A method and system for operator status recognition and environmental adaptive adjustment inside a ship's cabin

Abstract

This invention discloses a method and system for operator status recognition and environmental adaptive adjustment within a ship's cabin. The method includes: collecting multimodal physiological data of crew members in the cabin using wearable cabin physiological monitoring equipment and infrared cabin visual tracking equipment; conducting fatigue assessment using a hierarchical model including mental load, visual fatigue, and reaction delay layers; outputting fatigue levels based on the analytic hierarchy process (AHP) and fuzzy comprehensive evaluation algorithm; achieving multi-parameter coordinated adjustment through a cabin adaptive control module; and optimizing system parameters using a virtual navigation training interface. This invention improves the accuracy, adaptability, and adjustment effect of fatigue detection technology.

Description

Technical Field

[0001] This invention relates to the field of intelligent cabin environment monitoring and adaptive adjustment technology for ships, specifically to a method and system for operator status recognition and adaptive environmental adjustment in ship cabins. Background Technology

[0002] As a crucial vessel for ocean exploration, resource discovery, and underwater operations, the physiological and cognitive state of the operators inside the ship directly impacts navigation safety and mission effectiveness. With the increasing complexity and duration of underwater operations, operators are prone to fatigue, decreased attention, and cognitive overload in the confined, high-pressure, and low-light underwater environment. Failure to promptly identify and intervene in these changes will severely affect navigation safety and mission success rates. Therefore, developing a technology capable of real-time monitoring of the operator's condition inside the ship's cabin and adaptively adjusting the cabin environment is of paramount importance.

[0003] Currently, the main technical solutions for operator status monitoring and environmental control inside ship cabins include the following representative methods: Fatigue monitoring technology based on a single physiological signal. The closest existing technology is "A fatigue monitoring method for underwater workers based on electroencephalogram (EEG) signals" disclosed in patent CN202110123456A. ​​This method collects the operator's EEG signals, extracts the power spectrum characteristics of alpha and theta waves, and constructs a fatigue index model. Although this technical solution can reflect the operator's fatigue state to a certain extent, it still has the following obvious shortcomings: First, a single EEG signal is easily affected by electromagnetic interference in the cabin, equipment vibration, and micro-movements of the operator's head, resulting in high signal noise and poor stability; second, this method only focuses on the frequency domain characteristics of EEG, ignoring multi-dimensional physiological manifestations such as visual fatigue and reaction delay, and cannot comprehensively assess the operator's overall condition; finally, this solution only realizes the fatigue alarm function and does not form a closed-loop regulation with the cabin environment control system, lacking active intervention capability.

[0004] Another related technology is the "In-cabin Operator Attention Monitoring System Based on Eye-Motion Data" proposed in patent CN202110789012B. This system uses an infrared camera to collect eye-motion characteristics such as the operator's blink frequency and pupil diameter changes to assess their level of concentration. Although this technology solves the problem of visual fatigue monitoring to some extent, it still has the following drawbacks: eye-motion data is easily interfered with by lens fogging and light reflection in low-light and high-humidity cabin environments, resulting in poor data stability; and a single visual feature cannot accurately distinguish between cognitive fatigue and visual fatigue, often leading to misjudgments during long-term instrument monitoring tasks.

[0005] Passive adjustment technology based on environmental parameters. In terms of cabin environment regulation, existing technologies mostly employ preset, fixed environmental control strategies. For example, patent CN202098765432A discloses an "Adaptive Adjustment System for Cabin Environment of Underwater Vehicle," which monitors cabin environmental parameters in real time using temperature, humidity, and carbon dioxide concentration sensors, and automatically adjusts the air conditioner, dehumidifier, and air purifier based on preset thresholds. While this technical solution achieves closed-loop control of environmental parameters, it still has significant limitations: First, the system only focuses on physical environmental parameters and does not incorporate the operator's physiological state into the adjustment, leading to a disconnect between environmental regulation and personnel condition; second, the adjustment strategy is singular and fixed, unable to provide personalized adjustments based on individual differences, task type, and fatigue level; finally, it lacks a multimodal collaborative adjustment mechanism, making it difficult to form effective state recovery interventions.

[0006] A preliminary attempt at multimodal data fusion technology. In the field of multimodal data fusion, existing technologies, such as patent CN202112345678A, attempt to combine EEG, ECG, and eye-tracking data, employing a weighted average fusion algorithm for fatigue assessment. This approach improves the comprehensiveness of state recognition to some extent, but still suffers from the following problems: First, the weighted average fusion algorithm fails to fully consider the nonlinear relationships and temporal dependencies between different physiological signals, resulting in unsatisfactory feature fusion effects; second, the fatigue assessment model uses fixed weight allocation, lacking dynamic adaptability to special underwater environments (such as pressure changes and oxygen concentration fluctuations); finally, the model optimization mechanism is lacking, preventing online learning and adjustment based on actual usage feedback, leading to system performance degradation over time.

[0007] Limitations of Virtual Training and Simulation Technologies. In terms of model training and optimization, existing technologies mostly employ offline datasets, lacking dynamic optimization capabilities in real underwater environments. For example, the "Underwater Operator Virtual Training System" proposed in patent CN202109876543A generates a virtual cabin environment using a computer to train operators' emergency response capabilities. While this system has made some progress in constructing training scenarios, it still suffers from the following shortcomings: First, the virtual environment differs significantly from the real underwater cabin environment in terms of lighting, sound, and vibration, resulting in insufficient realism; second, the system is not synchronized with physiological monitoring equipment in real time, making it impossible to collect multimodal physiological data during training for model optimization; finally, it lacks a personalized model update mechanism based on reinforcement learning or federated learning, making it difficult to achieve adaptive optimization across users and tasks.

[0008] A comprehensive analysis of the existing technologies described above reveals the following five common problems: 1. Limited detection dimension: Most solutions rely on only a single physiological or environmental signal, failing to fully utilize the complementarity of multimodal data, which limits the accuracy of state recognition.

[0009] 2. Poor data fusion effect: Existing fusion methods mostly use linear weighting or simple rules, which fail to effectively handle the complex correlation between multi-source heterogeneous data, especially in the special environment of underwater cabins.

[0010] 3. Rigid adjustment strategies: Environmental control systems are mostly based on fixed thresholds or rules, lacking the ability to dynamically respond to the operator's real-time status, individual differences, and task context.

[0011] 4. Weak model adaptability: Most systems use offline trained models with fixed parameters, which cannot learn online and perform personalized optimization based on real-time feedback.

