediction unit, the thermal comfort score and its confidence level are obtained by performing multi-scale prediction on the final hidden state. Based on the cross-module joint loss function, loss optimization is performed according to the predicted values ​​output by the multi-scale prediction unit and the psychological mediation module.

[0079] In this embodiment, the thermal comfort assessment module consists of three components: a spatiotemporal feature integration unit, a multi-scale prediction unit, and a cross-module joint loss function. Its core function is to receive fused feature vectors and dynamic adjustment coefficients, integrate the spatiotemporal feature interaction model, and optimize the loss function to ultimately output multi-scale thermal comfort prediction results, providing a comprehensive and accurate basis for outdoor thermal environment assessment.

[0080] The impact of the environmental and physiological characteristics in the fused feature vector on thermal comfort varies with individual status and environmental conditions. The dynamic adjustment coefficient plays a key role in this, adjusting the contribution of these two characteristics in the evaluation process in real time. Specifically, initial weights are set for each environmental and physiological characteristic, and these weights are adjusted based on the magnitude and direction of the dynamic adjustment coefficient. After weight adjustment, the fused feature vector is integrated based on the interaction of spatiotemporal features. In the temporal dimension, the temporal trend of features is considered. For example, for fused feature vectors collected continuously at multiple time points, the rate of change of each feature in the time series is calculated to analyze whether it is increasing, decreasing, or remaining stable. By analyzing these trends, we can understand the dynamic adaptation process of the human body to the thermal environment. In the spatial dimension, data from monitoring points at different geographical locations are combined. If multiple monitoring points are set up in different parts of a city, the fused feature vectors of each monitoring point are classified according to their geographical location. The correlation and differences between the features of different regions are analyzed. For example, some regions may have similar environmental characteristics but different physiological characteristics. This may be related to the characteristics of the population in that region (such as age distribution and activity type). The feature vectors of different regions at each time point are concatenated to form a comprehensive feature vector that incorporates spatiotemporal information. The attention mechanism then weights the key features in the comprehensive feature vector, highlighting those that have a significant impact on thermal comfort, thereby obtaining the final hidden state. The attention mechanism automatically assigns different weights based on the relevance and importance of features, enabling the final hidden state to more accurately reflect the thermal comfort of the human body in complex outdoor environments.

[0081] Based on the final hidden state output by the spatiotemporal feature integration unit, a holistic thermal comfort prediction is performed. This holistic prediction aims to assess the average thermal comfort level across the entire study area (e.g., a city region). The thermal comfort score can be assigned on a scale of 1 to 10, with higher scores indicating greater thermal comfort. The confidence level of the holistic prediction is also determined based on the model's prediction confidence assessment method (e.g., based on sample size and feature consistency). The confidence level can be expressed as a percentage; for example, an 80% confidence level indicates an 80% probability that the prediction is accurate.

[0082] Local-scale prediction focuses on the thermal comfort conditions at specific locations or small areas within the study area (such as the perimeter of a single building or a park). For each local area, a corresponding final hidden state subset is extracted. More refined analysis methods, such as case-based reasoning, are employed. A large number of thermal comfort cases from different local areas under different environmental conditions are collected, including information such as environmental parameters, physiological responses, psychological indicators, and actual thermal comfort evaluations. The final hidden state subset of the current local area is matched with cases in the case library to identify the most similar cases. Based on the thermal comfort evaluation results of similar cases, the thermal comfort score and confidence level for the current local area are predicted. Because the characteristics of local areas may be more complex and diverse, case-based reasoning methods can better capture these subtle differences and improve the accuracy of local predictions. Through global and local-scale predictions, multi-scale thermal comfort scores and their confidence levels are obtained, providing a basis for decision-making at different levels.

[0083] A cross-module joint loss function is constructed to comprehensively consider the difference between the predicted values ​​of the multi-scale prediction unit and the psychological mediation module and the actual thermal comfort evaluation. The loss function can be designed as a weighted sum of multiple components.

