Air syncope early warning and protection system based on multi-modal data

Abstract

The present application relates to a kind of in-flight syncope early warning and protection system based on multi-modal data, solve the technical problem that the precision of real-time monitoring pilot head physiological parameter data in actual civil aviation flight environment is low in prior art, and the reliability of monitoring pilot anoxia and syncope state is low;Including multi-modal physiological sensing module and main control module;The multi-modal physiological sensing module is used to obtain the physiological parameter data and motion data in the process of pilot flight in real time;The main control module is used to receive the physiological parameter data and motion data and pre-process;Based on the physiological parameter data and motion data after pre-processing, it is judged whether pilot has in-flight syncope;If syncope, syncope warning is carried out and airplane is automatically changed flat, to ensure flight safety.The accuracy of obtaining the physiological parameter data of civil aviation pilot is improved, and the real-time reliable monitoring of pilot anoxia and syncope.

Description

Technical Field

[0001] This invention relates to the fields of aviation medicine, civil aviation piloting, human-computer interaction and artificial intelligence, and in particular to an in-flight syncope early warning and protection system based on multimodal data. Background Technology

[0002] Small civil aircraft, encompassing regional jets and general aviation aircraft, are the main aircraft types supporting short-haul transportation, the low-altitude economy, and flight training. Recently, the low-altitude economy has been identified as an emerging pillar industry, and small civil aircraft are moving towards a new stage of green intelligence and diversified application scenarios.

[0003] Pilots of small commercial aircraft must maintain a high level of concentration for extended periods, enduring continuous cognitive and physiological loads, and facing multiple physiological challenges in complex flight environments. Some small commercial aircraft can cruise at altitudes exceeding 30,000 feet, where atmospheric pressure drops significantly. Although the cockpit is pressurized, sudden events such as cabin pressurization failure, seal breaches, or oxygen supply system malfunctions can rapidly cause a drop in oxygen partial pressure within the cabin, leading to acute high-altitude hypoxia in the pilot. In a hypoxic state, the pilot's pulse oxygen saturation can rapidly drop from the normal level (95%–100%) to below 85%, a critical hypoxic range. Insufficient oxygen supply to brain tissue directly affects the function of the central nervous system.

[0004] Meanwhile, although small commercial aircraft do not experience severe high-G maneuvers during flight, they still experience a certain degree of inertial G-force when encountering strong turbulence, sudden attitude adjustments, emergency descents, or evasive maneuvers. This causes fluctuations in cerebral blood flow perfusion, further exacerbating insufficient blood and oxygen supply to the brain tissue. The changes in cerebral perfusion caused by high-altitude hypoxia and transient G-forces during flight create synergistic physiological stress, jointly triggering an acute state of cerebral ischemia and hypoxia. Physiological studies have shown that when local cerebral oxygen saturation (… When oxygen saturation drops below 50%, abnormal nerve conduction can occur, manifesting as symptoms such as confusion, decreased attention, slow reaction, and narrowed visual field. In severe cases, it can lead to a decline in consciousness, directly endangering flight safety. Therefore, real-time, continuous, and non-invasive physiological monitoring of civil aviation pilots, especially dynamic monitoring of key indicators such as brain oxygen and blood oxygen, is of great significance for early warning of dangerous conditions such as hypoxia and abnormal consciousness, and for ensuring civil aviation flight safety.