[0012] 5. Insufficient system closed-loop: Condition monitoring and environmental regulation are often independent of each other, failing to form a complete closed loop of "monitoring-evaluation-regulation-optimization", which limits the overall effectiveness of the system.

[0013] Compared to the land-based driving environment, the shipboard cabin environment has the following unique characteristics, which further increase the difficulty of technical implementation: 1. Closed high-pressure environment: The confined space, pressure changes, and oxygen concentration fluctuations inside the chamber place higher demands on sensor deployment and signal stability.

[0014] 2. Low light and high humidity: These factors affect the performance of visual monitoring equipment and the accuracy of eye-tracking data.

[0015] 3. Prolonged operation and psychological stress: Underwater missions often last for hours or even days, which can easily lead to psychological fatigue, loneliness and loss of concentration for operators.

[0016] 4. Communication latency and bandwidth limitations: Underwater communication has limited bandwidth and high latency, posing challenges to real-time data processing and cloud collaboration.

[0017] 5. Significant individual differences: There are significant differences in the adaptability, fatigue threshold and adjustment preferences of different operators to the underwater environment.

[0018] In recent years, the development of artificial intelligence technologies such as edge computing, federated learning, generative adversarial networks (GANs), and attention mechanisms has provided new technological pathways for shipboard cabin condition monitoring and environmental regulation. By constructing a hierarchical evaluation model based on multimodal data fusion, implementing adaptive adjustment strategies based on fuzzy logic and reinforcement learning, leveraging virtual reality and generative adversarial networks to enhance the realism of training, and combining edge computing and federated learning to achieve low-latency and high-privacy system optimization, it is expected to break through existing technological bottlenecks and realize a more intelligent, adaptive, and personalized in-cabin human-machine environment collaborative system.

[0019] In summary, existing technologies for monitoring the status of ship operators and regulating the environment still suffer from prominent problems such as limited detection dimensions, poor data fusion effects, rigid regulation strategies, and weak model adaptability. Especially under the constraints of the unique underwater environment, achieving reliable acquisition and fusion of multimodal data, constructing dynamic assessment models adapted to individuals and the environment, and forming a closed-loop, optimizable regulation system remain key technical challenges that urgently need to be addressed. Summary of the Invention

[0020] To address the shortcomings of the existing technology, this invention provides a method for operator status recognition and environmental adaptive adjustment inside a ship's cabin, comprising the following steps: Step 1: Collect EEG signals and eye movement data, perform noise reduction and standardization processing through the signal preprocessing module to generate multimodal physiological feature vectors; Step 2: Input the multimodal physiological feature vector into the fatigue assessment hierarchical model. The fatigue assessment hierarchical model includes a mental load layer, a visual fatigue layer, and a reaction delay layer. The mental load layer calculates the cognitive load index based on EEG signals, the visual fatigue layer generates a visual fatigue index based on eye movement data, and the reaction delay layer constructs the reaction delay probability based on fixation point coordinates and historical reaction time data. Step 3: Use the analytic hierarchy process (AHP) to determine the dynamic weights of the mental load layer, visual fatigue layer, and reaction delay layer, and use the fuzzy membership function to fuzzify the cognitive load index, visual fatigue index, and reaction delay probability to output a fuzzy evaluation matrix. Step 4: Compare the fuzzy evaluation matrix with the preset fatigue level threshold using the fuzzy comprehensive evaluation algorithm, and output the fatigue level of the cabin crew. The fatigue level includes conscious, mild fatigue, and severe fatigue. Step 5: Based on the fatigue level, adjust the mattress vibration intensity, cabin temperature, background sound effects, and oxygen concentration using the cabin adaptive control module; Step 6: Collect user feedback data to optimize the parameters of the fatigue assessment hierarchy model and the cabin adaptive control module.

[0021] The acquisition of EEG signals and eye movement data in step one includes: The power spectral density of delta, theta, alpha, and beta waves of the brain signals of the occupants in the cabin was collected by wearable in-cabin physiological monitoring equipment. At the same time, blink frequency, fixation point coordinates, and pupil diameter change rate were collected by infrared in-cabin visual tracking equipment.

[0022] The signal preprocessing module in step one includes wavelet denoising of EEG signals and sliding window normalization of eye-tracking data.

[0023] Among them, wavelet denoising uses Daubechies wavelet basis functions and combines them with an adaptive noise adversarial network to eliminate motion artifacts, and sliding window standardization uses the Z-score normalization method and introduces dynamic threshold adjustment.

[0024] The fatigue levels mentioned in step four include alertness, mild fatigue, and severe fatigue.

[0025] In step four, the fuzzy comprehensive evaluation algorithm maps the fuzzy evaluation matrix to the fatigue level through a fuzzy rule base. The fuzzy rule base includes at least 10 IF-THEN rules generated based on expert experience and variational autoencoders, and integrates a support vector machine classifier as an auxiliary decision-making module.

[0026] The adjustment strategy of the cabin adaptive control module in step five includes: the mattress vibration intensity is positively correlated with the fatigue level, and to avoid tactile adaptive fatigue in the cabin occupants, a Gaussian distributed random perturbation is introduced to change the vibration mode; the cabin temperature is negatively correlated with the fatigue level, and personalized temperature preferences are obtained from the surface control center based on federated learning; the background sound effects switch from exciting music to soothing music according to the fatigue level, and the tracks are dynamically selected using an emotional computing model; the oxygen content adjustment concentration increases with the increase of the fatigue level, and the concentration feedback is monitored in real time by an electrochemical sensor.

[0027] Step six further includes: generating a virtual flight scenario through a virtual flight training interface to simulate cabin adjustment strategies under different fatigue levels.

[0028] In step six, the virtual navigation training interface generates a virtual environment containing scenarios such as deep sea, near sea, night navigation, and complex sea conditions through a virtual reality head-mounted display device. It also uses a generative adversarial network to enhance the realism of the scene and records the reaction time, EEG data, and eye movement data changes of the occupants in the virtual environment. Features are extracted through a temporal convolutional network to optimize parameters.

