Physiological data analysis method and system for evaluating visual fatigue intervention effect of greenbelt

Abstract

The invention discloses a physiological data analysis method and system for evaluating a greenbelt visual fatigue intervention effect, and the method comprises the steps: carrying out the millisecond-level time alignment of three types of heterogeneous physiological signals, namely eye movement tracking data, an electroencephalogram signal and a sweat biomarker through a self-adaptive clock synchronization protocol, and generating a time-space coupled multi-modal physiological data set; carrying out nonlinear decomposition on the multi-modal physiological data set by adopting an improved variational auto-encoder, generating three orthogonal feature subspaces through implicit space projection, and carrying out tensor splicing on all orthogonal subspace features to form a decoupling feature vector; and constructing a multi-scale attention gating graph convolutional neural network model, and mapping the decoupling feature vector to a three-dimensional evaluation space containing a visual entropy value, a neural synchronization degree and a biochemical index. The deep coupling-decoupling architecture of the multi-modal physiological data is constructed, so that the three problems of signal misalignment, feature mixing and mechanism fuzziness are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of asthenopia evaluation, in particular to a physiological data analysis method and system for evaluating the intervention effect of green space asthenopia. BACKGROUND

[0002] With the acceleration of urbanization, asthenopia caused by high-density artificial environment has become a global public health problem.

[0003] The existing Chinese patent with publication number CN112568915B relates to a stereoscopic display asthenopia evaluation method, system and device based on multi-task learning, aiming to solve the problems of complex evaluation process and low accuracy rate of traditional methods and methods of collecting user's electroencephalogram signals and evaluating asthenopia through deep learning network. The method of the invention comprises: preprocessing the obtained set of electroencephalogram signal data; using the asthenopia grade evaluation model to calculate the probability of each asthenopia grade corresponding to the two-dimensional matrix data after preprocessing; taking the asthenopia grade with the maximum probability value as the obtained asthenopia grade and outputting it. The invention is trained through a multi-task learning model, through the interaction of two tasks, the intermediate layer shares features and has the ability of classification and reconstruction, the model generalization ability is improved, and in the case of limited electroencephalogram marker data, the accuracy of stereoscopic display asthenopia evaluation is improved.

[0004] However, in the process of implementing the related technical solutions, at least the following technical problems are found:

[0005] The physiological monitoring means uses a single-dimensional index (such as independent analysis of electroencephalogram signals), which is difficult to capture the complex coupling relationship between neural compensation, environmental stimulation and individual differences in the formation process of asthenopia. If multi-modal physiological signals are used, there are systematic defects in time synchronization, spatial correlation and feature decoupling at three levels. At the same time, green space is considered as an important carrier to relieve visual stress due to its natural properties, and the influence of green space is not considered in the process of implementing the related technical solutions. SUMMARY

[0006] In order to solve the above problems, the embodiments of the present application provide a physiological data analysis method for evaluating the intervention effect of green space asthenopia, which comprises:

[0007] Through the adaptive clock synchronization protocol, the eye movement tracking data, the electroencephalogram signals and the sweat biomarkers are time-aligned at the millisecond level, and a multi-modal physiological data set coupled in space and time is generated;

[0008] An improved variational autoencoder is used to perform nonlinear decomposition on the multi-modal physiological data set, three orthogonal feature subspaces are generated through hidden space projection, and all orthogonal subspaces features are tensor spliced to form a decoupling feature vector;

[0009] A multi-scale attention gate graph convolutional neural network model is constructed to map the decoupled feature vector to a three-dimensional evaluation space containing visual entropy, neural synchronization degree and biochemical indicators.

[0010] Based on the causal correlation analysis of neural responses and environmental factors, an augmented reality visualization decision scheme is generated, which contains a dynamic heat map and vegetation optimization parameters.

[0011] Further, the adaptive clock synchronization protocol comprises:

[0012] When generating an anti-interference time reference, the period characteristics of the photoelectric pulse wave and the clock drift compensation amount of the satellite time signal are fused;

[0013] When performing cross-modal signal reconstruction, the Legendre polynomial approximation algorithm is used to perform phase synchronization interpolation on eye movement trajectories and electroencephalogram signals with different sampling rates.

[0014] Further, the orthogonal feature subspace includes visual load components, environmental response components and individual baseline components; the generation method of the decoupled feature vector comprises:

[0015] The visual load component quantifies the chaos degree of the visual system by calculating the Lyapunov index of the pupil diameter oscillation wave, and is converted into a three-dimensional feature vector;

[0016] The environmental response component is characterized by analyzing the mutual information entropy of the brain electrical μ wave band energy and the ground vegetation reflectance spectrum, forming a two-dimensional feature vector;

[0017] The individual baseline component uses a dynamic time warping algorithm to match the baseline mode in the user historical biological feature database to generate a one-dimensional feature vector;

[0018] After standardizing the three types of feature vectors with different dimensions, they are spliced along the feature axis to form a decoupled feature vector.