[0005] Existing civil aviation flight safety monitoring systems mostly focus on the independent monitoring of single physiological indicators (such as pulse oximetry) or pilot operational status, lacking the ability to perform multimodal fusion analysis of physiological and flight-related data. Traditional systems often rely on subjective reports from pilots or post-event retrospective analysis by ground control towers, making it difficult to achieve real-time and accurate early warnings of dangerous situations. Furthermore, most systems only possess monitoring functions and lack automatically triggered protection mechanisms, failing to respond quickly to pilot anomalies to prevent accidents. For example, some airborne monitoring equipment only displays physiological parameter values, requiring pilots or ground personnel to manually assess the risk. In sudden scenarios such as high-altitude hypoxia, pilots may have already lost their normal operational capabilities, leading to delays in early warning and protection. In addition, the data storage and retrospective mechanisms of existing systems are inadequate, making it difficult to support subsequent flight safety assessments and event analysis. Summary of the Invention

[0006] Based on the above analysis, the present invention aims to provide an in-flight syncope early warning and protection system based on multimodal data, in order to solve the technical problems of low accuracy of real-time monitoring of civil aviation pilots' head physiological parameters in actual flight environments and low reliability of monitoring civil aviation pilots' hypoxia and syncope.

[0007] This invention provides an in-flight syncope early warning and protection system based on multimodal data, including a multimodal physiological sensing module and a main control module; The multimodal physiological sensing module is used to acquire multimodal data, including physiological parameter data and motion data, of civil aviation pilots during flight in real time; The main control module is used to receive the physiological parameter data and motion data and preprocess them; based on the preprocessed physiological parameter data and motion data, it determines whether the pilot has experienced in-flight syncope; if syncope occurs, it issues a syncope warning and automatically levels the aircraft to ensure the pilot's flight safety.

[0008] Furthermore, the main control module is installed on the rear side of the pilot's flight helmet and connected to the multimodal physiological sensing module, including a data processing chip, a storage module, a Bluetooth communication device, and a battery; The data processing chip is used to preprocess the physiological parameter data and motion data, and to execute a head physiological monitoring method for judging the pilot's in-flight hypoxia and syncope based on the preprocessed physiological parameter data and motion data, thereby determining whether the pilot has experienced in-flight syncope. The storage module is used to store the physiological parameter data and motion data, as well as the result of whether fainting is determined by the data processing module based on the preprocessed physiological parameter data and motion data. The Bluetooth communication device is used to send the preprocessed physiological parameter data and motion data, as well as the corresponding result of whether or not the patient has fainted, to the airborne terminal. The battery is used to power the multimodal physiological sensing module and the main control module.

[0009] Furthermore, the multimodal physiological sensing module includes a PPG sensor and a motion sensor; The PPG sensor is installed on the inner forehead area of ​​the pilot's flight helmet and is adjusted to fit snugly against the wearer's forehead for real-time acquisition of the pilot's physiological parameter data; wherein, the physiological parameter data includes Value and PI value; The motion sensor is installed inside the pilot's flight helmet at the center of the top of the head to acquire the pilot's motion data in real time; wherein, the motion data includes head motion data, pitch angle, yaw angle, and roll angle; the head motion data includes a time window. Number of nods Nodding frequency Angle of looking down and the percentage of people looking down Among them, time window Nodding frequency within and the percentage of people looking down based on , Calculated.

[0010] Furthermore, the system also includes an airborne terminal and a civil aviation ground control tower; The airborne terminal is used to, based on the received result of whether or not the aircraft has fainted, issue a warning voice broadcast if fainting occurs, and automatically level the aircraft; and to send the physiological parameter data and motion data, the preprocessed physiological parameter data and motion data, the result of whether or not the aircraft has fainted, and the result of whether or not the aircraft has automatically leveled the aircraft to the civil aviation ground control via a data link. The civil aviation ground control tower is used to display the pilot's physiological parameter trends and motion status trends in real time based on the received data. At the same time, it classifies and stores the received data to provide data support for subsequent flight safety assessments and incident tracing. If the pilot experiences in-flight fainting, it sends an emergency command to the airborne terminal through the data link.