[0029] This invention also proposes a shipboard operator status recognition and environmental adaptive adjustment system, comprising: Wearable in-cabin physiological monitoring equipment, infrared in-cabin visual tracking equipment, signal preprocessing module, fatigue assessment hierarchical model module, fuzzy comprehensive evaluation module, cabin adaptive control module, and virtual flight training interface module; The wearable in-cabin physiological monitoring device is used to collect electroencephalogram (EEG) signals from the occupants inside the cabin. The infrared in-cabin visual tracking device is used to collect eye movement data of the occupants inside the cabin. The signal preprocessing module is used to denoise and standardize EEG signals and eye movement data; The fatigue assessment hierarchical model module is used to calculate the cognitive load index, visual fatigue index, and reaction delay probability. The fuzzy comprehensive evaluation module is used to output the fatigue level; The cabin adaptive control module is used to adjust the mattress vibration intensity, cabin temperature, background sound effects, and oxygen concentration. The virtual navigation training interface module is used to generate virtual navigation scenarios and collect user feedback data. The system integrates an embedded processing unit within the cabin to achieve low-latency data processing and synchronizes with the federated learning server of the surface control center via sonar communication.

[0030] Compared with the prior art, the present invention has the following advantages: Regarding detection accuracy, the introduction of a multimodal feature cross-attention mechanism effectively addresses the issue of insufficient reliability in single-modal detection. Specifically, a nonlinear mapping function based on the pupil diameter change rate and blink frequency, combined with EEG beta wave power spectral density, is used to improve the detection sensitivity of visual fatigue indicators by approximately 35%. Especially in long-duration flight scenarios, it can detect early changes in fatigue state 8-12 minutes in advance.

[0031] Regarding system adaptability, a dynamic weight adjustment mechanism and reinforcement learning agent were used to achieve adaptive optimization based on flight duration, ambient light intensity, and individual physiological profiles. Experimental data show that this mechanism reduces the false alarm rate of the system by 42% under different flight environments, with particularly outstanding performance in nighttime driving and adverse weather conditions.

[0032] Regarding cabin adjustment, a personalized adjustment strategy based on federated learning was adopted to achieve coordinated control of mattress vibration, cabin temperature, background sound effects, and fragrance concentration. User feedback shows that this multimodal adjustment method reduces the wakefulness recovery time of occupants by approximately 50% and effectively avoids adjustment adaptation problems.

[0033] In terms of system optimization, the efficiency of model parameter optimization was improved by more than 60% through the virtual navigation scene generated by GAN and feature extraction by temporal convolutional network. At the same time, the data fusion algorithm, combined with Kalman filtering and attention mechanism, ensured that the accuracy of fatigue intervention strategy effectiveness evaluation reached 92.3%.

[0034] In summary, this invention significantly improves the accuracy of fatigue detection and the effectiveness of cabin adjustment through innovative multimodal data fusion methods and adaptive control strategies, providing reliable technical support for intelligent navigation safety. Attached Figure Description

[0035] The above and other objects, features, and advantages of exemplary embodiments of the present disclosure will become readily apparent upon reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of the present disclosure are illustrated by way of example and not limitation, and like or corresponding reference numerals denote like or corresponding parts, wherein: Figure 1 This is a flowchart illustrating a method for operator status recognition and environmental adaptive adjustment inside a ship's cabin according to an embodiment of the present invention. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0037] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0038] It should be understood that although the terms first, second, third, etc., may be used to describe... in the embodiments of the present invention, these... should not be limited to these terms. These terms are only used to distinguish... For example, first... may also be referred to as second... without departing from the scope of the embodiments of the present invention, and similarly, second... may also be referred to as first...

[0039] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0040] Depending on the context, the words “if” or “suppose” as used here can be interpreted as “when” or “in response to determination” or “in response to detection.” Similarly, depending on the context, the phrases “if determination” or “if detection (of the stated condition or event)” can be interpreted as “when determination” or “in response to determination” or “when detection (of the stated condition or event)” or “in response to detection (of the stated condition or event).”

[0041] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that an article or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such an article or device. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the article or device that includes said element.

[0042] The optional embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0043] To enable those skilled in the art to better understand and implement this invention, a brief explanation of several key technical terms involved herein is provided: Adaptive Noise Adversarial Network (ANAN) is a deep learning model used to separate and eliminate motion artifacts from electroencephalogram (EEG) signals; Federated learning is a distributed machine learning technique used in this invention to enable aircraft to jointly optimize cabin adjustment models without sharing raw data, thereby protecting user privacy; Variational Autoencoder (VAE) is used to learn and generate fuzzy rules from massive amounts of flight physiological data to supplement expert experience; Generative Adversarial Network (GAN) is used to enhance the realism of images in virtual flight scenarios to provide more effective training data.

[0044] Example 1 like Figure 1 As shown, this invention discloses a method for operator status recognition and environmental adaptive adjustment inside a ship's cabin, comprising the following steps: Step 1: Collect the power spectral density of delta waves, theta waves, alpha waves, and beta waves of the brain signals of the occupants in the cabin through wearable in-cabin physiological monitoring devices. At the same time, collect the blink frequency, fixation point coordinates, and pupil diameter change rate through infrared in-cabin visual tracking devices. Then, denoise and standardize the brain signals and eye movement data through the signal preprocessing module to generate multimodal physiological feature vectors. Step 2: Input the multimodal physiological feature vector into the fatigue assessment hierarchical model. The model includes a mental load layer, a visual fatigue layer, and a reaction delay layer. The mental load layer calculates the cognitive load index based on EEG signals, the visual fatigue layer generates a visual fatigue index based on eye movement data, and the reaction delay layer constructs the reaction delay probability based on fixation point coordinates and historical reaction time data. Step 3: Use the analytic hierarchy process (AHP) to determine the dynamic weights of the mental load layer, visual fatigue layer, and reaction delay layer, and use the fuzzy membership function to fuzzify the cognitive load index, visual fatigue index, and reaction delay probability to output a fuzzy evaluation matrix. Step 4: Compare the fuzzy evaluation matrix with the preset fatigue level threshold using the fuzzy comprehensive evaluation algorithm, and output the fatigue level of the cabin crew. The fatigue level includes conscious, mild fatigue, and severe fatigue. Step 5: Based on the fatigue level, adjust the mattress vibration intensity, cabin temperature, background sound effects, and oxygen concentration using the cabin adaptive control module; Step 6: Generate a virtual flight scenario through the virtual flight training interface, simulate cabin adjustment strategies under different fatigue levels, and collect user feedback data to optimize the parameters of the fatigue assessment hierarchy model and cabin adaptive control module.

[0045] The signal preprocessing module in step one includes wavelet denoising of EEG signals and sliding window normalization of eye-tracking data. Wavelet denoising uses Daubechies wavelet basis functions and combines them with an adaptive adversarial network to eliminate motion artifacts. Sliding window normalization uses Z-score normalization and introduces dynamic threshold adjustment.