[0084] For the multi-scale prediction unit, the loss component includes both global prediction error and local prediction error. The global prediction error measures the difference between the predicted overall thermal comfort score and the actual overall thermal comfort evaluation, while the local prediction error measures the difference between the predicted thermal comfort score for each local area and the actual local thermal comfort evaluation. For the psychological mediation module, the loss component measures the difference between its predicted assessment of the impact of psychological state on thermal comfort and the actual role played by psychological state in thermal comfort. Appropriate weights are assigned to each loss component, and the weights can be adjusted based on the importance of each component to thermal comfort assessment. An iterative optimization approach is used to adjust the parameters in the thermal comfort assessment module based on the value of the cross-module joint loss function. In each iteration, the current loss function value is calculated. Based on the gradient information of the loss function or other optimization strategies (such as heuristic search algorithms), the weight allocation parameters in the spatiotemporal feature integration unit and the model parameters in the multi-scale prediction unit are updated. Through multiple iterations, the parameters are continuously optimized to gradually reduce the value of the cross-module joint loss function, thereby improving the overall prediction performance of the thermal comfort assessment module and ensuring more accurate and reliable multi-scale prediction results.

[0085] Through the specific implementation steps of the above spatiotemporal feature integration unit, multi-scale prediction unit and cross-module joint loss function, the thermal comfort assessment module can make full use of the fused feature vector and dynamic adjustment coefficient, integrate the spatiotemporal feature interaction pattern, and output accurate and multi-scale thermal comfort prediction results through loss function optimization, providing strong support for the research and application of outdoor thermal environment.

[0086] Furthermore, the contribution weights of environmental features and physiological features in the fusion feature vector are adjusted in real time through the dynamic adjustment coefficient, and the spatial-temporal feature interaction pattern is integrated to obtain the final hidden state, which specifically includes: Construct a gated recurrent network, use the dynamic adjustment coefficient as a conditional constraint, and adjust the contribution weights of environmental and physiological characteristics in real time through the gated recurrent network to obtain integrated features; Obtain the historical hidden state, project the overall emotional disorder prediction result according to the integrated features, the candidate state weight matrix and the emotional state projection matrix, and obtain the candidate state vector; Constructing an emotion enhancement gate, inputting the integrated features and the candidate state vector into the emotion enhancement gate, performing gating adjustment on the overall emotion disorder prediction result through the reference weight matrix and the modulation matrix, and obtaining an updated gating vector, wherein the updated gating vector is used to control the fusion ratio of the historical hidden state and the current candidate state; Based on the updated gating vector, the final hidden state is obtained by fusing the candidate state vector and the historical hidden state.

[0087] In this embodiment, a gated recurrent network is designed to integrate temporal features and psychological state feedback, and a dynamic adjustment coefficient generated by the psychological mediation module is introduced. As a conditional constraint, the environmental-physiological characteristics are first dynamically weighted, through Adjust the contribution weights of environmental and physiological characteristics in real time and output integrated features :

[0088] in For environmental characteristics, Physiological characteristics ( Projected into 32-dimensional space).

[0089] Then we conduct the emotion enhancement gating design. To update the gate vector, used to control the historical hidden state With the current candidate status The designed gating mechanism can retain important historical states:

[0090] in is the baseline weight matrix, learning the general pattern of feature interaction; Predicting outcomes for overall mood disorders The modulation matrix is ​​mapped into a weight correction value to dynamically adjust the gate decision logic; Sigmoid compresses the output to interval, realizing the soft selection mechanism.

[0091] Based on the above, emotional state projection is performed. is the candidate state vector, which represents the new feature representation of the current time step and is calculated as follows:

[0092] in is the candidate state weight matrix, encoding the spatiotemporal feature interaction pattern. is the emotional state projection matrix, Mapping to a 256-dimensional feature space enables explanatory correction of psychological states to physiological responses. express Function, used to constrain the output to (-1,1) to alleviate the gradient explosion problem. Correction candidate status , suppressing feature shift caused by emotional fluctuations.

[0093] Furthermore, the thermal comfort score and its confidence level are obtained by performing multi-scale prediction on the final hidden state, specifically including: Constructing a probability prediction network based on joint psychological-physiological constraints, wherein the probability prediction network includes a psychological modulation prediction head, an uncertainty quantization unit, and an interpretable sampling unit; Based on the probability prediction network, the weight distribution of the final hidden state is dynamically adjusted through the dynamic adjustment coefficient to output the mean prediction result of psychological modulation; Based on the uncertainty quantification unit, the final hidden state is processed according to a first amplification coefficient to obtain an uncertainty quantification result of emotion enhancement, wherein the first amplification coefficient represents the amplification coefficient of the overall emotional disorder degree to the uncertainty; Based on the mean prediction results of psychological modulation and the uncertainty quantification results of emotion enhancement, Gaussian distribution processing is performed through interpretable sampling units to obtain the thermal comfort score and its confidence.