[0019] Further, the dynamic intervention evaluation step comprises:

[0020] The three-dimensional point cloud data of green space is converted into graph node features with spatial topological relationships;

[0021] The differential equation coupling relationship between the vegetation canopy light distribution model and the ciliary muscle accommodation activity is established;

[0022] The visual cortex activation pattern and scene semantic segmentation results are fused to generate a probability gradient map of visual fatigue relief efficiency.

[0023] Further, the causal correlation analysis comprises:

[0024] The electroencephalogram γ wave band energy distribution is mapped to the corresponding spatial coordinates in the augmented reality scene to form a response topography map;

[0025] A Bayesian association network between specific vegetation morphological parameters and optic nerve relaxation was constructed using a causal inference algorithm.

[0026] Furthermore, the construction of the three-dimensional evaluation space includes:

[0027] A visual system entropy dimension calculated based on the nonlinear characteristics of pupil oscillation of the visual load component in the decoupled feature vector;

[0028] Phase synchronization index dimension of the prefrontal-parietal brain network dynamics based on the environmental response component;

[0029] Partial least squares regression dimension of biochemical marker concentration gradient and spatial openness based on individual baseline components;

[0030] The visual system entropy dimension, phase synchronization index dimension and partial least squares regression dimension are integrated into a unified coordinate system, namely the three-dimensional evaluation space, where the x-axis is the visual system entropy dimension; the Y-axis is the phase synchronization index dimension; and the Z-axis is the partial least squares regression dimension.

[0031] Furthermore, the generated augmented reality visualization solution includes:

[0032] Overlaying a holographic projection layer with a gradient that alleviates visual fatigue in the user's real-world field of view;

[0033] A vibration waveform pattern positively correlated with visual recovery was generated through a tactile feedback device.

[0034] Furthermore, it also includes autonomous optimization steps:

[0035] Applying the meta-reinforcement learning framework to dynamically adjust the attention weight parameters of graph convolutional neural networks;

[0036] Simulate the plant light competition ecological model and iteratively generate the landscape spatial topology scheme for optimal intervention of visual fatigue.

[0037] A physiological data analysis system for evaluating the effects of green space visual fatigue intervention, including:

[0038] A spatiotemporal synchronization module, which uses an adaptive clock synchronization protocol to perform millisecond-level time alignment on three heterogeneous physiological signals: eye tracking data, EEG signals, and sweat biomarkers, to generate a spatiotemporally coupled multimodal physiological dataset;

[0039] an aggregation module, wherein the aggregation module uses an improved variational autoencoder to perform nonlinear decomposition on the multimodal physiological dataset, generates three orthogonal feature subspaces through latent space projection, and performs tensor splicing on all orthogonal subspace features to form a decoupled feature vector;

[0040] A collective mapping module, which maps the decoupled feature vectors to a three-dimensional evaluation space containing visual entropy, neural synchronization, and biochemical indicators by constructing a multi-scale attention-gated graph convolutional neural network model;

[0041] A solution output module generates an augmented reality visualization decision solution including a dynamic heat map and vegetation optimization parameters based on causal correlation analysis between neural responses and environmental factors.

[0042] The technical effects and advantages of the physiological data analysis method and system for evaluating the intervention effect of green space visual fatigue provided by the present invention are as follows:

[0043] By constructing a deep coupling-decoupling architecture for multimodal physiological data, this invention creatively solves the three core problems of "signal misalignment, feature confounding, and mechanism ambiguity" in visual fatigue assessment, while also taking into account the visual impact of green spaces. This invention uses an adaptive clock synchronization protocol to overcome the limitations of traditional interpolation algorithms, leveraging the periodic characteristics of photoelectric pulse waves to establish an anti-interference time benchmark, enabling dynamic anchoring of millisecond-level physiological events with spatial environmental parameters. This allows for the first quantifiable analysis of the phase coupling relationship between pupil tremor events and EEG gamma oscillations. An improved variational autoencoder, through orthogonal subspace constraints, separates the confounded visual load signal into three independent components: visual chaos, environmental responsiveness, and individual metabolic baseline. The combined calculation of the Lyapunov exponent and mutual information entropy reveals the dynamic equilibrium threshold of the eye-brain-environment system. A multiscale graph convolutional network couples the three-dimensional assessment space with the vegetation canopy light distribution model through differential equations, constructing a dose-effect relationship between a specific leaf area index and optic nerve relaxation through a Bayesian causal network, making the vegetation optimization parameters in the augmented reality solution neurobiologically interpretable. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a flow chart of the physiological data analysis method for evaluating the intervention effect of green space visual fatigue in Example 1;

[0045] Figure 2 This is a flow chart of the physiological data analysis method for evaluating the intervention effect of green space visual fatigue in Example 2;

[0046] Figure 3 This is a connection diagram of the physiological data analysis system for evaluating the intervention effect of green space visual fatigue in Example 3. DETAILED DESCRIPTION

[0047] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be clearly and completely described below, obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work belong to the scope of protection of the present application.

