Physiological data acquisition method for air traffic controller

Abstract

The invention discloses a physiological data acquisition method for an air traffic controller, and relates to the technical field of data acquisition, and the method comprises the steps: obtaining the multi-modal physiological data of the air traffic controller at the current moment; inputting the multi-modal physiological data into the trained fatigue-emotion evaluation model to obtain a predicted fatigue score and predicted emotion state probability distribution at the current moment; based on the predicted fatigue score, the predicted emotional state probability distribution and the multi-modal physiological data, processing tasks of different priorities are generated; processing tasks are executed according to priorities through a hierarchical scheduling algorithm, and physiological data collection is completed. Through multi-modal data fusion and model evaluation, the physiological state of the controller is accurately obtained, tasks are processed according to priorities, timely response of key data is guaranteed, reliable physiological information support is provided for air traffic control, and the flight safety guarantee level is improved.

Description

Technical Field

[0001] This application relates to the field of data acquisition technology, and in particular to a method for collecting physiological data for air traffic controllers. Background Technology

[0002] In the aviation industry, flight safety is always of paramount importance. Air traffic controllers, as the core of air traffic management, bear the critical responsibility of directing and coordinating flight activities, ensuring orderly flight operations, and guaranteeing flight safety. During flight, air traffic controllers need to make various decisions quickly and accurately in complex and ever-changing environments, and respond promptly to emergencies. Their decision-making ability and reaction speed directly affect the safety and efficiency of the entire air transport system.

[0003] During flight, air traffic controllers generate a vast amount of diverse physiological data, including a significant amount of both critical and non-critical data. Current technologies treat all data equally, lacking an effective hierarchical mechanism. This results in critical data not receiving priority resources during transmission, computation, and alerting. In the event of network congestion or computational resource constraints, critical physiological alerts may be delayed or even lost, failing to provide timely decision support for control positions and the command center, thus impacting flight safety.

[0004] Therefore, there is an urgent need for a physiological data collection method for air traffic controllers to solve the problems of the inability to classify collected data and the untimely response to critical data, so as to ensure that air traffic controllers are in the best condition at critical moments and to ensure flight safety. Summary of the Invention

[0005] The purpose of this application is to provide a method for collecting physiological data for air traffic controllers, which can solve the shortcomings of the inability to classify the collected data and the inability to guarantee the real-time response of critical data.

[0006] To achieve the above objectives, this application provides the following solution: In a first aspect, this application provides a method for collecting physiological data for air traffic controllers, including: Acquire multimodal physiological data of air traffic controllers at the current moment; the multimodal physiological data includes electrocardiogram and heart rate monitoring data, respiratory monitoring data, electroencephalogram (EEG) activity monitoring data, electrical skin activity monitoring data, blood oxygen saturation data, and behavioral and posture data; The air traffic controller's current multimodal physiological data is input into a pre-trained fatigue-emotion assessment model to obtain the predicted fatigue score and predicted emotional state probability distribution for the current moment. The pre-trained fatigue-emotion assessment model is obtained by iteratively training a preset machine learning network model using a sample dataset, which includes historical multimodal physiological data and corresponding actual fatigue scores and actual emotional state probability distributions. The emotional states include calmness and anxiety. Based on the predicted fatigue score at the current moment, the preset emotional state probability distribution, and multimodal physiological data, processing tasks with different priorities are generated; the priorities include critical level, high priority, routine level, and background level. The processing tasks are executed according to priority using a hierarchical scheduling algorithm to complete the collection of physiological data from air traffic controllers.

[0007] Secondly, this application provides a physiological data acquisition system for air traffic controllers, comprising: The multimodal physiological data acquisition module is used to acquire the air traffic controller's multimodal physiological data at the current moment; the multimodal physiological data includes electrocardiogram and heart rate monitoring data, respiratory monitoring data, electroencephalogram (EEG) activity monitoring data, electrical skin activity monitoring data, blood oxygen saturation data, and behavioral and posture data; The fatigue-emotion assessment module is used to input the air traffic controller's current multimodal physiological data into a pre-trained fatigue-emotion assessment model to obtain the predicted fatigue score and predicted emotional state probability distribution for the current moment. The pre-trained fatigue-emotion assessment model is obtained by iteratively training a preset machine learning network model using a sample dataset. The sample dataset includes historical multimodal physiological data and corresponding actual fatigue scores and actual emotional state probability distributions. The emotional states include calmness and anxiety. The task generation module is used to generate processing tasks of different priorities based on the predicted fatigue score, preset emotional state probability distribution, and multimodal physiological data at the current moment; the priorities include critical level, high priority, normal level, and background level. The task scheduling and execution module is used to execute the processing tasks according to priority through a hierarchical scheduling algorithm to complete the collection of physiological data of air traffic controllers.

[0008] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the physiological data acquisition method for air traffic controllers described in any of the above-described methods.

[0009] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the physiological data acquisition method for air traffic controllers described above.

[0010] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the physiological data acquisition method for air traffic controllers described above.

[0011] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides a method for collecting physiological data for air traffic controllers. By acquiring multimodal physiological data including electrocardiogram, heart rate, and respiration from air traffic controllers at the current moment, it solves the problem that single physiological data collection is incomplete and cannot accurately reflect the controller's overall physiological state, achieving comprehensive and accurate collection of controller physiological information. By inputting the multimodal physiological data into a trained fatigue-emotion assessment model, it obtains predicted fatigue scores and predicted emotional state probability distributions, solving the problem of difficulty in quantifying and assessing the controller's fatigue level and emotional state, achieving a digital and accurate assessment of the controller's physical and mental state. Based on the prediction results and multimodal physiological data, it generates processing tasks with different priorities and executes them according to priority through a hierarchical scheduling algorithm, solving the problem of untimely response to critical data, and achieving efficient and orderly processing of physiological data and real-time response to critical information. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is an application environment diagram of a physiological data collection method for air traffic controllers according to an embodiment of this application.

[0014] Figure 2 This is a flowchart illustrating a method for collecting physiological data from air traffic controllers, provided as an embodiment of this application.

[0015] Figure 3 This is a schematic diagram of the functional modules of a physiological data acquisition system for air traffic controllers, provided as an embodiment of this application.