[0011] Furthermore, the method for monitoring the head physiology of the pilot to determine in-flight hypoxia and syncope includes: Based on real-time acquisition The value and the predetermined driver's The baseline value is used to obtain the driver's blood oxygen decay ratio; based on the motion data and the pre-trained blood oxygen prediction model, a prediction result is obtained as to whether the driver is hypoxic. Based on the blood oxygen attenuation ratio and the real-time acquired... The value is used to determine whether the driver is suspected of having hypoxia; If hypoxia is suspected, the pilot's hypoxia status will be determined based on the predicted hypoxia results. If hypoxia occurs in the air, the driver's condition is determined by the PI attenuation coefficient obtained based on the PI value and the pre-determined PI reference value of the driver. If the pilot's condition is suspected of being unconscious, the system will determine whether the pilot is unconscious based on real-time motion data. If the pilot is unconscious, the aircraft will automatically level off.

[0012] Furthermore, the standard deviation of the pitch angle is calculated based on the pitch angle, yaw angle, and roll angle, respectively. Yaw angle standard deviation and roll angle standard deviation ; Will , , , and After preprocessing, the data is input into a pre-trained blood oxygen prediction model to obtain a prediction result of whether the driver is hypoxic.

[0013] Furthermore, if or If so, it is determined that the driver is suspected of having hypoxia; among them, This represents the percentage of blood oxygen depletion. The threshold for the proportion of blood oxygen decay. For every The real-time monitoring of the second The average value for Threshold.

[0014] Furthermore, if it is determined that the driver is suspected of hypoxia, and the prediction result of hypoxia is hypoxia, then it is determined that the driver has experienced in-flight hypoxia. Based on PI attenuation coefficient To determine the driver's condition; like If so, the driver's condition is judged to be suspected of fainting; like If so, the driver's condition is determined to be extremely low infusion. like If so, the driver's condition is determined to be low perfusion; like If so, the driver's status is determined to be normal (PI). in, , and These are the threshold values ​​for the first, second, and third PI attenuation coefficients, respectively. ; If the driver's condition is suspected to be syncope, the system will determine whether the driver is syncope based on the pitch angle in the real-time motion data. If the absolute value of the pitch angle in the real-time motion data is greater than the preset pitch angle threshold, then the driver is confirmed to be syncope.

[0015] Furthermore, the adjustment mechanism is installed in the forehead area of ​​the pilot's flight cap to adjust the fit position of the PPG sensor with the forehead based on the wearer's hairline position, so as to ensure that the PPG sensor can fit in close contact with the pilot's forehead skin; The adjustment mechanism includes a fixed base, a sliding base, a lateral adjustment device, a longitudinal adjustment device, and a driver; The mounting base is installed on the forehead area of ​​the pilot's flight helmet; the slide is mounted on the mounting base and can slide on the mounting base, the slide being used to mount the PPG sensor; The lateral adjustment device is used to control the slide to move laterally in the plane; The longitudinal adjustment device is used to control the slide to move longitudinally in the plane; The driver is used to control the operation of the lateral adjustment device and the longitudinal adjustment device respectively.

[0016] Furthermore, the blood oxygen prediction model includes an input layer, an LSTM temporal feature extraction layer, and a fully connected classification decision layer; the LSTM layer is used to extract temporal features; the fully connected layer makes classification decisions based on the temporal features. The pre-trained blood oxygen prediction model is obtained through the following process: Obtain physiological parameters and motion data of the heads of multiple drivers within a certain historical period to obtain corresponding... , , , and After preprocessing, the samples are used as training samples, and together with the corresponding samples labeled as hypoxic, they form a training sample set. The loss function is the binary cross-entropy loss between the prediction of hypoxia and the true label of the sample; the Adam optimizer is used for iterative training. If continuous The accuracy of the validation set in epochs is less than Or the F1 score on the validation set is greater than or equal to And the recall rate is greater than or equal to If the training stops, the well-trained blood oxygen prediction model is obtained. in, , and These are the accuracy threshold, F1 threshold, and recall threshold, respectively.