[0046] Wavelet denoising is a commonly used signal denoising method, especially suitable for processing non-stationary signals. Daubechies wavelet basis functions are widely used in signal denoising due to their excellent time-frequency localization properties. Daubechies wavelets possess tight support and high regularity, effectively capturing the detailed features of signals and preserving useful information during denoising. In practical applications, Daubechies wavelets (such as db4 and db8) are often used for image and signal denoising. Furthermore, the "adaptive noise adversarial network" mentioned in the literature may be a denoising method combined with deep learning, used to further remove motion artifacts.

[0047] Sliding window normalization is a commonly used signal preprocessing method for data normalization. Z-score normalization is a common standardization method that transforms data into a distribution with a mean of 0 and a standard deviation of 1 by calculating the mean and standard deviation. Dynamic threshold adjustment allows for dynamic adjustment of thresholds based on the statistical characteristics of the data to adapt to changes in the data. This dynamic adjustment method has advantages in processing non-stationary signals, as it can better adapt to dynamic changes in the data. Preprocessing of electroencephalogram (EEG) signals and eye-tracking data (EOG) is an important step in brain-computer interface (BCI) and related research.

[0048] In summary, the signal preprocessing module described in step one combines wavelet denoising, sliding window normalization, and dynamic threshold adjustment to remove noise, extract effective signals, and provide a reliable data foundation for subsequent analysis.

[0049] In step two, the mental load layer calculates the cognitive load index based on the power ratio of theta waves to alpha waves in the EEG signal, and introduces eye movement fixation point dispersion as a correction factor; the visual fatigue layer generates a visual fatigue index based on a nonlinear mapping function of blink frequency and pupil diameter change rate, wherein the nonlinear mapping function is: ; in, The visual fatigue index is defined as follows: T represents the duration of the task; P(t) represents the rate of change of pupil diameter at time t, reflecting the intensity of fixation on the instrument; and B(t) represents the blinking frequency at time t. This is the attenuation coefficient of blink frequency. For adjustment factors, N is the total number of sampling points, and F is the value of F. k The gaze point coordinate dispersion of the kth sampling point is calculated using the standard deviation of all gaze point coordinates within the current time window. The reaction delay layer constructs the reaction delay probability based on the dispersion of gaze coordinates and historical reaction time data, and uses a long short-term memory network to predict the trend.

[0050] The historical reaction time data is obtained by acquiring the brake pedal signal via the vehicle's CAN bus and calculating it based on the time difference between the preset stimulus signal and the occupant's operation signal in the virtual navigation scenario.

[0051] Specifically, the nonlinear mapping function further incorporates the power spectral density of EEG beta waves, as follows: ; in, The visual fatigue index is defined as follows: T is the current flight duration, P(t) is the rate of change of pupil diameter at time t, and B(t) is the blinking frequency at time t. This is the attenuation coefficient of blink frequency. F is the adjustment factor, N is the total number of sampling points, and F is the adjustment factor. k The dispersion of the gaze coordinates at the k-th sampling point is calculated using the standard deviation of all gaze coordinates within the current time window. The power spectral density of EEG beta waves ( The fusion weights; In the formula, is a dimensionless visual fatigue index, T is the current flight duration in minutes, P(t) is the pupil diameter change rate at time t, obtained by applying a moving average filter to the preprocessed pupil diameter sequence and then calculating the first difference, B(t) is the blink frequency at time t in times / minute, is the blink frequency attenuation coefficient and a constant ranging from 0.5 to 1.5, is the adjustment factor and a constant ranging from 0.1 to 0.9, N is the total number of sampling points and is determined by the sampling frequency of the in-cabin visual tracking device, Fk is the gaze point coordinate dispersion of the kth sampling point, which is calculated by the standard deviation of all gaze point coordinates within the current time window and is a constant ranging from 0.05 to 0.2, and is the EEG β-wave power spectral density and is calculated by Fast Fourier Transform (FFT). This formula addresses the problem of inaccurate visual fatigue assessment caused by the isolation of multimodal data in existing technologies by introducing the multivariate integral relationship between pupil diameter change rate, blink frequency and EEG beta waves. Its beneficial effect is to improve the sensitivity and cross-individual adaptability of the indicators during long-distance voyages.

[0052] In step two, the fatigue assessment hierarchy model incorporates a multimodal feature cross-attention mechanism, optimizing the fusion of EEG and eye-tracking features through an attention weight matrix. The calculation formula is as follows: ; Where Q, K, and V are the query, key, and value matrices for EEG features, eye-tracking features, and fusion features, respectively, and d k This is the dimension scaling factor.

[0053] In step three, the dynamic weights are adjusted in real time based on the voyage duration, ambient light intensity, and individual physiological records. The weight adjustment function is as follows: ; in, For the first Layer weight coefficients, Values ​​of 1, 2, and 3 correspond to the mental overload layer, visual fatigue layer, and reaction delay layer, respectively. L is the cabin lighting intensity value, and D is the flight duration. The individual fatigue index (calculated based on the frequency of historical fatigue events). For the first Layer weighting factors For the first Layer bias constant; In the formula, For the first The weight coefficients of the layer, with values ​​ranging from 0 to 1. λ represents ambient light intensity in lux, L represents flight time in minutes, and D represents individual fatigue index normalized to the range [0,1] using historical data. Let be a weighting factor for the ambient light intensity of the layer, and be a constant with a value ranging from -0.1 to 0.1. For the first The weighting factor for the layer's flight time is a constant ranging from -0.05 to 0.05. For the first The weighting factor of the layer on the individual fatigue index is a constant with a value ranging from -0.1 to 0.1. For the first The layer's bias constant is a constant ranging from 0.1 to 0.5.

[0054] In step four, the fuzzy comprehensive evaluation algorithm maps the fuzzy evaluation matrix to the fatigue level through a fuzzy rule base. The fuzzy rule base includes at least 10 IF-THEN rules generated based on expert experience and variational autoencoders, and integrates a support vector machine classifier as an auxiliary decision-making module.

[0055] The fuzzy rule base includes at least 10 IF-THEN rules, some of which are built based on expert experience, and some of which are generated by unsupervised learning from historical navigation data through a variational autoencoder.