[0016] Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, this application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] The physiological data collection method for air traffic controllers provided in this application embodiment can be applied to, for example... Figure 1 The application environment shown is illustrated. Terminal 101 communicates with server 102 via a network. A data storage system can store the data that server 102 needs to process. The data storage system can be set up independently, integrated into server 102, or placed in the cloud or on another server. Terminal 101 can send the air traffic controller's current multimodal physiological data to server 102. After receiving the air traffic controller's current multimodal physiological data, server 102 inputs the data into a trained fatigue-emotion assessment model to obtain the predicted fatigue score and predicted emotional state probability distribution for the current moment. The trained fatigue-emotion assessment model is obtained by iteratively training a preset machine learning network model using a sample dataset. The sample dataset includes historical multimodal physiological data and corresponding real fatigue scores and real emotional state probability distributions. The emotional states include calm and anxiety. Based on the predicted fatigue score, preset emotional state probability distribution, and multimodal physiological data for the current moment, processing tasks with different priorities are generated. The priorities include critical, high priority, routine, and background. The processing tasks are executed according to priority using a hierarchical scheduling algorithm to complete the collection of the air traffic controller's physiological data.

[0020] The terminal 101 can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. The server 102 can be implemented using a standalone server or a server cluster consisting of multiple servers, or it can be a cloud server.

[0021] In one exemplary embodiment, such as Figure 2As shown, a method for collecting physiological data for air traffic controllers is provided. This method is executed by computer equipment, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is applied to... Figure 1 Taking server 102 as an example, the explanation includes the following steps 201 to 204. Wherein: Step 201: Obtain the air traffic controller's current multimodal physiological data; the multimodal physiological data includes electrocardiogram and heart rate monitoring data, respiratory monitoring data, electroencephalogram (EEG) activity monitoring data, electrical skin activity monitoring data, blood oxygen saturation data, and behavioral and posture data.

[0022] Step 202: Input the air traffic controller's current multimodal physiological data into the trained fatigue-emotion assessment model to obtain the predicted fatigue score and predicted emotional state probability distribution at the current moment; the trained fatigue-emotion assessment model is obtained by iteratively training a preset machine learning network model using a sample dataset, the sample dataset including historical multimodal physiological data and corresponding real fatigue scores and real emotional state probability distributions; the emotional states include calmness and anxiety.

[0023] Step 203: Based on the predicted fatigue score, preset emotional state probability distribution, and multimodal physiological data at the current moment, generate processing tasks with different priorities; the priorities include critical level, high priority, regular level, and background level.

[0024] Step 204: The processing tasks are executed according to priority using a hierarchical scheduling algorithm to complete the collection of the air traffic controller's physiological data.

[0025] By implementing steps 201 to 204 above, this application can achieve comprehensive, real-time, and hierarchical response collection and analysis of the physiological state of air traffic controllers.

[0026] In another exemplary embodiment of this application, the air traffic controller's current multimodal physiological data is input into a trained fatigue-emotion assessment model to obtain a predicted fatigue score and a predicted emotional state probability distribution for the current moment, prior to which the following steps are taken: The ZigBee protocol with a star topology is used to acquire multimodal physiological data at the current moment.

[0027] The multimodal physiological data at the current moment are preprocessed to obtain preprocessed multimodal physiological data; the preprocessing operation includes data cleaning.

[0028] In another exemplary embodiment of this application, the training process of the fatigue-emotion assessment model specifically includes: Standardization and time alignment were performed on the multimodal physiological data of historical moments in the sample dataset to obtain spatiotemporally synchronized standardized multimodal physiological data.

[0029] The missing values ​​in the spatiotemporally synchronized standardized multimodal physiological data were filled in using an interpolation method based on Kalman filtering to obtain continuous temporal multimodal physiological data.

[0030] Based on the true fatigue scores and true emotional state probability distributions corresponding to multimodal physiological data at historical moments, a mutual information algorithm is used to extract features from continuous temporal multimodal physiological data to obtain fatigue-sensitive feature data and emotion-sensitive feature data. The fatigue-sensitive feature data includes low-frequency power of heart rate variability, standard deviation of RR interval (SDNN), the theta wave power ratio of EEG, abnormal trend value of respiratory rate, pitch angle fluctuation amplitude of posture stability, persistently low blood oxygen saturation trend index, and fixation point dwell time of eye movement. The emotion-sensitive feature data includes peak frequency of skin conductance response, alpha wave power ratio, beta wave power ratio, and respiratory rhythm entropy value.

[0031] The fatigue-sensitive feature data and the emotion-sensitive feature data were nonlinearly reduced by using the kernel principal component analysis method to obtain low-dimensional fatigue-sensitive feature data and low-dimensional emotion-sensitive feature data.

[0032] Low-dimensional fatigue-sensitive feature data and low-dimensional emotion-sensitive feature data are concatenated into a joint feature vector. (These two types of low-dimensional features, low-dimensional fatigue-sensitive feature data and low-dimensional emotion-sensitive feature data, are concatenated together to form a single joint feature vector).

[0033] A single-hidden-layer feedforward neural network is constructed, and the hidden layer node parameters are iteratively optimized using a crow search algorithm until a preset iteration termination condition is met, resulting in optimized hidden layer node parameters. The hidden layer node parameters include the weight matrix from the input layer to the hidden layer and the bias vector of the hidden layer node. The preset iteration termination condition includes reaching a preset maximum number of iterations or the rate of change of the fitness value for a consecutive preset number of iterations being less than a preset fitness rate of change threshold. The fitness value is calculated based on the weighted sum of the mean square error of the fatigue score and the cross-entropy of the probability distribution of the emotional state.

[0034] The optimal output weights are obtained by solving the Moore-Penrose generalized inverse matrix using the optimized hidden layer node parameters.

[0035] A well-trained fatigue-emotion assessment model is obtained based on the optimized hidden layer node parameters and the optimal output weights.

[0036] In another exemplary embodiment of this application, based on the predicted fatigue score at the current moment, the preset emotional state probability distribution, and multimodal physiological data, processing tasks with different priorities are generated, specifically including: If the predicted fatigue score is greater than or equal to the first preset fatigue threshold, or the probability of anxiety in the preset emotional state probability distribution is greater than or equal to the first preset anxiety threshold, or a first type of abnormal physiological state is detected based on multimodal physiological data, a critical-level processing task is generated; the first type of abnormal physiological state includes arrhythmia, apnea, epileptic brain waves, or sudden loss of consciousness.

[0037] If the predicted fatigue score is greater than or equal to the second preset fatigue threshold and less than the first preset fatigue threshold, or if the anxiety probability in the preset emotional state probability distribution is greater than or equal to the second preset anxiety threshold and less than the first preset anxiety threshold, or if multimodal physiological data detects a second type of abnormal physiological state, a high-priority processing task is generated; the second type of abnormal physiological state includes fatigue warning, emotional stress, early hypoxia, or attention distraction.