[0048] Embodiment one:

[0049] Please refer to Figure 1 The embodiment of the present application provides a physiological data analysis method for evaluating the intervention effect of green space visual fatigue, the method comprises:

[0050] Through an adaptive clock synchronization protocol, three types of heterogeneous physiological signals, namely eye movement tracking data, electroencephalogram signals and sweat biomarkers, are time-aligned at a millisecond level to generate a spatio-temporal coupled multi-modal physiological data set;

[0051] An improved variational autoencoder is used for nonlinear decomposition of the multi-modal physiological data set, three orthogonal feature subspaces are generated through hidden space projection, and all orthogonal subspace features are tensor spliced to form a decoupled feature vector;

[0052] A multi-scale attention gate graph convolutional neural network model is constructed to map the decoupled feature vector to a three-dimensional evaluation space containing visual entropy, neural synchronization degree and biochemical indicators;

[0053] Based on the causal association analysis of neural responses and environmental factors, an augmented reality visualization decision scheme containing dynamic heat maps and vegetation optimization parameters is generated.

[0054] The adaptive clock synchronization protocol comprises:

[0055] When generating an anti-interference time reference, the period characteristics of the photoelectric pulse wave and the clock drift compensation amount of the satellite time signal are fused;

[0056] The construction of the anti-interference time reference, i.e. the bio-physical hybrid clock, is based on the acquisition of the fingertip blood vessel volume pulse signal by the photoelectric pulse wave sensor, the extraction of its cycle stability characteristics (about 0.5-2Hz physiological rhythm), and the fusion with the microsecond time stamp of the Beidou satellite time module. Through the Kalman filtering algorithm, the pulse wave period is taken as the biological clock reference, and the clock drift (typical value in the range of ±10ms) caused by the atmospheric delay of the satellite signal is compensated. The hybrid time reference generated thereby has both biological rhythm adaptability and physical clock accuracy, and provides a unified time coordinate for subsequent multi-source data.

[0057] When reconstructing cross-modal signals, a Legendre polynomial approximation algorithm is used to perform phase synchronization interpolation on eye movement trajectories and electroencephalogram signals with different sampling rates.

[0058] Different sampling rates, for example, eye movement trajectory from eye tracker (sampling rate 120Hz), electroencephalogram from electroencephalogram device (sampling rate 1000Hz), using Legendre polynomial approximation algorithm for phase synchronization interpolation, including:

[0059] In the time window of each segment of eye movement data, based on the kinematic continuity assumption of eye movement, a 5th order Legendre orthogonal basis function is constructed;

[0060] The analytical expression of the eye movement trajectory is obtained by least square fitting, and the interpolation point of the eye movement trajectory at the electroencephalogram sampling time is calculated;

[0061] The cross-correlation analysis is performed on the interpolated eye movement data and the original electroencephalogram signal, and the phase offset is dynamically adjusted until the maximum synchronization entropy is reached.

[0062] At the same time, the sweat biomarker also needs to be time-corrected, which can use the fluid dynamics control unit of the sweat biomarker collection device to adjust the sample transmission rate through micro-channel pressure feedback, eliminate the sampling interval fluctuation caused by the discontinuity of sweat secretion on the skin surface, and establish a continuous time sequence of cortisol concentration changes combined with the time stamp of the mixing time reference.

[0063] Through the cooperative compensation mechanism of biological rhythm and physical clock, the synchronization deviation of traditional single clock source in complex environment is effectively overcome; the orthogonal polynomial interpolation is used instead of linear interpolation, which significantly reduces the phase distortion of high-frequency signals in cross-rate matching, and at the same time, the active control strategy of fluid dynamics ensures the time continuity of sweat data, so that the synchronization confidence of the three types of physiological signals in the fixed time window reaches a high level, laying a reliable time and space reference for subsequent multi-modal analysis.

[0064] The orthogonal characteristic subspace includes a visual load component, an environmental response component and an individual baseline component; the generation method of the decoupled characteristic vector includes:

[0065] The visual load component quantifies the chaos degree of the visual system by calculating the Lyapunov exponent of the pupil diameter oscillation wave, and converts it into a three-dimensional characteristic vector, including:

[0066] The maximum Lyapunov exponent (used to quantify the chaotic response degree of the visual system to external stimuli) is calculated by phase space reconstruction of the pupil diameter change in a continuous 10-second time window;

[0067] The obtained Lyapunov exponent value is mapped to a three-dimensional space through a radial basis function to generate a characteristic vector reflecting the visual tension degree, adjustment sensitivity and fatigue accumulation rate.