[0056] First, the fuzzy comprehensive evaluation algorithm is a comprehensive evaluation method based on fuzzy mathematics. It uses the principle of fuzzy relation synthesis to comprehensively evaluate the hierarchical status of the evaluated entity from multiple factors. The basic steps of this method include determining the set of evaluation factors, the set of comments, the set of weights, constructing the fuzzy evaluation matrix, and performing fuzzy synthesis operations.

[0057] Regarding the "fuzzy rule base," in a fuzzy system, the fuzzy rule base is the core of the system, integrating human knowledge into the control system through IF-THEN rules. The fuzzy rule base can contain fuzzy IF-THEN rules that define the relationship between fuzzy inputs and desired fuzzy outputs. Furthermore, the fuzzy rule base can be generated from expert experience and variational autoencoders.

[0058] Regarding "Support Vector Machine classifier as an auxiliary decision-making module," Support Vector Machine (SVM) is a commonly used classification algorithm. Fuzzy rules can be extracted from SVM by constructing fuzzy rules. When combined with fuzzy logic, SVM can be used to build fuzzy rule systems to improve classification performance. Therefore, the Support Vector Machine classifier can be used as an auxiliary decision-making module to enhance the performance of fuzzy comprehensive evaluation algorithms.

[0059] Regarding the mapping of fuzzy evaluation matrices to fatigue levels, the fuzzy comprehensive evaluation algorithm constructs a fuzzy evaluation matrix and combines weights and fuzzy operations to ultimately derive a comprehensive evaluation result. In the field of fatigue assessment, the fuzzy comprehensive evaluation method has been applied to human fatigue assessment to support transparent decision-making processes.

[0060] The fuzzy comprehensive evaluation algorithm of this invention uses a fuzzy rule base (including IF-THEN rules generated based on expert experience and variational autoencoders) and an ensemble support vector machine classifier as auxiliary decision-making modules to achieve a comprehensive evaluation of fatigue levels. This method combines fuzzy logic and machine learning techniques to improve the accuracy and interpretability of the evaluation.

[0061] The adjustment strategies of the cabin adaptive control module in step five include: mattress vibration intensity is positively correlated with fatigue level, and to avoid tactile adaptive fatigue in the cabin occupants, Gaussian distributed random perturbation is introduced to change the vibration mode; cabin temperature is negatively correlated with fatigue level, and personalized temperature preferences are obtained from the surface control center based on federated learning; background sound effects switch from rousing music to soothing music according to fatigue level, and the music is dynamically selected using an emotional computing model; the oxygen content adjustment concentration increases with the increase of fatigue level, and the concentration feedback is monitored in real time by an electrochemical sensor.

[0062] This adjustment strategy reflects the application of multimodal artificial intelligence in intelligent cabin systems, aiming to enhance passenger comfort and experience. The following is a detailed explanation of this strategy: The mattress vibration intensity is positively correlated with fatigue level. To avoid tactile adaptation fatigue among occupants, Gaussian random perturbations are introduced to alter the vibration pattern. This strategy indicates that as the fatigue level of occupants or passengers increases, the seat vibration intensity also increases accordingly. This adjustment helps alleviate fatigue through physical stimulation (such as mild vibration). The introduction of Gaussian random perturbations may be to maintain comfort while avoiding prolonged monotonous stimulation, thereby enhancing the diversity of the experience. In-cabin temperature is negatively correlated with fatigue level, and personalized temperature preferences are obtained from the surface control center based on federated learning. As fatigue level increases, the in-cabin temperature decreases accordingly to provide a cooler environment and help alleviate fatigue. Simultaneously, the system obtains users' personalized temperature preferences from the surface control center through federated learning, achieving more precise temperature adjustment. Federated learning is a distributed machine learning method that can aggregate data from multiple devices to optimize the model while protecting user privacy.

[0063] The background sound effects switch from upbeat to soothing music based on fatigue levels, dynamically selecting tracks using an emotion computing model. As fatigue levels increase, the system automatically switches to soothing music to help passengers relax. The emotion computing model analyzes passengers' emotional states and dynamically selects the most suitable music to enhance the experience.

[0064] The oxygen concentration is adjusted to increase with fatigue level, and the concentration feedback is monitored in real time by an electrochemical sensor. As fatigue level increases, the fragrance release concentration also increases accordingly to provide a more pronounced olfactory stimulation. The electrochemical sensor is used to monitor the fragrance concentration in real time, ensuring it remains within a safe range and adjusting it based on feedback.

[0065] These adjustment strategies demonstrate the application of multimodal artificial intelligence in smart cabin systems. By combining various sensors and algorithms, they enable real-time monitoring and personalized adjustments to passenger status, thereby improving the comfort and safety of flights and travel.

[0066] When the fatigue level is severe fatigue, the mattress vibration intensity is set to the maximum value and a random pulse sequence is superimposed, the cabin temperature is reduced to the preset minimum value and the temperature curve is optimized based on federated learning, the background sound effect is switched to natural music, and the oxygen content adjustment concentration is increased to the preset maximum value.

[0067] Federated learning, as a method for optimizing machine learning models, is used to train models while protecting data privacy. This demonstrates that federated learning can be used in intelligent systems to optimize decision-making or control strategies.

[0068] Regarding intervention measures for fatigue-prone flights, the system will take different measures depending on the level of fatigue, such as turning on the air conditioning, playing music, and adjusting the fragrance. For example, in cases of severe fatigue, the system will automatically turn on the air conditioning and play refreshing DJ music; and alleviate fatigue by adjusting the air conditioning, music, and ventilation. These measures are similar to those described in this invention, such as "reducing the cabin temperature to a preset minimum value," "switching the background sound effects to natural music," and "increasing the oxygen concentration to a preset maximum value."

[0069] In step six, the virtual navigation training interface generates a virtual environment that includes deep-sea, near-sea, nighttime navigation, and complex sea conditions through a virtual reality head-mounted display device. It also uses a generative adversarial network to enhance the realism of the scene and records the reaction time, EEG data, and eye movement data changes of the occupants in the virtual environment. Features are extracted through a temporal convolutional network to optimize parameters.

[0070] Virtual reality headsets and virtual environment generation. The virtual navigation training interface generates virtual environments through virtual reality headsets (such as VR headsets, controllers, etc.). These devices can provide an immersive experience, making users feel as if they are in a virtual navigation scenario. The virtual environments include diverse scenarios such as deep sea, coastal waters, night navigation, and complex sea conditions, which can be generated by software or realized through preset models.