[0017] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects: 1. This invention, through the integration of the PPG sensor and adjustment mechanism in the multimodal physiological sensing module, ensures that the PPG sensor is in close contact with the driver's forehead skin, reducing interference from the driver's hair and improving cerebral oxygen saturation. The accuracy of collecting physiological parameter data of PI perfusion index; 2. Based on preprocessed physiological parameter data and motion data, this invention combines a blood oxygen prediction model extracted by LSTM time-series features to achieve a progressive judgment from driver hypoxia to syncope, solving the problem of traditional methods that separate hypoxia and syncope monitoring and have delayed early warning, and improving the reliability of hypoxia and syncope monitoring. 3. This invention combines two types of data—the driver's physiological parameters and head movement data—for collaborative judgment. It uses dual verification of the proportion of blood oxygen decay (physiological parameters) and movement characteristics (standard deviations of head nodding frequency, head-down percentage, pitch angle, yaw angle, and roll angle) to determine whether the driver is experiencing in-flight hypoxia. This is because individual cerebral oxygen saturation varies. The differences are significant, and the decrease in oxygen levels varies greatly under hypoxic conditions. Therefore, relying on a single indicator can lead to a large number of misjudgments. Furthermore, the head movement characteristics of drivers also change significantly after hypoxia. Combining multi-dimensional data on head movement characteristics for verification greatly improves the accuracy of judging whether a driver is hypoxic and reduces the false positive rate. 4. The personalized baseline values ​​and dynamic thresholds of this invention are adapted to individual pilot differences. Before flight, the individual pilot's... Benchmark values ​​and PI benchmark values; to avoid judgment bias caused by individual differences and improve the targeting of monitoring; 5. This invention utilizes a time window in the process of determining hypoxia by analyzing the driver's head movement characteristics. Nodding frequency within and the percentage of people looking down Standard deviation of pitch angle Yaw angle standard deviation and roll angle standard deviation Five indicators, all of which change significantly under hypoxic conditions, can be combined using machine learning methods to automatically obtain a binary classification result of whether hypoxia has occurred. The entire process requires no manual intervention. Furthermore, since the blood oxygen prediction model has been trained on a large amount of data, the classification result is accurate. This classification method can identify different head movements caused by hypoxia, and has a wider recognition range than a single indicator. 6. The method in this invention provides real-time monitoring at the second level, controlling the delay from data acquisition to status judgment to the second level, thus providing time for subsequent intervention in case of pilot fainting. If pilot fainting is confirmed, the aircraft is automatically leveled without manual intervention, preventing loss of aircraft control due to pilot incapacitation and reducing the risk of accidents.

[0018] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained from the description and drawings, which are particularly pointed out. Attached Figure Description

[0019] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.

[0020] Figure 1 This is a schematic diagram of a multimodal data-based in-flight syncope early warning and protection system module in an embodiment of the present invention; Figure 2 This is one of the structural schematic diagrams of the aviation flight cap in the embodiments of the present invention; Figure 3 This is a second schematic diagram of the structure of the aviation flight cap in this embodiment of the invention; Figure 4 This is one of the structural schematic diagrams of the adjustment mechanism in an embodiment of the present invention; Figure 5 This is a second schematic diagram of the adjustment mechanism in an embodiment of the present invention; Figure 6 This is a cross-sectional view of the adjustment mechanism in an embodiment of the present invention; Figure 7 for Figure 6 A magnified view of a portion of the image; Figure 8 This is a flowchart of a head physiological monitoring method for determining hypoxia and syncope in an embodiment of the present invention; Figure 9 This is a schematic diagram of the blood oxygen prediction model structure and training in an embodiment of the present invention.