[0068] The environmental response component is characterized by analyzing the mutual information entropy between the EEG μ-band energy and the surface vegetation reflectance spectrum to form a two-dimensional feature vector, including:

[0069] Perform wavelet packet decomposition on the EEG signal to extract the energy proportion of the μ band in the frontal cortex;

[0070] Calculate the mutual information entropy between this energy sequence and the 550nm (chlorophyll reflection peak) and 680nm (red light absorption valley) bands in the vegetation spectrum (indicating the efficiency of the nervous system's adaptation to environmental light signals);

[0071] The mutual information entropy matrix was reduced to two-dimensional eigenvectors by principal component analysis, corresponding to spectral adaptability and environmental pressure response, respectively.

[0072] The individual baseline component uses a dynamic time warping algorithm to match the baseline pattern in the user's historical biometric database to generate a one-dimensional feature vector, including:

[0073] The dynamic time warping (DTW) algorithm is used to compare the morphological similarity between the current heart rate variability curve and the historical baseline pattern;

[0074] A scalar feature value is generated according to the curvature of the matching path to quantify the degree to which the individual's current physiological state deviates from the normal state.

[0075] The three types of feature vectors of different dimensions are normalized and concatenated along the feature axis to form the final decoupled feature vector, including:

[0076] The three types of eigenvectors were subjected to Z-score standardization to eliminate dimensional differences;

[0077] Splice along the feature axis in the order of [visual load characteristics, environmental response characteristics, individual baseline characteristics];

[0078] Orthogonal projection is used to eliminate the residual correlation between features and form the final decoupled feature vector with biological interpretability.

[0079] Through a hierarchical feature decoupling strategy, a reliable conversion of multimodal physiological signals into fatigue assessment indicators is achieved. Compared with the traditional PCA dimensionality reduction method, the nonlinear feature extraction of the visual load component reduces the quantification error of the stimulation effect of green landscape complexity on the visual system. The two-dimensional mapping of environmental response characteristics can effectively distinguish the differential effects of natural light environment and artificial lighting on the nervous system. The dynamic matching mechanism of individual baselines increases the model's adaptability to differences in users' physiological rhythms by more than 2 times, making it particularly suitable for universal assessment of visual fatigue across populations.

[0080] The dynamic intervention assessment steps include:

[0081] Convert green space 3D point cloud data into graph node features with spatial topological relationships;

[0082] Green space 3D point cloud data is processed using an improved octree structure. Voxel downsampling is used to retain key geometric features of vegetation canopy morphology (such as leaf density and branch fractal dimension). A spatial clustering algorithm is used to construct hierarchical topological relationships. The construction of hierarchical topological relationships includes:

[0083] Region growing segmentation is performed under the condition that the angle between the normal vectors of adjacent point clouds is less than 15°;

[0084] The light propagation pathway connections between canopy nodes are established using a radius search method.

[0085] Finally, a graph attention network is used to extract node embedding features with spatial awareness.

[0086] Establish the coupling relationship between the vegetation canopy light distribution model and the differential equation of ciliary muscle regulatory activity;

[0087] The system of coupled differential equations in the differential equations is used to calculate the three-dimensional distribution of photosynthetically active radiation within the canopy based on the radiation transfer model. The calculation method includes: ;

[0088] Where, Height within the canopy The photosynthetically active radiation intensity at is the leaf projection function, describing the leaf in the incident direction The projection ratio on ; is the leaf area index, which is the total leaf area per unit ground area; is the diffuse light ratio coefficient (dimensionless), which represents the proportion of diffuse light in the incident light; It is a diffuse light source term that describes the scattered light distribution of the sky or surrounding environment.

[0089] The ciliary muscle accommodation dynamics equations in the differential equation coupling relationship include: ;

[0090] Where, To adjust tension (such as the degree of pupil constriction or changes in lens curvature); is the time constant of the regulation process, indicating the dynamic response speed; are gain coefficients, which respectively control the effects of average light intensity and light intensity fluctuation on regulating tension; is the average illuminance of a certain area of ​​the retina or canopy; is the root mean square of light intensity fluctuations, reflecting the spatiotemporal heterogeneity of light intensity.

[0091] The visual cortex activation pattern is integrated with the scene semantic segmentation results to generate a probability gradient map of visual fatigue relief effectiveness.

[0092] The method for generating a probability gradient map of visual fatigue relief effectiveness includes:

[0093] Use 3D U-Net to perform semantic segmentation on point cloud data (functional divisions such as trees, shrubs, and lawns);

[0094] The energy distribution of the gamma band (30-80 Hz) of the visual cortex in the EEG signal is encoded as an activation heat map;

[0095] Design a cross-modal gating network to align the semantic segmentation results with the neural activation patterns;

[0096] Monte Carlo sampling is used to generate a gradient distribution map of the probability of visual fatigue improvement, and sensitive areas for landscape optimization are marked.

[0097] The hierarchical extraction of scene topological features improves the quantification accuracy of the impact of vegetation spatial layout on visual comfort. The photobiological coupling equation can capture the critical relationship between light fluctuation frequency and ciliary muscle spasm. The generation of probability gradient maps can improve the design efficiency of landscape optimization solutions and accurately identify areas with insufficient shade and nodes with excessive color contrast.