[0038] If the predicted fatigue score is less than the first preset fatigue threshold, and the probability of anxiety in the preset emotional state probability distribution is less than the first preset anxiety threshold, and no first or second type of abnormal physiological state is detected in the multimodal physiological data, a routine-level processing task is generated.

[0039] If system resources are idle, and all current critical and high-priority processing tasks have been completed, and no critical or high-priority tasks are being generated or waiting to be processed, then generate background-level processing tasks.

[0040] In another exemplary embodiment of this application, the processing tasks are executed according to priority using a hierarchical scheduling algorithm to complete the collection of physiological data from air traffic controllers, specifically including: A preset proportion of computing resources is reserved for critical processing tasks using a hard real-time approach.

[0041] A priority-based resource preemption strategy is adopted, and the ready queue of high-priority processing tasks is monitored in real time. If the ready queue of high-priority processing tasks is not empty, a preemption signal is triggered. If preset threshold judgment conditions and trend prediction conditions are met, the preemption timing and the computing resources occupied by regular and background processing tasks are dynamically adjusted. The preset threshold judgment conditions include the execution time of regular or background tasks exceeding the corresponding maximum continuous execution time threshold, or the resource occupancy rate of regular or background tasks exceeding the corresponding resource occupancy rate upper limit. The trend prediction conditions include, based on the changing trend of real-time multimodal physiological data, predicting that the trigger probability of high-priority processing tasks in the future preset time window will exceed a first probability threshold; or based on the temporal analysis of historical multimodal physiological data, predicting that the generation frequency of high-priority processing tasks in the future preset time window will exceed a second frequency threshold. That is, based on the changing trend of multimodal physiological data, it is predicted that the probability of high-priority tasks will increase significantly, or based on historical task scheduling data, it is predicted that the generation frequency of high-priority processing tasks will show an upward trend.

[0042] Use Linux cgroups or Kubernetes namespaces to isolate resource access permissions for processing tasks of different priorities.

[0043] In another exemplary embodiment of this application, a dual-algorithm dynamic switching mechanism of AES-256 encryption algorithm and national standard SM4 algorithm is used for data encryption transmission.

[0044] The following example, using the physiological data collection process of a specific air traffic controller, illustrates this application.

[0045] Example 1: A method for collecting physiological data for air traffic controllers includes the following steps: S1. Monitor the electrocardiogram and heart rate, respiratory rate, brain electrical activity, and skin electrical activity of air traffic controllers to collect multimodal physiological data.

[0046] As an optional implementation, the ECG and heart rate monitoring includes: real-time acquisition of ECG signals using a wearable ECG sensor, elimination of motion artifact interference using a filtering algorithm, and simultaneous calculation of heart rate variability (HRV) to assess stress levels. The specific steps are as follows: (1) The real-time optimized Pan-Tompkins algorithm is used for QRS wave detection: the baseline drift and electromyographic noise are eliminated by bandpass filter; the QRS wave is enhanced by differential square operation; the R wave peak is detected by dynamic threshold and the RR interval sequence is generated. The RR interval sequence refers to the time distance between adjacent QRS complex R peaks, in ms.

[0047] (2) Motion artifact removal: Synchronously read triaxial accelerometer data. When the acceleration change of any axis is greater than 2g, mark the corresponding RRi interval as invalid and replace the invalid / abnormal value with cubic spline interpolation.

[0048] (3) HRV calculation: Calculate the standard deviation of NN intervals (SDNN) for the effective RR interval series, and use SDNN to quantify the overall HRV.

[0049] (4) Pressure classification: SDNN < 50ms is judged as high pressure state, SDNN > 100ms is judged as low pressure state.

[0050] As an optional implementation, respiratory rate monitoring includes: using a chest-strap piezoelectric sensor or impedance-based respiratory monitoring device to capture respiratory waveforms through changes in chest cavity expansion, for analyzing respiratory rate and rhythm abnormalities, specifically: (1) Based on the improved Teager Energy Operator (TEO), the inspiratory / expiratory inflection point is enhanced, and the TEO output is dynamically thresholded to detect peak values ​​and mark the inspiratory peak and expiratory valley.

[0051] (2) Count the number of respiratory cycles within a 30s sliding window and calculate the average respiratory rate (RR).

[0052] (3) Threshold rule: RR>25 breaths / minute for 1 minute is considered tachypnea, RR<8 breaths / minute for 30 seconds is considered bradypnea; the threshold can be updated adaptively according to the individual's resting respiratory rate.

[0053] (4) Detection of rhythm abnormalities: Autocorrelation analysis is performed on the respiratory signal. If the peak value of the cycle is >0.7 and meets the time window requirements, the Cyclic Sleep Respiration (CSR) marker is triggered and verified in combination with blood oxygen saturation (SpO2) data.

[0054] As an optional implementation, the brain activity monitoring includes: using a lightweight electroencephalogram (EEG) device to focus on acquiring alpha and beta wave frequency signals, and combining this with event-related potentials (ERPs) to analyze cognitive load and attention concentration. The experimental paradigm design and signal processing flow are as follows: (1) Experimental paradigm of aviation mission simulation events.

[0055] Target stimulus: Sudden conflict alarm, requiring immediate response; Non-target stimulus: Regular radar signal update; Background stimulus: Airspace background noise; Timing control: Stimulus presentation time is 100ms, with random intervals to avoid the expected effect.

[0056] (2) Data collection and preprocessing.

[0057] Electrode configuration: Dry electrode caps are used to focus on leads Fz, Cz, and Pz.

[0058] The noise reduction process includes: ① Bandpass filtering: using a 4th-order Butterworth filter to eliminate low-frequency drift and high-frequency noise; ② Independent Component Analysis (ICA): removing eye movement and electromyography artifacts; ③ Segment locking: extracting time segments (epochs) starting from the stimulus occurrence; ④ Baseline correction: eliminating DC offset using the first 200ms of the stimulus as the baseline, and key ERP component analysis: P300 component (250ms-400ms): reflecting attention allocation and decision-making, with a decrease in amplitude indicating cognitive overload; N1 component (0ms-120ms): reflecting early sensory processing, with a latency extension >20ms indicating inattention; and contingent negative changes. Variation (CNV, -500ms - 0ms): A decrease in amplitude indicates insufficient task readiness, meaning that the brain does not fully mobilize cognitive resources to prepare for the subsequent task during the preparation phase before facing the task. The peak amplitude is measured in lead Cz. The amplitude decreases when the cognitive load increases. An extension of >20ms from the stimulus to the N1 peak indicates a decrease in attention. The negative wave area in the first 500ms of the stimulus is calculated to reflect the degree of task readiness.