[0021] Figure label: 1-Pilot's flight cap; 2-PPG sensor; 3-Motion sensor; 4-Main control module; 5-Adjustment mechanism; 51-Fixed seat; 511-Limit slide groove; 512-Limit guide rail; 52-Slide seat; 521-Limit slide plate; 522-Mounting plate; 53-Lateral adjustment device; 531-First adjustment component; 5311-Push plate; 5312-Drive screw; 5313-Drive lead screw; 5314-Transmission gear sleeve; 5315-Power shaft; 532-Second adjustment component; 54-Longitudinal adjustment device; 541-Third adjustment component; 542-Fourth adjustment component; 55-Driver; 551-Drive gear plate; 552-Drive shaft; 553-Adjusting sleeve. Detailed Implementation

[0022] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0023] This invention discloses an in-flight syncope early warning and protection system based on multimodal data, which solves the technical problems of low accuracy of real-time monitoring of pilot head physiological parameters in actual flight environments and low reliability of monitoring pilot hypoxia and syncope in existing methods.

[0024] One specific embodiment of the present invention, such as Figure 1 As shown, a mid-air syncope early warning and protection system based on multimodal data is disclosed, including a multimodal physiological sensing module and a main control module; The multimodal physiological sensing module is used to acquire multimodal data, including physiological parameter data and motion data, of civil aviation pilots during flight in real time; The main control module 4 is used to receive the physiological parameter data and motion data and preprocess them; based on the preprocessed physiological parameter data and motion data, it determines whether the pilot has experienced in-flight syncope; if syncope occurs, it issues a syncope warning and automatically levels the aircraft to ensure the pilot's flight safety.

[0025] The main control module 4 is installed on the rear side of the pilot's flight helmet 1 (in the inner lining of the flight helmet). Figure 3 The location is only marked in the figure, and the specific shape is not shown. The main control module 4 is connected to the multimodal physiological sensing module and includes a data processing chip (not shown in the figure), a storage module (not shown in the figure), a Bluetooth communication device (not shown in the figure) and a battery. The data processing chip is used to preprocess the physiological parameter data and motion data, and to execute a head physiological monitoring method for judging the pilot's in-flight hypoxia and syncope based on the preprocessed physiological parameter data and motion data, thereby determining whether the pilot has experienced in-flight syncope. The storage module is used to store the physiological parameter data and motion data, as well as the result of whether fainting is determined by the data processing module based on the preprocessed physiological parameter data and motion data. The Bluetooth communication device is used to send the preprocessed physiological parameter data and motion data, as well as the corresponding result of whether or not the patient has fainted, to the airborne terminal. The battery is used to power the multimodal physiological sensing module and the main control module 4.

[0026] The multimodal physiological sensing module includes a PPG sensor 2 and a motion sensor 3; For example, the present invention starts with the flight helmet worn by pilots of small civil aircraft during flight, in which the PPG sensor 2 (Photoplethysmography Sensor) and motion sensor 3 are embedded in the liner of the pilot's flight helmet 1, as shown. Figure 3 As shown, a PPG sensor 2 is placed on the pilot's forehead, a motion sensor 3 is placed on the top of the pilot's flight helmet, and a main control module 4 is integrated on the back of the pilot's flight helmet; all three are covered by the liner of the flight helmet. Figure 3 In the diagram, both PPG sensor 2 and motion sensor 3 indicate their positions inside the flight cap, but their specific shapes are not shown.

[0027] The flight cap is a soft flight cap that fully covers the ears without a hard shell, completely enclosing both ears to secure the pilot's aviation headset and intercom; the PPG sensor 2 and motion sensor 3 used in this invention are integrated into the flight cap; Typical application scenarios: training for small unpressurized fixed-wing aircraft, private flights, agricultural and forestry spraying, and basic training for light helicopters.

[0028] The PPG sensor 2 is installed on the inner forehead area of ​​the pilot's flight helmet and is closely fitted to the wearer's forehead skin via an adjustment mechanism 5, for real-time acquisition of the pilot's physiological parameter data; wherein, the physiological parameter data includes Values ​​and PI (Perfusion Index); The PI value is a real-time hemodynamic parameter that reflects the blood flow status of peripheral microvessels. A higher value indicates good local arterial dilation and perfusion; a lower value suggests vasoconstriction or insufficient perfusion.