[0094] In this embodiment, a probability prediction network with psychological-physiological joint constraints is constructed to achieve thermal comfort score output. First, the psychological modulation prediction head is designed. Dynamically adjust the final hidden state after integrating spatiotemporal features Weight distribution, output psychological modulation mean prediction :

[0095] and are the weight matrix and bias term of the linear transformation, respectively, used to output the mean thermal comfort score. is the weight matrix of the dynamic adjustment coefficient, and the scalar Expanded to 256-dimensional space. express Function, used to Constrain to , to prevent the adjustment range from being too large. For Hadamard product, realize Hidden state Personalized modulation.

[0096] Then, uncertainty quantification design is carried out. To quantify the uncertainty of sentiment enhancement, it is calculated as follows:

[0097] in , Generates standard deviation base values ​​for the linear transformation parameters (weight matrix and bias term). To ensure the standard deviation , while keeping the function smooth and differentiable. Overall mood disorder Amplification factor for uncertainty. Overall mood disturbance Amplify the forecast standard deviation to reflect the impact of psychological fluctuations on the stability of the assessment. Output uncertainty quantification results To assist users in understanding the reliability of predictions.

[0098] Finally, interpretable sampling is performed to output the thermal comfort score And its confidence:

[0099] in is standard Gaussian noise, implementing the reparameterization technique. It is characterized by is the variational posterior distribution conditional on . is a variable The contribution coefficient of , balances the influence of psychological characteristics and random noise. This part realizes the differentiability of the random sampling process and supports gradient back propagation to the psychological mediation module. , establish a mapping relationship between psychological traits and physiological responses.

[0100] Furthermore, the cross-module joint loss function includes a thermal comfort regression loss function and a psychological modeling loss function; the loss optimization based on the cross-module joint loss function and the predicted values ​​output by the multi-scale prediction unit and the psychological mediation module specifically includes: Based on the thermal comfort regression loss function, the output value of the multi-scale prediction unit is optimized according to the thermal comfort prediction value and the actual thermal comfort value output by the multi-scale prediction unit; Based on the psychological modeling loss function, the output value of the psychological mediation module is optimized according to the cross entropy loss and KL divergence loss of the emotional disorder prediction task.

[0101] In this embodiment, a cross-module joint loss function is designed to achieve end-to-end multi-task optimization. First, the thermal comfort regression loss is defined , the mean square error (MSE) loss prioritizes thermal comfort prediction accuracy:

[0102] in is the number of training samples in a batch, and Respectively Thermal comfort prediction value and true value of samples; and Respectively The predicted value of the dynamic adjustment coefficient of each sample and its true value, Calibrated by experimental psychology to enhance the physical meaning of the adjustment coefficient, is the weight parameter of the adjustment coefficient constraint term.

[0103] The first MSE forced model output Approximating the true thermal comfort score , ensuring the reliability of core functions. The second item passed Weighted Constraints Close to the true value , preventing the psychological mediation module from outputting meaningless adjustment coefficients. This loss jointly optimizes thermal comfort prediction and psychological adjustment coefficient generation, avoiding feature deviation caused by module separation.

[0104] Secondly, we introduce the psychological modeling loss , CE+KL loss maintains the rationality of psychological state modeling:

[0105]

[0106]

[0107] in is the cross entropy loss for the task of predicting the degree of emotional disorder, For the The true label of the overall emotional disorder level (one-hot encoding), The output The overall probability of mood disorders. is the KL (Kullback-Leibler) divergence loss, The output of the psychological mediation module The mean of the dimensional variable, For the dimensional variable standard deviation. The variable generated by the psychological mediation module is dimension; , , is the weight coefficient of each loss item, represents the L1 regularization function.

[0108] In this loss definition, Ensure that the overall emotional disturbance prediction is consistent with the thermal comfort score logic; Constrained variables Obey the standard normal distribution , to prevent overfitting; (L1 regularization) forces non-critical dimensions to approach zero, retaining core psychological factors and improving the interpretability of psychological traits.

[0109] Finally, we use the annealing training strategy to match the human cognitive law (perceive the environment first and then understand the psychology) to improve the convergence stability of the model:

[0110] in For the The total weight of the psychological modeling loss during step training, is the initial total weight ( ), The number of warm-up training steps (about 5 epochs).