[0059] (3) Multimodal fusion and attention assessment.

[0060] Eye-tracking data collaboration: The gaze concentration index (GCI) is calculated by using an eye tracker to measure the time spent between the gaze point and the key area of ​​the radar screen, and the pupil diameter (PD) is recorded. The PD is positively correlated with the P300 amplitude.

[0061] Task performance feedback: Record the target stimulus response time (RT) and error rate (ER) to verify the effectiveness of the ERP indicators. RT is obtained by calculating the key press delay of the target stimulus; ER is obtained by calculating the number of times a non-target stimulus is mistakenly identified as a target stimulus.

[0062] As an optional implementation, electrodermal activity monitoring includes: measuring changes in electrodermal activity (EDA) via finger or wrist electrodes to reflect sympathetic nerve excitation for identifying stress responses; the identification steps include: feature extraction of stress responses. (1) Temporal characteristics. Peak detection of skin conductance response: The first derivative + dynamic threshold method was used to detect the SCR onset point; stress-related parameters include amplitude: peak amplitude of SCR (Skin Conductance Response); rise time: time from onset to peak; recovery time: time required for the peak to drop by 50%; non-specific skin conductance response (NS-SCR): the number of SCRs per minute was counted, >3 times / minute under stress.

[0063] (2) Frequency domain characteristics. Sympathetic nerve index: Calculate the power in the 0.045Hz-0.25Hz frequency band. The power rises by more than 50% of the baseline during stress.

[0064] (3) Nonlinear characteristics. Skin conductance level (SCL) trend: SCL rises slowly using exponential weighted moving average analysis.

[0065] (4) Multimodal fusion and stress determination.

[0066] Acute stress triggering conditions: SCR amplitude > 1 μS and rise time < 1 second, and heart rate > 120 bpm is detected simultaneously.

[0067] Chronic stress triggering conditions: NS-SCR > 5 times / minute for 10 minutes and SCL rise slope > 0.1 μs / min; or when an abnormal increase in heart rate > 120 bpm or a sudden increase in the proportion of delta waves in EEG is detected, the critical task will automatically trigger an alarm and push it to the command center.

[0068] S2. The ZigBee protocol with a star topology is used to aggregate the collected scattered physiological sensor data to the local receiving module. The local receiving module has a built-in embedded processor to perform preliminary data cleaning and feature extraction, reducing the transmission pressure to the cloud.

[0069] As an optional implementation, the preliminary data cleaning includes: (1) aligning the sampling clocks of each sensor with hardware timestamps, controlling the error within ±2ms, and eliminating the impact of timing misalignment on data analysis; (2) deploying a ring buffer to temporarily store the original physiological data to cope with sudden data stream impacts; (3) using an adaptive notch filter to eliminate the impact of power line interference on ECG signals; (4) constructing a motion noise model based on accelerometer data to eliminate motion artifacts. Here, acceleration refers to the acceleration of human motion measured by a triaxial accelerometer. The motion noise model is a mathematical representation method used to quantify and eliminate the interference of body movement on physiological signals. Its core is to establish a mapping relationship between motion and noise through accelerometer data, thereby separating pure physiological components from mixed signals; (5) using wavelet transform to separate high-frequency noise and physiological signal components, and dynamically adjusting the baseline of the respiratory signal using the moving average method.

[0070] This implementation method utilizes wavelet transform to separate high-frequency noise from physiological signal components. The baseline of the respiratory signal is dynamically adjusted using a moving average method. Specifically, it includes: selecting the Daubechies 4 wavelet, balancing tight support and smoothness, suitable for physiological signal characteristics; performing a 5-level decomposition on the ECG / EEG signal; using Stein's unbiased risk estimation threshold to adaptively determine the threshold for each level; and reconstructing the signal after soft thresholding. Real-time optimization includes: sliding window processing, inputting a window every 5 seconds with a 50% overlap to avoid edge effects; and hardware acceleration, utilizing the STM32's FPU (Floating Point Unit) to accelerate floating-point operations, with a single decomposition taking less than 10ms.

[0071] Respiratory signal baseline dynamic correction: Causes of baseline drift: Sensor factors: temperature drift of piezoelectric materials, changes in electrode contact impedance; Physiological factors: slow amplitude fluctuations caused by deep breathing.

[0072] The moving average method is implemented by setting the following parameters: window length W: adaptively adjusted according to the respiratory rate; step size: each time one sampling point is slid, the moving average of the signal within the window is calculated, and the exponential weighted moving average is used to reduce the computational cost. The estimated baseline is subtracted from the original signal.

[0073] S3. The preprocessed data is uploaded to a remote server through an encrypted channel and combined with the CSA-ELM (Crow Search Algorithm-Extreme Learning Machine) algorithm to perform multimodal data fusion to achieve real-time assessment of fatigue and emotional state. Specifically, the following steps are included: (1) Input layer standardization: Z-score normalization is performed on multimodal data from different sensors to eliminate dimensional differences; (2) Dynamic time warping is performed on the asynchronously sampled data stream based on hardware timestamps to ensure synchronization of the time window of multimodal data; (3) Interpolation based on Kalman filtering is used to complete the data missing caused by the temporary failure of the sensor; (4) Mutual information algorithm is used to screen the fatigue / emotion sensitive features in each modality. Among them, emotion-related factors include: peak frequency of skin conductance response, β / α wave power ratio, and respiratory rhythm entropy value; (5) High-dimensional features are mapped to low-dimensional space through kernel principal component analysis, retaining more than 90% of the variance contribution rate, constructing a single hidden layer feedforward neural network, randomly generating the input layer to hidden layer weight matrix WW and bias bb, setting the number of hidden layer nodes to 1.5 times the input features through trial and error, encoding the hidden layer node parameters of ELM as the crow position vector, defining the weighted sum of classification error and model sparsity as the optimization objective, updating parameters through the memory tracking and random walk mechanism of the crow group, avoiding local optima, and the iteration termination condition is that the fitness change is <0.1% for 10 consecutive generations. Using the optimized ELM parameters, the output weight β is solved through Moore-Penrose generalized inverse. (6) The multimodal features after dimensionality reduction are concatenated into a joint feature vector Xfused, inputting Xfused into the CSA-ELM model, and outputting the fatigue score F∈[0,1] and the emotional state probability distribution E=[p 焦虑 p 平静 The classification threshold is dynamically updated based on individual historical baseline data to reduce the impact of individual differences. Continuous evaluation is performed with a 30-second window and a 5-second step size. The output results are smoothed by an exponentially weighted moving average. A red alert (high-priority task) is triggered when any of the following conditions are met: fatigue level F>0.8 lasts for more than 1 minute, anxiety probability pan anxiety>0.7 is accompanied by a sudden increase in delta wave power. The results are pushed to the control console interface via WebSocket, and the controller's status is displayed in real time with a color gradient. Abnormal targets are automatically focused on and displayed.