[0029] The motion sensor 3 is installed inside the pilot's flight cap at the center of the top of the head to acquire the pilot's motion data in real time; wherein, the motion data includes head motion data, pitch angle, yaw angle, and roll angle; the head motion data includes a time window. Number of nods Nodding frequency Angle of looking down and the percentage of people looking down Among them, time window Nodding frequency within and the percentage of people looking down based on , Calculated.

[0030] The PPG sensor 2, motion sensor 3, and main control module 4 used in the method of this invention are integrated into the pilot's flight cap, or into the headset of a civil aviation pilot; or into any other device worn on the pilot's head, such as a headband or glasses, as long as the device can keep the PPG sensor 2 in contact with the wearer's forehead skin.

[0031] The PPG sensor 2, motion sensor 3 and main control module 4 used in the method of this invention are directly integrated into the driver's existing head-mounted work equipment, without adding any extra burden to the wearer, and achieving "non-interference, long-term, and imperceptible monitoring".

[0032] like Figure 2 and 3 As shown, PPG sensor 2 is installed on the inner forehead area of ​​the pilot's flight cap 1, and can fit closely to the wearer's forehead to collect the wearer's physiological parameter data in real time; motion sensor 3 is installed in the middle inner area of ​​the pilot's flight cap 1, and is opposite to the center area of ​​the wearer's head to collect the wearer's head movement data in real time; main control module 4 is installed on the rear side of the pilot's flight cap 1 and is connected to PPG sensor 2 and motion sensor 3; main control module 4 is also connected to an airborne terminal (such as a laptop or tablet computer) and sends the physiological parameter data and motion data collected by PPG sensor 2 and motion sensor 3 to the airborne terminal; this realizes the synchronous and high-precision collection of the pilot's physiological parameter data and motion data, providing reliable data support for subsequent assessment of the pilot's physiological stress response under extreme conditions such as high overload, hypoxia, and syncope.

[0033] like Figure 1 As shown, the system also includes an airborne terminal and a civil aviation ground control tower; The airborne terminal is used to, based on the received result of whether the pilot has fainted, issue a warning voice broadcast if fainting occurs, and automatically level the aircraft; and to send the physiological parameter data and motion data, the preprocessed physiological parameter data and motion data, the result of whether fainting occurs, and the result of whether the aircraft has automatically leveled the aircraft to the civil aviation ground control via a data link. The civil aviation ground control tower is used to display the pilot's physiological parameter trends and motion status trends in real time based on the received data. At the same time, it classifies and stores the received data to provide data support for subsequent flight safety assessments and incident tracing. If the pilot experiences in-flight fainting, it sends an emergency command to the airborne terminal through the data link.

[0034] For example, emergency instructions may include: Confirm that the onboard automatic leveling system has been activated and is maintaining the current stable flight attitude; Notify fellow passengers (if any) to immediately enter emergency cooperation mode and assist in monitoring aircraft parameters; Continuously and frequently transmit the pilot's real-time physiological data and aircraft dynamic parameters back to the ground civil aviation dispatch center; After the pilot regains consciousness, guide him to gradually confirm his physical condition through voice interaction, and then slowly adjust the flight attitude to a safe cruise mode. If the pilot does not regain consciousness after a preset time, the emergency command instructs the onboard system to switch to fully automatic mode and plan the optimal route to the nearest alternate airport. These emergency commands are transmitted to the airborne terminal in real time via data link and executed in conjunction with the system's built-in emergency logic to ensure a closed-loop response between the ground civil aviation dispatch center and the airborne system in the event of a fainting incident, minimizing the risk of an accident.