[0111] Initial training ( ) gradually increase The weight of the task is to avoid the psychological prediction task from interfering with the thermal comfort return task too early. hour, , fully enable psychological modeling loss. Through progressive weight adjustment, balanced feature extraction ( ) and mental modeling ( ) to prevent optimization oscillation.

[0112] Reference Figure 3 An embodiment of the present invention provides an outdoor thermal comfort intelligent assessment system 3 using psychological state as a medium. The system 3 specifically includes: Data acquisition module 301, acquiring psychological index data and multi-scale parameters, wherein the psychological index data includes multiple emotional factor data that can be converted into overall emotional disorder degree data, and the multi-scale parameters include environmental physical parameters and physiological response parameters; The feature fusion module 302 is used to input the environmental physical parameters and physiological response parameters into the dynamic feature extraction module, and generate a fused feature vector by encoding the environmental physical parameters and extracting physiological time series features, and performing cross-modal feature fusion; The dynamic adjustment module 303 is used to construct a psychological mediation prediction module based on the psychological index data, input the fused feature vector into the psychological mediation prediction module, and generate a dynamic adjustment coefficient by predicting the degree of emotional disorder; The prediction output module 304 is used to input the fused feature vector and the dynamic adjustment coefficient into the thermal comfort assessment module, and output the multi-scale prediction results by integrating the spatiotemporal feature interaction mode and loss function optimization.

[0113] It is understandable that if Figure 1 The contents of the embodiment of the outdoor thermal comfort intelligent evaluation method using psychological state as the medium are applicable to the embodiment of the outdoor thermal comfort intelligent evaluation system using psychological state as the medium. The functions specifically implemented by the embodiment of the outdoor thermal comfort intelligent evaluation system using psychological state as the medium are similar to those in the embodiment of the outdoor thermal comfort intelligent evaluation method using psychological state as the medium. Figure 1 The embodiment of the outdoor thermal comfort intelligent evaluation method using psychological state as the medium is the same as that shown in FIG. Figure 1 The beneficial effects achieved by the embodiment of the outdoor thermal comfort intelligent assessment method using psychological state as a medium are also the same.

[0114] It should be noted that the information interaction, execution process and other contents between the above-mentioned systems are based on the same concept as the embodiment of the method of the present invention. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0115] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0116] Reference Figure 4An embodiment of the present invention further provides a computer device 4, comprising: a memory 402, a processor 401, and a computer program 403 stored in the memory 402. When the computer program 403 is executed on the processor 401, the outdoor thermal comfort intelligent assessment method using psychological state as a medium as described in any one of the above methods is implemented.

[0117] The computer device 4 can be a desktop computer, a notebook computer, a palmtop computer, a cloud server or other computing devices. The computer device 4 can include, but is not limited to, a processor 401 and a memory 402. Those skilled in the art will understand that Figure 4 This is merely an example of the computer device 4 and does not constitute a limitation on the computer device 4 . The computer device 4 may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the computer device 4 may also include input and output devices, network access devices, etc.

[0118] The processor 401 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor may be a microprocessor or any conventional processor.

[0119] In some embodiments, the memory 402 may be an internal storage unit of the computer device 4, such as a hard drive or memory of the computer device 4. In other embodiments, the memory 402 may also be an external storage device of the computer device 4, such as a plug-in hard drive, a Smart Media Card (SMC), a Secure Digital (SD) card, a flash memory card, etc. equipped on the computer device 4. Furthermore, the memory 402 may include both an internal storage unit of the computer device 4 and an external storage device. The memory 402 is used to store an operating system, application programs, a boot loader, data, and other programs, such as the program code of the computer program. The memory 402 may also be used to temporarily store data that has been output or is about to be output.

[0120] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for intelligently assessing outdoor thermal comfort using psychological state as a medium as described in any one of the above methods is implemented.

[0121] In this embodiment, if the integrated unit is implemented as a software functional unit and sold or used as a standalone product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application can implement all or part of the process steps in the above-mentioned method embodiments by using a computer program to instruct the relevant hardware. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a camera / terminal device, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signals, telecommunication signals, and software distribution media. Examples include USB flash drives, removable hard drives, magnetic disks, or optical disks. In some jurisdictions, based on legislation and patent practice, computer-readable media cannot be electric carrier signals or telecommunication signals.