[0074] Among them, implementing this method, the preprocessed data is uploaded to the remote server through an encrypted channel, including: using the dual algorithm of AES-256 encryption and the national cryptographic SM4 algorithm to dynamically switch between them, and selecting the encryption strength in real time according to the network environment, specifically: (1) Conditions for selecting the AES-256 encryption algorithm: ① Network environment: bandwidth ≥ 10Mbps and latency < 50ms and packet loss rate < 0.1%; ② Security scenario: IDS (Intrusion Detection System) detects potential man-in-the-middle attacks; ③ Transmitted data contains critical physiological alarms; ④ Device status: edge node CPU utilization < 60%.

[0075] (2) Conditions for selecting the national cryptographic SM4 algorithm: ① Network environment: bandwidth <10Mbps or latency ≥50ms or packet loss rate ≥0.1%; ② Security scenario: meet the compliance requirements of China's Information Security Protection 2.0; ③ Regular level data batch upload; ④ Device status: CPU utilization ≥60%.

[0076] The key is updated regularly through quantum key distribution technology, and an independent transmission channel is allocated based on 5G private network slicing technology. Combined with anti-interference modulation, physiological data is prioritized for transmission, and the packet loss rate is suppressed to within 0.05%. The pre-processed multimodal data is encapsulated into standardized data packets, embedded with timestamps and unique device identifiers, and a data fingerprint is generated using the SHA-3 hash algorithm. The receiving end verifies consistency and prevents data tampering during transmission.

[0077] 5G private network slicing is divided into three types: URLLC slices, eMBB slices, and mMTC slices. URLLC slices have a latency of <10ms and a reliability of >99.999%, eMBB slices have a bandwidth of >50Mbps, and mMTC slices have a connection density of >1M devices / km. 2 .

[0078] S4. Define urgency levels for preprocessed data and adopt a four-level priority model, including critical, high priority, normal, and background levels; set hard real-time constraints for critical tasks and deduce resource requirements based on task response time.

[0079] Critical task generation conditions include: arrhythmia: ventricular fibrillation, heart rate >150 bpm or <40 bpm lasting for 5 seconds; apnea: respiratory waveform disappearance for more than 10 seconds; epileptic brain waves: high-frequency spike-and-wave complexes detected; sudden loss of consciousness: falling to the ground without warning detected by posture sensors.

[0080] High-priority generation conditions include: fatigue warning: HRV low-frequency power exceeds the threshold for 3 consecutive cycles + alpha wave ratio >40%; emotional stress: skin conductivity rise slope >0.5μs / s + sudden increase in β wave power ratio; early hypoxia: blood oxygen saturation <92% for 30 seconds; attention distraction: eye tracking shows gaze deviating from the radar screen for >5 seconds + increased EEG theta waves.

[0081] Routine tasks include: recording basic physiological parameters: periodic storage of heart rate, respiratory rate, and body temperature; device self-test: sensor battery level and signal quality; non-urgent feature calculation: long-term trend analysis of RR interval standard deviation, respiratory entropy, etc.; and data synchronization: transmitting non-real-time historical physiological data to a backup server.

[0082] Background tasks include: offline model training: updating the CSA-ELM classifier based on historical data; log compression and archiving: compressing and storing expired data to a local database; software updates: downloading firmware patches; and non-critical statistics: generating fatigue distribution charts for weekly / monthly reports.

[0083] S5. Through a hierarchical scheduling algorithm, high-priority tasks are allowed to preempt resources from low-priority tasks, reserving 20%-30% of computing resources for critical tasks.

[0084] In this embodiment, the hierarchical scheduling algorithm allows high-priority tasks to preempt resources from low-priority tasks by: isolating resource access permissions for tasks at different levels through Linux cgroups or Kubernetes namespaces to prevent low-priority tasks from occupying high-priority resources; setting hardware-level protection policies to prohibit low-priority tasks from accessing critical peripherals; monitoring the ready queue of high-priority tasks in real time; triggering a preemption signal when the queue is not empty; and dynamically adjusting the preemption timing by combining threshold judgment and trend prediction.

[0085] Example 2: A method for collecting physiological data for air traffic controllers includes the following steps: S1. Monitor the electrocardiogram and heart rate, respiratory rate, electroencephalogram (EEG) activity, and electrodermal activity of air traffic controllers, and collect monitoring data.

[0086] S2. The ZigBee protocol with a star topology is used to aggregate the collected scattered physiological sensor data to the local receiving module. The local receiving module has a built-in embedded processor to perform preliminary data cleaning and feature extraction, reducing the transmission pressure to the cloud.

[0087] S3. The preprocessed data is uploaded to a remote server through an encrypted channel for multimodal data fusion to achieve real-time assessment of fatigue and emotional state.

[0088] As an alternative implementation, the fatigue-emotion assessment model for assessing fatigue and emotional state can also be a weighted scoring model constructed by integrating HRV low-frequency power, eye-tracking data, and posture stability indicators. The weighted scoring model combines the alpha / beta wave power ratio of EEG and the peak frequency of skin conductance responses. It uses a support vector machine (SVM) to classify anxiety / calm states, adaptively adjusting the alarm threshold based on individual baseline data to reduce misjudgment rates caused by individual differences. A sliding time window is introduced to achieve continuous state tracking and capture sudden abnormal fluctuations. Evaluation results are pushed to the command center via WebSocket protocol, displaying the group's fatigue distribution in the form of a heatmap. Key targets are automatically marked with red alerts. The weighted scoring model includes an input layer, a feature weighting layer, a nonlinear decision layer, and a dynamic feedback layer. The input layer extracts multimodal features, the feature weighting layer assigns weights, and the nonlinear decision layer uses SVM for classification: Input: weighted score + original features; Kernel function: RBF (Radial Basis Function); Output: probability of anxiety / calm state. The dynamic feedback layer features: threshold adaptation; baseline calibration: daily resting test updates to individual mean μ and standard deviation σ; alarm triggering: Score > μ + 2σ and SVM probability > 0.7; sliding window optimization: window length: 60 seconds; step size: 5 seconds.

[0089] S4. Define the urgency level of the preprocessed data and adopt a four-level priority model, including critical, high priority, normal, and background levels; set hard real-time constraints for critical tasks and deduce resource requirements based on task response time. Critical tasks include arrhythmia and apnea.