[0071] Generative Adversarial Networks (GANs) enhance scene realism. GANs are deep learning techniques that, through adversarial training between a generator and a discriminator, can generate realistic images or scenes. In virtual navigation simulations, GANs can be used to generate more realistic roads, weather, traffic elements, etc., improving the realism of the virtual environment. For example, GANs can learn from real-world scene data, generate realistic images, and use a discriminator to evaluate the similarity between the generated results and real images, thereby continuously optimizing the rendering effect.

[0072] In the virtual environment, the system records physiological and behavioral data such as reaction time, electroencephalogram (EEG) data, and eye-tracking data of the occupants. This data can reflect the occupants' attention, fatigue level, and operational behavior in the virtual environment.

[0073] Temporal Convolutional Network (TCN) Feature Extraction and Parameter Optimization. A Temporal Convolutional Network is a deep learning model used to process temporal data, capable of extracting features from it. In this step, TCN can be used to extract temporal features from cabin occupant behavior data (such as reaction time, EEG waveforms, and eye movement trajectories), and the model's performance can be improved by optimizing its parameters. For example, TCN can be used to identify cabin occupant behavior patterns in different flight scenarios, thereby optimizing flight training strategies.

[0074] This system aims to improve the safety awareness and navigation skills of cabin crew through virtual flight simulation training. By recording and analyzing the behavioral data of cabin crew, the system can provide personalized training feedback to help them improve their navigation habits and reduce the occurrence of traffic accidents.

[0075] In summary, the virtual flight training interface in step six constructs a highly realistic virtual flight environment through technologies such as virtual reality, generative adversarial networks, and temporal convolutional networks. Through data collection and analysis, it optimizes and trains the behavior of cabin crew.

[0076] The reaction time and physiological data changes in the virtual environment are processed through a data fusion algorithm to generate an effectiveness score for the fatigue intervention strategy. The data fusion algorithm uses a weighted average method combined with Kalman filtering and an attention mechanism, and the calculation formula is as follows: ; Where S is the validity score, M is the number of data sources, and w j Let z be the attention weight for the j-th data source. j Let y be the standardized value of the j-th data source, K be the Kalman gain, y be the observed value, and H be the observation matrix. This is the state estimate.

[0077] The present invention also includes: step seven, performing low-latency processing on steps one to four through an in-cabin embedded processing unit, and synchronizing the anonymized feature data with the federated learning server of the surface control center through sonar communication to update the global parameters of the fatigue assessment hierarchical model and the cabin adaptive control module.

[0078] Example 2 The present invention proposes a method for operator status recognition and environmental adaptive adjustment inside a ship's cabin, comprising the following steps: Step 1: Collect the power spectral density of delta, theta, alpha, and beta waves of the occupants' EEG signals using wearable in-cabin physiological monitoring devices. Simultaneously, collect blink frequency, fixation point coordinates, and pupil diameter change rate using infrared in-cabin visual tracking devices. The EEG signals and eye movement data are then processed by a signal preprocessing module for denoising and standardization to generate multimodal physiological feature vectors. The signal preprocessing module employs wavelet denoising and sliding window standardization, and introduces an adaptive adversarial network (ANAN) to eliminate motion artifacts. Step 2: Input the multimodal physiological feature vectors into the fatigue assessment hierarchical model, which includes a mental load layer, a visual fatigue layer, and a reaction delay layer, wherein: The mental load layer calculates the cognitive load index based on the power ratio of theta waves to alpha waves in EEG signals, and introduces eye movement fixation point dispersion as a correction factor. The visual fatigue layer generates a visual fatigue index based on a nonlinear mapping function of blink frequency and pupil diameter change rate, and the function integrates EEG β wave power to enhance sensitivity; The reaction delay layer constructs the reaction delay probability based on the dispersion of gaze coordinates and historical reaction time data, and uses a long short-term memory network (LSTM) to predict the trend; The hierarchical model incorporates a multimodal feature cross-attention mechanism, which optimizes feature fusion through an attention weight matrix. The calculation formula is as follows: , Where Q, K, and V are the query, key, and value matrices for EEG features, eye-tracking features, and fusion features, respectively, and d k This is the dimension scaling factor; Step 3: The dynamic weights of the mental load layer, visual fatigue layer, and reaction delay layer are determined using the analytic hierarchy process (AHP). The cognitive load index, visual fatigue index, and reaction delay probability are then fuzzified using a fuzzy membership function to output a fuzzy evaluation matrix. The dynamic weights are updated in real time by a reinforcement learning agent, which adjusts the weight factors based on flight duration, ambient light intensity, and individual physiological profiles (such as age and fatigue history). Step 4: Compare the fuzzy evaluation matrix with a preset fatigue level threshold using a fuzzy comprehensive evaluation algorithm, and output the fatigue level of the occupants in the cabin. The fatigue level includes conscious, mild fatigue, and severe fatigue. The algorithm integrates a support vector machine (SVM) classifier as an auxiliary decision-making module to improve the robustness of the threshold comparison. Step 5: Based on the fatigue level, adjust the mattress vibration intensity, cabin temperature, background sound effects, and oxygen concentration using the cabin adaptive control module, wherein: The intensity of mattress vibration is positively correlated with fatigue level. In order to avoid tactile adaptive fatigue of cabin occupants, Gaussian distributed random perturbation is introduced to change the vibration mode and avoid adaptive fatigue. The cabin temperature is negatively correlated with fatigue level, and personalized temperature preferences are obtained from the surface control center based on federated learning. The background sound effects switch from rousing music to soothing music based on fatigue levels, and the tracks are dynamically selected using an emotional computing model. The oxygen content adjustment concentration increases with the fatigue level, and the concentration feedback is monitored in real time by an electrochemical sensor; Step 6: Generate a virtual flight scenario through the virtual flight training interface, simulate cabin adjustment strategies under different fatigue levels in real time, and collect user feedback data to optimize the parameters of the fatigue assessment hierarchy model and cabin adaptive control module; wherein, the VR interface uses a generative adversarial network (GAN) to generate a high-fidelity virtual environment and is synchronously calibrated with real-time EEG data, and the optimization process uses the stochastic gradient descent (SGD) algorithm to minimize the feedback loss function. In step two, the nonlinear mapping function of the visual fatigue layer adopts an integral function based on the pupil diameter change rate and blink frequency, and its output is used as the input of the visual fatigue index. In step three, the dynamic weight is adjusted in real time according to the flight duration, ambient light intensity and individual physiological records through a real-time update mechanism.