[0122] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0123] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0124] In the embodiments disclosed in the present application, it should be understood that the disclosed devices / terminal equipment and methods can be implemented in other ways. For example, the device / terminal equipment embodiments described above are merely schematic. For example, the division of the modules or units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0125] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

Claims

1. An intelligent outdoor thermal comfort assessment method based on psychological state, characterized in that: The method specifically includes: Acquiring psychological index data and multi-scale parameters, wherein the psychological index data includes multiple emotional factor data that can be converted into overall emotional disorder degree data, and the multi-scale parameters include environmental physical parameters and physiological response parameters; The environmental physical parameters and physiological response parameters are input into the dynamic feature extraction module, and the fused feature vector is generated by encoding the environmental physical parameters and extracting the physiological time series features, and performing cross-modal feature fusion. Construct a psychological mediation prediction module based on psychological indicator data, input the fusion feature vector into the psychological mediation prediction module, and generate a dynamic adjustment coefficient by predicting the degree of emotional disorder; The fused feature vector and dynamic adjustment coefficient are input into the thermal comfort assessment module, and the multi-scale prediction results are output by integrating the spatiotemporal feature interaction pattern and loss function optimization.

2. The method according to claim 1, characterized in that The dynamic feature extraction module includes an environmental physical parameter encoder, a physiological time series feature extractor, and a cross-modal feature fusion unit. The environmental physical parameters and physiological response parameters are input into the dynamic feature extraction module, and a fusion feature vector is generated through environmental physical parameter encoding and physiological time series feature extraction, and cross-modal feature fusion is performed. Specifically, the following steps are performed: The environmental physical parameters are input into the environmental physical parameter encoder, and the air temperature parameters, relative humidity parameters, radiation temperature parameters and wind speed parameters are processed through a four-layer fully connected neural network to generate an environmental feature vector; Inputting the physiological response parameters into a physiological time series feature extractor, processing the physiological time series signals in multiple time windows through a bidirectional long short-term memory network to obtain a physiological feature vector, wherein the physiological time series signals include skin temperature, heart rate variability (HRV), and local sweat rate; The environmental feature vector and the physiological feature vector are input into the cross-modal feature fusion unit, and the environmental feature vector and the physiological feature vector are independently projected in two paths. The importance of each feature dimension is determined by similarity calculation to generate attention weights. Based on the attention weight, the environmental feature vector and the physiological feature vector are element-wise multiplied and added to obtain the fused feature vector.

3. The method according to claim 1, characterized in that The psychological mediation prediction module includes a feature decoupling compression network and a dynamic adjustment coefficient generation unit. The fusion feature vector is input into the psychological mediation prediction module, and the dynamic adjustment coefficient is generated by predicting the overall emotional disorder level. Specifically, the following steps are performed: The fused feature vector is input into the feature decoupling and compression network, and the overall emotional disorder degree data is obtained through feature compression and emotion sensitivity factor mapping; The overall emotional disorder degree data is input into the dynamic adjustment coefficient generation unit, and the dynamic adjustment coefficient is generated by constraining the range through the hyperbolic tangent function.

4. The method according to claim 1, wherein The fused feature vector is input into the feature decoupling and compression network, and the overall emotional disorder degree data is obtained through feature compression and emotional sensitivity factor mapping, specifically including: The fusion feature vector is subjected to feature dimension reduction and noise suppression through the bottleneck fully connected layer of the feature decoupling and compression network to generate primary compression features. Based on the layer normalization-fully connected joint distillation layer of the feature decoupling compression network, the emotion sensitivity factor that is highly correlated with the overall degree of emotional disorder is extracted from the primary compressed features; Based on the final layer of the feature decoupling compression network, the emotion sensitivity factor is mapped to the normalized overall emotion disorder scalar through affine transformation and Tanh activation to obtain the original overall emotion disorder prediction value; The target overall mood disorder degree prediction value is obtained by normalizing the original overall mood disorder degree prediction value.