[0098] Causal analysis includes:

[0099] Mapping the energy distribution of the EEG gamma band to the corresponding spatial coordinates in the augmented reality scene to form a response topography;

[0100] In AR scenarios, high-density EEG acquisition equipment is first used to record the gamma-band neural oscillation signals generated by the user's visual cortex in real time. The gamma band refers to EEG activity with a frequency range of 30 to 80 Hz. Its energy changes can reflect the intensity of visual information processing. To determine the spatial source of neural activity within the brain, a source localization algorithm based on solving electromagnetic inverse problems is used. The source localization algorithm includes the following:

[0101] The EEG signals recorded from the scalp are combined with a three-dimensional head model, and mathematical inversion calculations are used to determine the specific location of the signal in the cerebral cortex, such as the primary visual cortex or fusiform gyrus.

[0102] To achieve accurate mapping between neural activity locations and AR scene coordinates, a dynamic affine transformation model is established to realize coordinate conversion through two key steps. First, the rotation angle is determined based on the user's real-time line of sight direction (obtained by eye tracking), so that the horizontal reference plane of the brain coordinate system is aligned with the projection plane of the AR scene. Second, translation parameters are used to compensate for individual physiological differences (such as interpupillary distance) to ensure that the neural activity hotspots of different users can be accurately mapped to the corresponding scene locations. Finally, a spatial interpolation algorithm is used to convert discrete neural activity intensity values into a continuous energy distribution map covering the entire AR scene, forming a visual response topography, where the highlighted areas represent landscape elements that trigger strong gamma oscillations.

[0103] A Bayesian association network between specific vegetation morphological parameters and optic nerve relaxation is constructed through causal reasoning algorithms.

[0104] Three-dimensional morphological features of vegetation are extracted from the AR scene, including branch and leaf distribution complexity, color contrast, and shading characteristics.

[0105] The branch and leaf complexity is quantified by calculating the fractal dimension of the vegetation surface geometry, with higher fractal dimension indicating more complex structure. The color contrast is determined by analyzing the difference in hue and lightness between the vegetation and the background. The shading characteristics are evaluated by simulating the attenuation of light passing through the vegetation canopy.

[0106] The optic nerve functional status is represented by multiple modalities of physiological indicators, including:

[0107] The response speed of pupil diameter to changes in illumination is recorded using a dynamic pupilometer, reflecting the regulatory sensitivity of the optic nerve. The focusing lag amount of the eye when switching between different distances is measured using an optometry device to assess the degree of visual fatigue. The natural blink rate per unit time is also counted as an indirect indicator of eye muscle relaxation.

[0108] Based on the above parameters, a Bayesian causal network model is constructed to reveal the internal relationship between vegetation characteristics and optic nerve function, including:

[0109] First, the data is standardized to eliminate dimensional differences. Then, a constrained causal discovery algorithm is used to identify potential causal relationships between variables, such as determining whether changes in branch and leaf complexity directly cause changes in pupil regulatory speed. Finally, a probabilistic graphical model is used to quantify the influence of each factor, distinguishing between direct causal effects and indirect associations. Time lag analysis is introduced during model training to ensure that the causal relationships conform to the time logic of physiological responses (such as changes in illumination preceding pupil responses).

[0110] Through the neural response topography, specific areas in the AR scene that are prone to causing visual load (such as high-contrast, texture-dense areas) can be visually identified, providing spatial positioning for landscape optimization.

[0111] The construction of the three-dimensional evaluation space includes:

[0112] A visual system entropy dimension calculated based on the nonlinear characteristics of pupil oscillation of the visual load component in the decoupled feature vector;

[0113] Methods for extracting nonlinear features of pupil oscillation include:

[0114] Using high-precision pupil tracking equipment to continuously record pupil diameter micro-fluctuations during natural observation, we capture the chaotic characteristics of pupil self-regulation through nonlinear dynamics analysis methods, including:

[0115] The pupil fluctuation signal is decomposed into different time scales, and the complexity of the signal at each scale is calculated. Complexity reflects the efficiency of information integration when the visual system processes external stimuli. Higher entropy values ​​indicate more active and unpredictable neural regulation.

[0116] The pupil time series was converted into a two-dimensional recurrence graph, and the stability of pupil accommodation patterns was assessed by statistically analyzing the distribution characteristics of diagonal structures. Frequent occurrence of short-line structures indicated accommodation disorders caused by visual overload.

[0117] The calculation method of the visual system entropy dimension includes:

[0118] The above-mentioned nonlinear characteristics were subjected to cross-modal correlation analysis with the γ-band energy in the EEG signal that represents visual load. The canonical correlation analysis method was used to screen entropy indicators that were significantly correlated with visual fatigue, and a comprehensive visual system entropy dimension was constructed. The comprehensive visual system entropy dimension can quantify the neurometabolic stress caused by specific visual scenes and is used to assess the potential load of environmental factors on the visual system.