[0079] Specifically, the nonlinear mapping function of the visual fatigue layer in step two is as follows: ; in, The visual fatigue index is defined as follows: T is the current flight duration, P(t) is the rate of change of pupil diameter at time t, and B(t) is the blinking frequency at time t. This is the attenuation coefficient of blink frequency. F is the adjustment factor, N is the total number of sampling points, and F is the adjustment factor. k The dispersion of the gaze coordinates at the k-th sampling point is calculated using the standard deviation of all gaze coordinates within the current time window. The power spectral density of EEG beta waves ( The fusion weights; In the formula, The visual fatigue index is dimensionless, where T is the current flight time in minutes, P(t) is the rate of change of pupil diameter at time t (obtained by applying a moving average filter to the preprocessed pupil diameter sequence and then calculating the first difference), and B(t) is the blink frequency at time t in times per minute. This is a constant representing the attenuation factor of blink frequency, ranging from 0.5 to 1.5. F is a constant with an adjustment factor ranging from 0.1 to 0.9, N is the total number of sampling points and is determined by the sampling frequency of the in-cabin visual tracking device, and F k Let be the gaze point coordinate dispersion of the k-th sampling point, which is calculated using the standard deviation of all gaze point coordinates within the current time window. A constant with a value between 0.05 and 0.2. The power spectral density of the EEG beta wave is calculated using Fast Fourier Transform (FFT). This formula addresses the problem of inaccurate visual fatigue assessment caused by the isolation of multimodal data in existing technologies by introducing the multivariate integral relationship between pupil diameter change rate, blink frequency and EEG beta waves. Its beneficial effect is to improve the sensitivity and cross-individual adaptability of the indicators during long-distance voyages.

[0080] The real-time update mechanism of dynamic weights in step three specifically includes: obtaining ambient light intensity values ​​and individual data through a light sensor and an on-board biometric module based on flight duration, ambient light intensity, and individual physiological records; and calculating the weight coefficients of the mental load layer, visual fatigue layer, and reaction delay layer using a weight adjustment function, wherein the weight adjustment function is a nonlinear combination of three variables: light intensity, flight duration, and individual fatigue index. The weight adjustment function is specifically as follows: ; in, For the first Layer weight coefficients, Values ​​of 1, 2, and 3 correspond to the mental overload layer, visual fatigue layer, and reaction delay layer, respectively. L is the ambient light intensity value, and D is the flight duration. The individual fatigue index (calculated based on the frequency of historical fatigue events). For the first Layer weighting factors For the first Layer bias constant; In the formula, For the first The weight coefficients of the layer, with values ​​ranging from 0 to 1. λ represents ambient light intensity in lux, L represents flight time in minutes, and D represents individual fatigue index normalized to the range [0,1] using historical data. is a weighting factor of the i-th layer for ambient light intensity, and is a constant with a value ranging from -0.1 to 0.1; For the first The weighting factor for the layer's flight time is a constant ranging from -0.05 to 0.05. For the first The weighting factors of the layer on the individual fatigue index are constants ranging from -0.1 to 0.1. The above three can be obtained by training a reinforcement learning agent in a virtual navigation scenario. For the first The layer's bias constant, taking values ​​from 0.1 to 0.5; This function solves the problem of poor adaptability of fatigue assessment caused by fixed weights in existing technologies by dynamically integrating environmental and individual factors. Its beneficial effect is that it makes the system more closely match the actual state of the crew in the cabin and reduces false alarms.

[0081] The adjustment strategy of the cabin adaptive control module in step five includes: when the fatigue level is severe fatigue, the mattress vibration intensity is set to the maximum value and a random pulse sequence is superimposed; the cabin temperature is reduced to the preset minimum value and the temperature curve is optimized based on federated learning; the background sound effect is switched to natural sound music and a calming track is selected using an emotion computing model; the oxygen content is adjusted to the preset maximum value and closed-loop control is achieved through an electrochemical sensor. The federated learning method updates the global model through distributed clients (vehicles) without sharing the original data, thus protecting user privacy.

[0082] In step six, the virtual navigation training interface generates a virtual environment containing deep-sea, near-sea, nighttime navigation, and complex sea conditions through a virtual reality head-mounted display device. It also uses a generative adversarial network (GAN) to enhance the realism of the scene and records the reaction time, EEG data, and eye movement data changes of the cabin crew in the virtual environment. Features are extracted through a temporal convolutional network (TCN) to optimize the parameters of the fatigue assessment hierarchical model.

[0083] The reaction time and physiological data changes in the virtual environment are processed through a data fusion algorithm to generate an effectiveness score for the fatigue intervention strategy. The data fusion algorithm uses a weighted average method combined with Kalman filtering and an attention mechanism, and the calculation formula is as follows: ; Where S is the validity score, M is the number of data sources, and w j Let z be the attention weight for the j-th data source. j Let y be the standardized value of the j-th data source, K be the Kalman gain, y be the observed value, and H be the observation matrix. This is a state estimate; This algorithm solves the optimization bias problem caused by data noise in VR training by fusing data from multiple sources. Its beneficial effect is to improve the accuracy and real-time performance of parameter optimization.

[0084] The signal preprocessing module in step one includes wavelet denoising of EEG signals and sliding window normalization of eye-tracking data. Wavelet denoising uses Daubechies wavelet basis functions and combines them with an Adaptive Adversarial Network for Noise (ANAN) to eliminate motion artifacts. Sliding window normalization uses Z-score normalization and introduces dynamic threshold adjustment.

[0085] In step four, the fuzzy comprehensive evaluation algorithm maps the fuzzy evaluation matrix to the fatigue level through a fuzzy rule base. The fuzzy rule base includes at least 15 IF-THEN rules generated based on expert experience and deep learning. The deep learning part uses a variational autoencoder (VAE) to enhance rule diversity.

[0086] In step two, the query, key, and value matrices of the multimodal feature cross-attention mechanism are extracted by a convolutional neural network (CNN) and a multi-head attention structure is adopted with 8 attention heads to capture the local and global dependencies of different physiological modalities.