[0090] S5. Through a hierarchical scheduling algorithm, high-priority tasks are allowed to preempt resources from low-priority tasks, reserving 25% of computing resources for critical tasks.

[0091] In this embodiment, ECG and heart rate monitoring includes: real-time acquisition of ECG signals through a wearable ECG sensor, elimination of motion artifact interference by combining a filtering algorithm, and simultaneous calculation of heart rate variability to assess stress level. Respiratory rate monitoring includes: capturing respiratory waveforms through changes in chest cavity expansion using a chest strap piezoelectric sensor or impedance method respiratory monitoring device to analyze respiratory frequency and rhythm abnormalities.

[0092] In this embodiment, EEG activity monitoring includes: using lightweight EEG equipment to focus on collecting signals in the alpha and beta wave frequency bands, and combining event-related potentials to analyze cognitive load and attention concentration. Skin conductance activity monitoring includes: measuring changes in skin conductivity through finger or wrist electrodes to reflect sympathetic nerve excitability and to identify stress responses. When an abnormal increase in heart rate (>120 bpm) or a sudden increase in the proportion of delta waves in the EEG is detected, an alarm is automatically triggered and pushed to the command center.

[0093] In this embodiment, the preliminary data cleaning includes: aligning the sampling clocks of each sensor with hardware timestamps to control the error within ±2ms, eliminating the impact of timing misalignment on data analysis; deploying a circular buffer to temporarily store the raw data to cope with sudden data stream surges; using an adaptive notch filter to eliminate the impact of power line interference on ECG signals; constructing a motion noise model based on accelerometer data; using wavelet transform to separate high-frequency noise from physiological signal components; and dynamically adjusting the baseline for low-frequency components such as respiratory signals using the moving average method.

[0094] In this embodiment, feature extraction includes: using the Pan-Tompkins algorithm to identify the R-wave peak value in real time, calculating the instantaneous heart rate and the standard deviation of the RR interval, extracting the ST segment slope within a fixed time window of 100ms after the R-wave, assessing the risk of myocardial ischemia, dividing the inspiratory / expiratory phase based on extreme point detection, calculating the relative changes in respiratory rate and tidal volume, using a dynamic time warping algorithm to match preset abnormal patterns, and using a Kalman filter to fuse accelerometer and gyroscope data to output pitch / roll angles.

[0095] In this embodiment, the preprocessed data is uploaded to the remote server through an encrypted channel, including: dynamically switching between AES-256 encryption and the national standard SM4 encryption algorithm; selecting encryption strength in real time according to the network environment; periodically updating the key using quantum key distribution technology; allocating an independent transmission channel based on 5G private network slicing technology; combining anti-interference modulation to ensure priority transmission of physiological data; suppressing the packet loss rate to within 0.05%; encapsulating the preprocessed multimodal data into standardized data packets; embedding timestamps and unique device identifiers; generating data fingerprints using the SHA-3 hash algorithm; verifying consistency at the receiving end; and preventing data tampering during transmission.

[0096] In this embodiment, the assessment of fatigue and emotional state includes: fusing HRV low-frequency power, eye-tracking data, and posture stability indicators to construct a weighted scoring model; combining the EEG α / β wave power ratio and skin conductance response peak frequency; classifying anxiety / calm states using support vector machines; adaptively adjusting alarm thresholds based on individual baseline data to reduce misjudgment rates caused by individual differences; introducing a sliding time window to achieve continuous state tracking and capture sudden abnormal fluctuations; pushing assessment results to the command center via the WebSocket protocol; displaying the group fatigue distribution in the form of a heatmap; and automatically marking key targets with red warnings.

[0097] In this embodiment, the hierarchical scheduling algorithm allows high-priority tasks to preempt resources from low-priority tasks by: isolating resource access permissions for tasks at different levels through Linux cgroups or Kubernetes namespaces to prevent low-priority tasks from occupying high-priority resources; setting hardware-level protection policies to prohibit low-priority tasks from accessing critical peripherals; monitoring the ready queue of high-priority tasks in real time; triggering a preemption signal when the queue is not empty; and dynamically adjusting the preemption timing by combining threshold judgment and trend prediction.

[0098] Example 3: A method for collecting physiological data for air traffic controllers includes the following steps: S1. Monitor the electrocardiogram and heart rate, respiratory rate, electroencephalogram (EEG) activity, and electrodermal activity of air traffic controllers, and collect monitoring data.

[0099] S2. The ZigBee protocol with a star topology is used to aggregate the collected scattered physiological sensor data to the local receiving module. The local receiving module has a built-in embedded processor to perform preliminary data cleaning and feature extraction, reducing the transmission pressure to the cloud.

[0100] S3. The preprocessed data is uploaded to a remote server through an encrypted channel. Multimodal data fusion is performed through a weighted scoring model to achieve real-time assessment of fatigue and emotional state.

[0101] S4. Define the urgency level of the preprocessed data and adopt a four-level priority model, including critical, high priority, normal, and background levels; set hard real-time constraints for critical tasks and deduce resource requirements based on task response time. Critical tasks include arrhythmia and apnea.

[0102] S5. Through a hierarchical scheduling algorithm, high-priority tasks are allowed to preempt resources from low-priority tasks, reserving 30% of computing resources for critical tasks.

[0103] In this embodiment, ECG and heart rate monitoring includes: real-time acquisition of ECG signals through a wearable ECG sensor, elimination of motion artifact interference by combining a filtering algorithm, and simultaneous calculation of heart rate variability to assess stress level. Respiratory rate monitoring includes: capturing respiratory waveforms through changes in chest cavity expansion using a chest strap piezoelectric sensor or impedance method respiratory monitoring device to analyze respiratory frequency and rhythm abnormalities.

[0104] In this embodiment, EEG activity monitoring includes: using lightweight EEG equipment to focus on collecting signals in the alpha and beta wave frequency bands, and combining event-related potentials to analyze cognitive load and attention concentration. Skin conductance activity monitoring includes: measuring changes in skin conductivity through finger or wrist electrodes to reflect sympathetic nerve excitability and to identify stress responses. When an abnormal increase in heart rate (>120 bpm) or a sudden increase in the proportion of delta waves in the EEG is detected, an alarm is automatically triggered and pushed to the command center.

[0105] In this embodiment, the preliminary data cleaning includes: aligning the sampling clocks of each sensor with hardware timestamps to control the error within ±2ms, eliminating the impact of timing misalignment on data analysis; deploying a circular buffer to temporarily store the raw data to cope with sudden data stream surges; using an adaptive notch filter to eliminate the impact of power line interference on ECG signals; constructing a motion noise model based on accelerometer data; using wavelet transform to separate high-frequency noise from physiological signal components; and dynamically adjusting the baseline for low-frequency components such as respiratory signals using the moving average method.