[0119] Phase synchronization index dimension of the prefrontal-parietal brain network dynamics based on the environmental response component;

[0120] EEG signals from the frontal and parietal regions are collected synchronously. Frequency band decomposition is used to obtain rhythmic components such as θ, α, and β. A Hilbert transform is performed on the EEG signals to extract the real-time phase angles of the oscillation waveforms in each frequency band. The phase difference consistency between the signals in the two brain regions is calculated using a sliding window method, generating a time-varying phase synchronization curve. Higher synchronization strength values ​​indicate more efficient information transfer between brain regions. Furthermore, standardized visual interference events (such as sudden bright light stimulation) can be set in augmented reality scenarios. The magnitude of the change in phase synchronization strength before and after the event can be analyzed to assess the brain network's ability to adapt to environmental impact.

[0121] Partial least squares regression dimension of biochemical marker concentration gradient and spatial openness based on individual baseline components.

[0122] Saliva and blood samples were collected from the subjects to measure the concentrations of stress-related biomarkers (such as cortisol and immunoglobulin A) and neurotransmitter metabolites (such as serotonin). A multidimensional gradient matrix reflecting the individual's physiological baseline was constructed. Then, a partial least squares regression model was established to establish the association between the concentrations of biochemical markers and spatial perception. This model includes:

[0123] Taking biochemical indicators as independent variables and spatial openness score as dependent variable, the combination of latent variables with the strongest explanatory power was extracted through iterative optimization.

[0124] Calculate the variable importance weights of each biochemical indicator and identify key physiological factors that affect space demand (such as cortisol circadian rhythm characteristics);

[0125] The output regression coefficient matrix reveals the prediction patterns of specific biochemical characteristics on spatial openness preference, supporting personalized environmental adaptation.

[0126] The above three dimensions are integrated into a unified coordinate system, namely the three-dimensional evaluation space, where the x-axis is the visual system entropy dimension, which is used to quantify the metabolic pressure of the visual system caused by the environment; the y-axis is the phase synchronization index dimension, which is used to reflect the environmental adaptation efficiency of the brain network; and the z-axis is the partial least squares regression dimension, which is used to characterize the matching degree between the individual physiological basis and spatial requirements.

[0127] Exemplary:

[0128] In the virtual office scenario assessment:

[0129] Areas with high visual entropy values ​​are mostly located in decorative interfaces with complex textures, suggesting that the density of surface patterns should be reduced to alleviate visual fatigue.

[0130] Frequently fluctuating lighting conditions led to a decrease in the phase synchronization index, suggesting that dynamic lighting design may interfere with brain network collaboration;

[0131] Subjects with abnormally elevated cortisol levels tended to have low openness requirements on the Z-axis, and were instructed to provide them with a semi-enclosed workstation layout.

[0132] This three-dimensional model breaks through the limitations of traditional single-dimensional evaluation, realizes the collaborative diagnosis of visual load, neural adaptation and individual differences, and provides a quantifiable physiological basis for environmental design optimization.

[0133] The generated augmented reality visualization includes:

[0134] Overlaying a holographic projection layer with a gradient that alleviates visual fatigue in the user's real-world field of view;

[0135] The holographic projection layer includes:

[0136] Apply low-saturation filter to high-load area to reduce the intensity of retinal photoreceptor stimulation; for example, superimpose a warm filter layer on the cool-toned high-light area to balance the spectral distribution;

[0137] Generate a ring-shaped luminance attenuation band outward along the visual focus to guide the natural movement of the eyeball to the low-luminance area and relieve the sustained tension of the ciliary muscle;

[0138] Embed dynamic diffusion light spots in the periphery of the visual field to induce autonomous eye movement through slow-moving soft light points, promoting uniform distribution of tear film;

[0139] Generate a vibration waveform pattern that is positively correlated with visual recovery degree through a haptic feedback device.

[0140] Generate a vibration waveform that dynamically matches the recovery index through a haptic feedback device worn on the wrist or neck, including:

[0141] When the recovery index rises, the frequency of haptic vibration gradually decreases, and the waveform transitions from rapid pulses to long-period fluctuations, simulating the physiological rhythm in a natural state of relaxation;

[0142] The vibration intensity changes non-linearly with the recovery index, with moderate intensity intermittent vibration in the initial stage to awaken attention; and weak continuous vibration in the later recovery stage to maintain the subconscious level of relaxation suggestion;

[0143] Achieve directional haptic stimulation through a multi-vibration unit array (such as a wave sequence extending from the wrist to the elbow), guiding the user to perform neck stretching or eye line turning movements.

[0144] The haptic signal is time-synchronized with the optical changes of the holographic projection layer, forming a cross-modal perception synergy; for example, when the holographic layer starts a gradual change in brightness, the haptic device triggers a directional vibration wave at the same time, strengthening the user's active adaptation to environmental intervention.