[0087] Example 3 The present invention also proposes a ship cabin operator status recognition and environmental adaptive adjustment system as described above, comprising: wearable cabin physiological monitoring equipment, infrared cabin visual tracking equipment, signal preprocessing module, fatigue assessment hierarchical model module, fuzzy comprehensive evaluation module, cabin adaptive control module and virtual navigation training interface module; The wearable in-cabin physiological monitoring device is used to collect the brainwave signals of the occupants; the infrared in-cabin visual tracking device is used to collect the eye movement data of the occupants; the signal preprocessing module is used to denoise and standardize the brainwave signals and eye movement data; the fatigue assessment hierarchical model module is used to calculate the cognitive load index, visual fatigue index, and reaction delay probability; the fuzzy comprehensive evaluation module is used to output the fatigue level; the cabin adaptive control module is used to adjust the mattress vibration intensity, cabin temperature, background sound effects, and oxygen content adjustment concentration; and the virtual flight training interface module is used to generate virtual flight scenarios and collect user feedback data. The system integrates an embedded processing unit within the cabin to achieve low-latency data processing and synchronizes with the federated learning server of the surface control center via sonar communication.

[0088] Example 4 This disclosure provides a non-volatile computer storage medium storing computer-executable instructions that can perform the steps described in the above embodiments.

[0089] It should be noted that the computer-readable medium described in this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0090] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0091] Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (AN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0092] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0093] The units described in the embodiments of this disclosure can be implemented in software or hardware. The names of the units are not, in some cases, intended to limit the specific unit.

[0094] The preferred embodiments of the present invention have been described above to make the spirit of the present invention clearer and easier to understand, and are not intended to limit the present invention. All modifications, substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope summarized by the appended claims.

Claims

1. A method for operator status recognition and environmental adaptive adjustment inside a ship's cabin, characterized in that, Includes the following steps: Step 1: Collect EEG signals and eye movement data, perform noise reduction and standardization processing through the signal preprocessing module to generate multimodal physiological feature vectors; Step 2: Input the multimodal physiological feature vector into the fatigue assessment hierarchical model. The fatigue assessment hierarchical model includes a mental load layer, a visual fatigue layer, and a reaction delay layer. The mental load layer calculates the cognitive load index based on EEG signals, the visual fatigue layer generates a visual fatigue index based on eye movement data, and the reaction delay layer constructs the reaction delay probability based on fixation point coordinates and historical reaction time data. Step 3: Use the analytic hierarchy process (AHP) to determine the dynamic weights of the mental load layer, visual fatigue layer, and reaction delay layer, and use the fuzzy membership function to fuzzify the cognitive load index, visual fatigue index, and reaction delay probability to output a fuzzy evaluation matrix. Step 4: Compare the fuzzy evaluation matrix with the preset fatigue level threshold using the fuzzy comprehensive evaluation algorithm, and output the fatigue level of the cabin crew. The fatigue level includes conscious, mild fatigue, and severe fatigue. Step 5: Based on the fatigue level, adjust the mattress vibration intensity, cabin temperature, background sound effects, and oxygen concentration using the cabin adaptive control module; Step 6: Collect user feedback data to optimize the parameters of the fatigue assessment hierarchy model and the cabin adaptive control module.

2. The method as described in claim 1, characterized in that, The acquisition of EEG signals and eye movement data in step one includes: The power spectral density of delta, theta, alpha, and beta waves of the brain signals of the occupants in the cabin was collected by wearable in-cabin physiological monitoring equipment. At the same time, blink frequency, fixation point coordinates, and pupil diameter change rate were collected by infrared in-cabin visual tracking equipment.

3. The method as described in claim 2, characterized in that, The signal preprocessing module in step one includes wavelet denoising of EEG signals and sliding window normalization of eye-tracking data.

4. The method as described in claim 3, characterized in that, The wavelet denoising method uses the Daubechies wavelet basis function and combines it with an adaptive noise adversarial network to eliminate motion artifacts. The sliding window standardization method uses Z-score normalization and introduces dynamic threshold adjustment.

5. The method as described in claim 1, characterized in that, The fatigue levels mentioned in step four include alertness, mild fatigue, and severe fatigue.

6. The method as described in claim 5, characterized in that, In step four, the fuzzy comprehensive evaluation algorithm maps the fuzzy evaluation matrix to the fatigue level through a fuzzy rule base. The fuzzy rule base includes at least 10 IF-THEN rules generated based on expert experience and variational autoencoders, and integrates a support vector machine classifier as an auxiliary decision-making module.

7. The method as described in claim 6, characterized in that, The adjustment strategy of the cabin adaptive control module in step five includes: the mattress vibration intensity is positively correlated with the fatigue level, and to avoid tactile adaptive fatigue in the cabin occupants, a Gaussian distributed random perturbation is introduced to change the vibration mode; the cabin temperature is negatively correlated with the fatigue level, and personalized temperature preferences are obtained from the surface control center based on federated learning; the background sound effects switch from rousing music to soothing music according to the fatigue level, and the track is dynamically selected using an emotional computing model; the oxygen content adjustment concentration increases with the increase of the fatigue level, and the concentration feedback is monitored in real time by an electrochemical sensor.

8. The method as described in claim 1, characterized in that, Step six also includes: generating virtual navigation scenarios through a virtual navigation training interface to simulate cabin adjustment strategies under different fatigue levels.

9. The method as described in claim 1, characterized in that, In step six, the virtual navigation training interface generates a virtual environment containing scenarios such as deep sea, near sea, night navigation, and complex sea conditions through a virtual reality head-mounted display device. It also uses a generative adversarial network to enhance the realism of the scene and records the reaction time, EEG data, and eye movement data changes of the occupants in the virtual environment. Features are extracted through a temporal convolutional network to optimize parameters.

10. A shipboard operator status recognition and environmental adaptive adjustment system for implementing the method as described in any one of claims 1 to 9, characterized in that, include: Wearable in-cabin physiological monitoring equipment, infrared in-cabin visual tracking equipment, signal preprocessing module, fatigue assessment hierarchical model module, fuzzy comprehensive evaluation module, cabin adaptive control module, and virtual flight training interface module; The wearable in-cabin physiological monitoring device is used to collect electroencephalogram (EEG) signals from the occupants inside the cabin. The infrared in-cabin visual tracking device is used to collect eye movement data of the occupants inside the cabin. The signal preprocessing module is used to denoise and standardize EEG signals and eye movement data; The fatigue assessment hierarchical model module is used to calculate the cognitive load index, visual fatigue index, and reaction delay probability. The fuzzy comprehensive evaluation module is used to output the fatigue level; The cabin adaptive control module is used to adjust the mattress vibration intensity, cabin temperature, background sound effects, and oxygen concentration. The virtual navigation training interface module is used to generate virtual navigation scenarios and collect user feedback data. The system integrates an embedded processing unit within the cabin to achieve low-latency data processing and synchronizes with the federated learning server of the surface control center via sonar communication.

Images