5. The method according to claim 4, characterized in that The thermal comfort assessment module includes a spatiotemporal feature integration unit, a multi-scale prediction unit, and a cross-module joint loss function. The fusion feature vector and the dynamic adjustment coefficient are input into the thermal comfort assessment module, and the multi-scale prediction result is output by integrating the spatiotemporal feature interaction mode and the loss function optimization, specifically including: Based on the spatiotemporal feature integration unit, the contribution weights of environmental features and physiological features in the fusion feature vector are adjusted in real time through the dynamic adjustment coefficient, and the spatiotemporal feature interaction pattern is integrated to obtain the final hidden state; Based on the multi-scale prediction unit, the thermal comfort score and the confidence level of the thermal comfort score are obtained by performing multi-scale prediction on the final hidden state. Based on the cross-module joint loss function, loss optimization is performed according to the predicted values ​​output by the multi-scale prediction unit and the psychological mediation module.

6. The method according to claim 5, characterized in that The method of adjusting the contribution weights of environmental features and physiological features in the fusion feature vector in real time through the dynamic adjustment coefficient, integrating the spatiotemporal feature interaction pattern, and obtaining the final hidden state specifically includes: Construct a gated recurrent network, use the dynamic adjustment coefficient as a conditional constraint, and adjust the contribution weights of environmental and physiological characteristics in real time through the gated recurrent network to obtain integrated features; Obtain the historical hidden state, project the overall emotional disorder prediction result according to the integrated features, the candidate state weight matrix and the emotional state projection matrix, and obtain the candidate state vector; Constructing an emotion enhancement gate, inputting the integrated features and the candidate state vector into the emotion enhancement gate, performing gating adjustment on the overall emotion disorder prediction result through the reference weight matrix and the modulation matrix, and obtaining an updated gating vector, wherein the updated gating vector is used to control the fusion ratio of the historical hidden state and the current candidate state; Based on the updated gating vector, the final hidden state is obtained by fusing the candidate state vector and the historical hidden state.

7. The method according to claim 5, characterized in that The thermal comfort score and its confidence level are obtained by performing multi-scale prediction on the final hidden state, specifically including: Constructing a probability prediction network based on joint psychological-physiological constraints, wherein the probability prediction network includes a psychological modulation prediction head, an uncertainty quantization unit, and an interpretable sampling unit; Based on the probability prediction network, the weight distribution of the final hidden state is dynamically adjusted through the dynamic adjustment coefficient to output the mean prediction result of psychological modulation; Based on the uncertainty quantification unit, the final hidden state is processed according to a first amplification coefficient to obtain an uncertainty quantification result of emotion enhancement, wherein the first amplification coefficient represents the amplification coefficient of the overall emotional disorder degree to the uncertainty; Based on the mean prediction results of psychological modulation and the uncertainty quantification results of emotion enhancement, Gaussian distribution processing is performed through interpretable sampling units to obtain the thermal comfort score and its confidence.

8. The method according to claim 5, characterized in that The cross-module joint loss function includes a thermal comfort regression loss function and a psychological modeling loss function. The loss optimization based on the cross-module joint loss function and the predicted values ​​output by the multi-scale prediction unit and the psychological mediation module specifically includes: Based on the thermal comfort regression loss function, the output value of the multi-scale prediction unit is optimized according to the thermal comfort prediction value and the actual thermal comfort value output by the multi-scale prediction unit; Based on the psychological modeling loss function, the output value of the psychological mediation module is optimized according to the cross entropy loss and KL divergence loss of the emotional disorder prediction task.

9. An intelligent outdoor thermal comfort assessment system based on psychological state, characterized by: The system specifically includes: a data acquisition module for acquiring psychological index data and multi-scale parameters, wherein the psychological index data includes a plurality of emotional factor data that can be converted into overall emotional disorder degree data, and the multi-scale parameters include environmental physical parameters and physiological response parameters; The feature fusion module is used to input the environmental physical parameters and physiological response parameters into the dynamic feature extraction module, and generate a fused feature vector by encoding the environmental physical parameters and extracting the physiological time series features, and performing cross-modal feature fusion; The dynamic adjustment module is used to construct a psychological mediation prediction module based on psychological indicator data, input the fused feature vector into the psychological mediation prediction module, and generate a dynamic adjustment coefficient by predicting the degree of emotional disorder; The prediction output module is used to input the fused feature vector and dynamic adjustment coefficient into the thermal comfort assessment module, and output multi-scale prediction results by integrating the spatiotemporal feature interaction pattern and loss function optimization.

10. A computer device, characterized in that: include: A memory, a processor, and a computer program stored in the memory, which, when executed on the processor, implements the outdoor thermal comfort intelligent assessment method using psychological state as a medium as described in any one of claims 1 to 8.

Images