[0145] Embodiment Two:

[0146] As shown in Figure 2 , this embodiment further improves the design based on Embodiment 1, except that Embodiment 1 does not consider the dynamic coupling feedback mechanism and ecological synergy constraint relationship of individual multi-modal physiological signals and environmental parameters in actual operation, only completes the spatiotemporal alignment of physiological signals, and does not establish a real-time mapping channel between environmental sensor data (light intensity, vegetation spectrum, microclimate parameters) and user physiological state, which will result in the inability to capture the causal relationship between glare stimulation and dynamic pupil contraction, the missing of the nonlinear coupling between temperature and humidity gradient and skin conductivity. Based on this, the physiological data analysis method for evaluating the intervention effect of green space visual fatigue also includes a self-optimization step:

[0147] Apply the meta-reinforcement learning framework to dynamically adjust the attention weight parameters of the graph convolutional neural network;

[0148] The meta-reinforcement learning framework is a dual-channel meta-reinforcement learning architecture, consisting of a meta-policy controller and a graph convolutional execution network. The meta-policy controller receives a multimodal physiological dataset, calculates the current level of compensation demand of the visual system, and generates attention weight adjustment instructions. The graph convolutional execution network adopts a multi-head dynamic attention mechanism to automatically enhance the topological feature extraction capability of high-load areas according to the attention weight adjustment instructions, while suppressing interference from redundant landscape information.

[0149] Simulate the plant light competition ecological model and iteratively generate a landscape spatial topology scheme for optimal intervention of visual fatigue; the simulated plant light competition ecological model includes a three-dimensional ecological game field and a dynamic recalibration protocol;

[0150] The methods for establishing a three-dimensional ecological game field include:

[0151] Load the vegetation photosynthetically active radiation absorption spectrum and the user's field of view requirements into the digital twin scene;

[0152] The light interception competition process of different tree-shrub combinations was simulated and the visual relief gain coefficient of each species was calculated;

[0153] A set of candidate solutions is generated through Monte Carlo strategy sampling and fed back to the EEG alpha wave coherence evaluation module to verify the effectiveness of neural compensation.

[0154] When the user's gaze stays on a specific landscape area, the dynamic recalibration protocol is triggered. The dynamic recalibration protocol includes:

[0155] According to the fluctuation pattern of extraocular muscle electromyography, the adjacency matrix construction rule of the graph convolutional network is reversely modified;

[0156] Combined with the sweat electrolyte concentration change curve, the intensity of the heat stress compensation factor in the light competition model was adjusted;

[0157] Output a landscape transformation plan that meets the requirements of "minimum intervention area, maximum nerve recovery efficiency".

[0158] By simulating the resource competition mechanism of natural communities, we ensure that the landscape plan not only conforms to the physiological characteristics of plants, but also forms a spatial rhythm that conforms to the laws of human visual cognition. At the same time, the meta-strategy knowledge base accumulated through long-term operation can enable the system to predict the evolution path of visual fatigue in different occupational groups and arrange preventive landscape intervention measures in advance.

[0159] Example 3:

[0160] like Figure 3As shown, based on the same inventive concept as the physiological data analysis method for evaluating the intervention effect of green space visual fatigue in the aforementioned embodiment, this application provides a physiological data analysis system for evaluating the intervention effect of green space visual fatigue. The system and method embodiments in the embodiments of this application are based on the same inventive concept. The system includes:

[0161] The spatiotemporal synchronization module uses an adaptive clock synchronization protocol to align the three heterogeneous physiological signals of eye tracking data, EEG signals, and sweat biomarkers at the millisecond level to generate a spatiotemporally coupled multimodal physiological dataset.

[0162] an aggregation module, which uses an improved variational autoencoder to perform nonlinear decomposition on the multimodal physiological dataset, generates three orthogonal feature subspaces through latent space projection, and performs tensor splicing on all orthogonal subspace features to form a decoupled feature vector;

[0163] A collective mapping module maps the decoupled feature vectors to a three-dimensional evaluation space containing visual entropy, neural synchronization, and biochemical indicators by constructing a multi-scale attention-gated graph convolutional neural network model;

[0164] The solution output module generates an augmented reality visualization decision solution that includes dynamic heat maps and vegetation optimization parameters based on the causal relationship analysis between neural responses and environmental factors.

[0165] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

[0166] The above is only a preferred specific implementation method of the embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes based on the technical solution and concept of the present application within the technical scope disclosed in the present application, and they should be covered by the scope of protection of the present application.

Claims

1. A physiological data analysis method for evaluating the intervention effect of green space visual fatigue, characterized in that: include: Through an adaptive clock synchronization protocol, three types of heterogeneous physiological signals, namely eye tracking data, EEG signals, and sweat biomarkers, are time-aligned at the millisecond level to generate a spatiotemporally coupled multimodal physiological dataset. An improved variational autoencoder is used to perform nonlinear decomposition on the multimodal physiological dataset, three orthogonal feature subspaces are generated by latent space projection, and all orthogonal subspace features are tensor-concatenated to form a decoupled feature vector; A multi-scale attention-gated graph convolutional neural network model is constructed to map the decoupled feature vector to a three-dimensional evaluation space containing visual entropy, neural synchronization, and biochemical indicators; Based on the causal correlation analysis between neural responses and environmental factors, an augmented reality visualization decision-making solution is generated, which includes dynamic heat maps and vegetation optimization parameters.