[0106] In this embodiment, feature extraction includes: using the Pan-Tompkins algorithm to identify the R-wave peak value in real time, calculating the instantaneous heart rate and the standard deviation of the RR interval, extracting the ST segment slope within a fixed time window of 120ms after the R-wave, assessing the risk of myocardial ischemia, dividing the inspiratory / expiratory phases based on extreme point detection, calculating the relative changes in respiratory rate and tidal volume, using a dynamic time warping algorithm to match preset abnormal patterns, and using a Kalman filter to fuse accelerometer and gyroscope data to output pitch / roll angles.

[0107] In this embodiment, the preprocessed data is uploaded to the remote server through an encrypted channel, including: dynamically switching between AES-256 encryption and the national standard SM4 encryption algorithm; selecting encryption strength in real time according to the network environment; periodically updating the key using quantum key distribution technology; allocating an independent transmission channel based on 5G private network slicing technology; combining anti-interference modulation to ensure priority transmission of physiological data; suppressing the packet loss rate to within 0.05%; encapsulating the preprocessed multimodal data into standardized data packets; embedding timestamps and unique device identifiers; generating data fingerprints using the SHA-3 hash algorithm; verifying consistency at the receiving end; and preventing data tampering during transmission.

[0108] In this embodiment, the assessment of fatigue and emotional state includes: fusing HRV low-frequency power, eye-tracking data, and posture stability indicators to construct a weighted scoring model; combining the EEG α / β wave power ratio and skin conductance response peak frequency; classifying anxiety / calm states using support vector machines; adaptively adjusting alarm thresholds based on individual baseline data to reduce misjudgment rates caused by individual differences; introducing a sliding time window to achieve continuous state tracking and capture sudden abnormal fluctuations; pushing assessment results to the command center via the WebSocket protocol; displaying the group fatigue distribution in the form of a heatmap; and automatically marking key targets with red warnings.

[0109] In this embodiment, the hierarchical scheduling algorithm allows high-priority tasks to preempt resources from low-priority tasks by: isolating resource access permissions for tasks at different levels through Linux cgroups or Kubernetes namespaces to prevent low-priority tasks from occupying high-priority resources; setting hardware-level protection policies to prohibit low-priority tasks from accessing critical peripherals; monitoring the ready queue of high-priority tasks in real time; triggering a preemption signal when the queue is not empty; and dynamically adjusting the preemption timing by combining threshold judgment and trend prediction.

[0110] Based on the same inventive concept, this application also provides a physiological data acquisition system for air traffic controllers to implement the above-described method for acquiring physiological data for air traffic controllers. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations of one or more embodiments of the physiological data acquisition system for air traffic controllers provided below can be found in the limitations of the physiological data acquisition method for air traffic controllers described above, and will not be repeated here.

[0111] In one exemplary embodiment, such as Figure 3 As shown, a physiological data acquisition system for air traffic controllers is provided, comprising: The multimodal physiological data acquisition module 301 is used to acquire the multimodal physiological data of the air traffic controller at the current moment; the multimodal physiological data includes electrocardiogram and heart rate monitoring data, respiratory monitoring data, electroencephalogram activity monitoring data, skin conductance activity monitoring data, blood oxygen saturation data, and behavior and posture data.

[0112] The fatigue-emotion assessment module 302 is used to input the air traffic controller's current multimodal physiological data into a trained fatigue-emotion assessment model to obtain the predicted fatigue score and predicted emotional state probability distribution for the current moment. The trained fatigue-emotion assessment model is obtained by iteratively training a preset machine learning network model using a sample dataset. The sample dataset includes multimodal physiological data from historical moments and the corresponding actual fatigue scores and actual emotional state probability distributions. The emotional states include calmness and anxiety.

[0113] The task generation module 303 is used to generate processing tasks with different priorities based on the predicted fatigue score, preset emotional state probability distribution, and multimodal physiological data at the current moment; the priorities include critical level, high priority, normal level, and background level.

[0114] The task scheduling and execution module 304 is used to execute the processing tasks according to priority through a hierarchical scheduling algorithm to complete the collection of physiological data of air traffic controllers.

[0115] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 4 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores the physiological data of air traffic controllers. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a method for collecting physiological data from air traffic controllers.

[0116] Those skilled in the art will understand that Figure 4 The structures shown are merely block diagrams of some structures related to the present application and do not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0117] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0118] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0119] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0120] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0121] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0122] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0123] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for collecting physiological data for air traffic controllers, characterized in that, The method for collecting physiological data for air traffic controllers includes: Acquire multimodal physiological data of air traffic controllers at the current moment; the multimodal physiological data includes electrocardiogram and heart rate monitoring data, respiratory monitoring data, electroencephalogram (EEG) activity monitoring data, electrical skin activity monitoring data, blood oxygen saturation data, and behavioral and posture data; The air traffic controller's current multimodal physiological data is input into a pre-trained fatigue-emotion assessment model to obtain the predicted fatigue score and predicted emotional state probability distribution for the current moment. The pre-trained fatigue-emotion assessment model is obtained by iteratively training a preset machine learning network model using a sample dataset, which includes historical multimodal physiological data and corresponding actual fatigue scores and actual emotional state probability distributions. The emotional states include calmness and anxiety. Based on the predicted fatigue score at the current moment, the preset emotional state probability distribution, and multimodal physiological data, processing tasks with different priorities are generated; the priorities include critical level, high priority, routine level, and background level. The processing tasks are executed according to priority using a hierarchical scheduling algorithm to complete the collection of physiological data from air traffic controllers.

2. The method for collecting physiological data for air traffic controllers according to claim 1, characterized in that, The air traffic controller's current multimodal physiological data is input into a trained fatigue-emotion assessment model to obtain the predicted fatigue score and predicted emotional state probability distribution for the current moment, including: The ZigBee protocol with a star topology is used to acquire multimodal physiological data at the current moment; The multimodal physiological data at the current moment are preprocessed to obtain preprocessed multimodal physiological data; the preprocessing operation includes data cleaning.