2. The physiological data analysis method for evaluating the intervention effect of green space visual fatigue according to claim 1, characterized in that: The adaptive clock synchronization protocol includes: When generating an anti-interference time reference, the periodic characteristics of the photoelectric pulse wave and the clock drift compensation of the satellite timing signal are integrated; When reconstructing cross-modal signals, the Legendre polynomial approximation algorithm is used to perform phase-synchronous interpolation of eye movement trajectories and EEG signals with different sampling rates.

3. The physiological data analysis method for evaluating the intervention effect of green space visual fatigue according to claim 1, characterized in that: The orthogonal feature subspace includes visual load component, environmental response component and individual baseline component; The generation methods of decoupled feature vectors include: The visual load component is used to quantify the chaos of the visual system by calculating the Lyapunov exponent of the pupil diameter oscillation wave and converting it into a three-dimensional feature vector. The environmental response component is characterized by analyzing the mutual information entropy between the EEG μ-band energy and the surface vegetation reflectance spectrum to form a two-dimensional feature vector; The individual baseline component uses a dynamic time warping algorithm to match the benchmark pattern in the user's historical biometric database to generate a one-dimensional feature vector; The three types of feature vectors with different dimensions are normalized and concatenated along the feature axis to form a decoupled feature vector.

4. The physiological data analysis method for evaluating the intervention effect of green space visual fatigue according to claim 1, characterized in that: The dynamic intervention assessment steps include: Convert green space 3D point cloud data into graph node features with spatial topological relationships; Establish the coupling relationship between the vegetation canopy light distribution model and the differential equation of ciliary muscle regulatory activity; The visual cortex activation pattern is integrated with the scene semantic segmentation results to generate a probability gradient map of visual fatigue relief effectiveness.

5. The physiological data analysis method for evaluating the intervention effect of green space visual fatigue according to claim 1, characterized in that: Causal analysis includes: Mapping the energy distribution of the EEG gamma band to the corresponding spatial coordinates in the augmented reality scene to form a response topography; A Bayesian association network between specific vegetation morphological parameters and optic nerve relaxation was constructed using a causal inference algorithm.

6. The physiological data analysis method for evaluating the intervention effect of green space visual fatigue according to claim 3, characterized in that: The construction of the three-dimensional evaluation space includes: A visual system entropy dimension calculated based on the nonlinear characteristics of pupil oscillation of the visual load component in the decoupled feature vector; Phase synchronization index dimension of the prefrontal-parietal brain network dynamics based on the environmental response component; Partial least squares regression dimension of biochemical marker concentration gradient and spatial openness based on individual baseline components; The visual system entropy dimension, phase synchronization index dimension and partial least squares regression dimension are integrated into a unified coordinate system, namely the three-dimensional evaluation space, where the x-axis is the visual system entropy dimension; the Y-axis is the phase synchronization index dimension; and the Z-axis is the partial least squares regression dimension.

7. The physiological data analysis method for evaluating the intervention effect of green space visual fatigue according to claim 1, characterized in that: The generated augmented reality visualization includes: Overlaying a holographic projection layer with a gradient that alleviates visual fatigue in the user's real-world field of view; A vibration waveform pattern positively correlated with visual recovery was generated through a tactile feedback device.

8. The physiological data analysis method for evaluating the intervention effect of green space visual fatigue according to claim 1, characterized in that: It also includes autonomous optimization steps: Applying the meta-reinforcement learning framework to dynamically adjust the attention weight parameters of graph convolutional neural networks; Simulate the plant light competition ecological model and iteratively generate the landscape spatial topology scheme for optimal intervention of visual fatigue.

9. A physiological data analysis system for evaluating the intervention effect of green space visual fatigue, characterized in that: The system includes: A spatiotemporal synchronization module, which uses an adaptive clock synchronization protocol to perform millisecond-level time alignment on three heterogeneous physiological signals: eye tracking data, EEG signals, and sweat biomarkers, to generate a spatiotemporally coupled multimodal physiological dataset; an aggregation module, wherein the aggregation module uses an improved variational autoencoder to perform nonlinear decomposition on the multimodal physiological dataset, generates three orthogonal feature subspaces through latent space projection, and performs tensor splicing on all orthogonal subspace features to form a decoupled feature vector; A collective mapping module, which maps the decoupled feature vectors to a three-dimensional evaluation space containing visual entropy, neural synchronization, and biochemical indicators by constructing a multi-scale attention-gated graph convolutional neural network model; A solution output module generates an augmented reality visualization decision solution including a dynamic heat map and vegetation optimization parameters based on causal correlation analysis between neural responses and environmental factors.

Images