3. The method for collecting physiological data for air traffic controllers according to claim 2, characterized in that, The training process of the fatigue-emotion assessment model specifically includes: Standardization and time alignment are performed on the multimodal physiological data of historical moments in the sample dataset to obtain spatiotemporally synchronized standardized multimodal physiological data; The missing values ​​in the spatiotemporally synchronized standardized multimodal physiological data were filled in using an interpolation method based on Kalman filtering to obtain continuous temporal multimodal physiological data; Based on the true fatigue scores and true emotional state probability distributions corresponding to multimodal physiological data at historical moments, a mutual information algorithm is used to extract features from continuous time-series multimodal physiological data to obtain fatigue-sensitive feature data and emotion-sensitive feature data. The fatigue-sensitive feature data includes low-frequency power of heart rate variability, standard deviation of RR interval, the theta wave power ratio of EEG, abnormal trend value of respiratory rate, pitch angle fluctuation amplitude of posture stability, persistently low blood oxygen saturation trend index, and fixation point dwell time of eye movement. The emotion-sensitive feature data includes peak frequency of skin conductance response, alpha wave power ratio, beta wave power ratio, and respiratory rhythm entropy value. The fatigue-sensitive feature data and the emotion-sensitive feature data were nonlinearly reduced by kernel principal component analysis to obtain low-dimensional fatigue-sensitive feature data and low-dimensional emotion-sensitive feature data. Low-dimensional fatigue-sensitive feature data and low-dimensional emotion-sensitive feature data are concatenated into a joint feature vector; A single-hidden-layer feedforward neural network is constructed, and the hidden layer node parameters are iteratively optimized using a crow search algorithm until a preset iteration termination condition is met, resulting in optimized hidden layer node parameters. The hidden layer node parameters include the weight matrix from the input layer to the hidden layer and the bias vector of the hidden layer node. The preset iteration termination condition includes reaching a preset maximum number of iterations or the rate of change of the fitness value for a consecutive preset number of iterations being less than a preset fitness change rate threshold. The fitness value is calculated based on the weighted sum of the mean square error of the fatigue score and the cross-entropy of the probability distribution of the emotional state. Using the optimized hidden layer node parameters, the optimal output weights are obtained by solving the Moore-Penrose generalized inverse matrix. A well-trained fatigue-emotion assessment model is obtained based on the optimized hidden layer node parameters and the optimal output weights.

4. The method for collecting physiological data for air traffic controllers according to claim 1, characterized in that, Based on the predicted fatigue score at the current moment, the preset emotional state probability distribution, and multimodal physiological data, processing tasks with different priorities are generated, specifically including: If the predicted fatigue score is greater than or equal to the first preset fatigue threshold, or the probability of anxiety in the preset emotional state probability distribution is greater than or equal to the first preset anxiety threshold, or a first type of abnormal physiological state is detected based on multimodal physiological data, a critical-level processing task is generated; the first type of abnormal physiological state includes arrhythmia, apnea, epileptic brain waves, or sudden loss of consciousness. If the predicted fatigue score is greater than or equal to the second preset fatigue threshold and less than the first preset fatigue threshold, or the anxiety probability in the preset emotional state probability distribution is greater than or equal to the second preset anxiety threshold and less than the first preset anxiety threshold, or the multimodal physiological data detects a second type of abnormal physiological state, a high-priority processing task is generated; the second type of abnormal physiological state includes fatigue warning, emotional stress, early hypoxia, or attention deficit. If the predicted fatigue score is less than the first preset fatigue threshold, and the probability of anxiety in the preset emotional state probability distribution is less than the first preset anxiety threshold, and no first or second type of abnormal physiological state is detected in the multimodal physiological data, a routine processing task is generated. If system resources are idle, and all current critical and high-priority processing tasks have been completed, and no critical or high-priority tasks are being generated or waiting to be processed, then generate background-level processing tasks.

5. The method for collecting physiological data for air traffic controllers according to claim 1, characterized in that, The processing tasks are executed according to priority using a hierarchical scheduling algorithm to complete the collection of physiological data from air traffic controllers, specifically including: A predetermined proportion of computing resources is reserved for critical processing tasks using hard real-time methods. A priority-based resource preemption strategy is adopted, and the ready queue of high-priority processing tasks is monitored in real time. If the ready queue of high-priority processing tasks is not empty, a preemption signal is triggered. If preset threshold judgment conditions and trend prediction conditions are met, the preemption timing and the computing resources occupied by regular and background processing tasks are dynamically adjusted. The preset threshold judgment conditions include the execution time of regular or background tasks exceeding the corresponding maximum continuous execution time threshold, or the resource occupancy rate of regular or background tasks exceeding the corresponding resource occupancy rate upper limit. The trend prediction conditions include the prediction that the trigger probability of high-priority processing tasks will exceed a first probability threshold in the future preset time window based on the changing trend of real-time multimodal physiological data; or the prediction that the generation frequency of high-priority processing tasks will exceed a second frequency threshold in the future preset time window based on the temporal analysis of historical multimodal physiological data. Use Linux cgroups or Kubernetes namespaces to isolate resource access permissions for processing tasks of different priorities.

6. The method for collecting physiological data for air traffic controllers according to claim 1, characterized in that, A dual-algorithm dynamic switching mechanism using AES-256 encryption and the national standard SM4 algorithm is employed for encrypted data transmission.

7. A physiological data acquisition system for air traffic controllers, characterized in that, The physiological data acquisition system for air traffic controllers uses the physiological data acquisition method for air traffic controllers according to any one of claims 1-6, and the physiological data acquisition system for air traffic controllers includes: The multimodal physiological data acquisition module is used to acquire the air traffic controller's multimodal physiological data at the current moment; the multimodal physiological data includes electrocardiogram and heart rate monitoring data, respiratory monitoring data, electroencephalogram (EEG) activity monitoring data, electrical skin activity monitoring data, blood oxygen saturation data, and behavioral and posture data; The fatigue-emotion assessment module is used to input the air traffic controller's current multimodal physiological data into a pre-trained fatigue-emotion assessment model to obtain the predicted fatigue score and predicted emotional state probability distribution for the current moment. The pre-trained fatigue-emotion assessment model is obtained by iteratively training a preset machine learning network model using a sample dataset. The sample dataset includes historical multimodal physiological data and corresponding actual fatigue scores and actual emotional state probability distributions. The emotional states include calmness and anxiety. The task generation module is used to generate processing tasks of different priorities based on the predicted fatigue score, preset emotional state probability distribution, and multimodal physiological data at the current moment; the priorities include critical level, high priority, normal level, and background level. The task scheduling and execution module is used to execute the processing tasks according to priority through a hierarchical scheduling algorithm to complete the collection of physiological data of air traffic controllers.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the physiological data acquisition method for air traffic controllers as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the physiological data acquisition method for air traffic controllers as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the physiological data acquisition method for air traffic controllers as described in any one of claims 1-6.

Images