[0035] Because of individual differences in hairline among wearers, hair can affect the fit between the PPG sensor 2 and the forehead skin when wearing the flight cap, resulting in poor fit and consequently, low quality of cerebral oxygen saturation data acquisition. To address this issue, such as... Figure 4 As shown, the pilot's flight helmet 1 also includes an adjustment mechanism 5. The adjustment mechanism 5 is used to adjust the fit position of the PPG sensor 2 and the pilot's forehead according to the wearer's hairline position, thereby ensuring that the PPG sensor 2 can fit tightly against the wearer's forehead skin, avoiding interference from hair in signal acquisition, thereby improving the photoelectric signal-to-noise ratio of the PPG sensor 2, and thus improving the quality of the cerebral blood oxygen saturation data acquired by the PPG sensor 2, so as to reduce the error rate of subsequent data processing and ensure the accuracy of monitoring results.

[0036] The adjustment mechanism 5 is installed on the forehead area of ​​the pilot's flight cap 1 and is used to adjust the fit position of the PPG sensor 2 with the pilot's forehead based on the wearer's hairline position to ensure that the PPG sensor 2 can fit in close contact with the wearer's forehead skin. The adjustment mechanism 5 includes a fixed base 51, a sliding base 52, a lateral adjustment device 53, a longitudinal adjustment device 54, and a driver 55; The mounting base 51 is installed on the forehead area of ​​the pilot's flight helmet; the slide is mounted on the mounting base 51 and can slide on the mounting base 51, and the slide 52 is used to install the PPG sensor 2; The lateral adjustment device 53 is used to control the slide block 52 to move laterally in the plane; The longitudinal adjustment device 54 is used to control the slide block 52 to move longitudinally in the plane; The driver 55 is used to control the operation of the lateral adjustment device 53 and the longitudinal adjustment device 54 respectively.

[0037] like Figure 4 and Figure 5 As shown, the adjustment mechanism 5 includes a fixed base 51, a slide 52, a lateral adjustment device 53, a longitudinal adjustment device 54, and a driver 55. The fixed base 51 is installed on the inner forehead area of ​​the flight cap body 1. The slide 52 is installed on the fixed base 51 and can slide on the fixed base 51. The slide 52 is used to install the PPG sensor 2. The lateral adjustment device 53 is used to control the slide 52 to move laterally in the plane (i.e., move along the X-axis). The longitudinal adjustment device 54 is used to control the slide 52 to move longitudinally in the plane (i.e., move along the Y-axis). The driver 55 is used to control the operation of the lateral adjustment device 53 and the longitudinal adjustment device 54 respectively. Specifically, the actuator 55 controls the movement of the lateral adjustment device 53 and the longitudinal adjustment device 54 respectively, thereby enabling the slide 52 to move precisely in the horizontal or vertical direction within the plane. This allows for the micro-adjustment of the PPG sensor 2 within the plane, thus adapting the PPG sensor 2 to the hairline position of different wearers. This ensures that the PPG sensor 2 can fit closely to the wearer's forehead skin, avoiding interference from hair in signal acquisition. Consequently, the photoelectric signal-to-noise ratio of the PPG sensor 2 is improved, thereby enhancing the quality of physiological parameter data acquired by the PPG sensor 2. This reduces the error rate of subsequent data processing and ensures the accuracy of the monitoring results.

[0038] The fixed base 51 is provided with a limiting slide groove 511, which is used to install the slide 52 and make the slide 52 only slide within the plane of the fixed base 51, and not move perpendicular to the plane, thereby reducing the bounce of the slide 52 and ensuring the stability of the slide 52. This ensures that the PPG sensor 2 remains in continuous and stable contact with the wearer's forehead skin, thereby improving the quality of physiological parameter data collected by the PPG sensor 2.

[0039] like Figure 6As shown, the slide 52 includes a limiting slide plate 521, a connecting column, and a mounting plate 522. The connecting column is used to fix the limiting slide plate 521 and the mounting plate 522, and the mounting plate 522 is used to mount the PPG sensor 2. The limiting slide plate 521 is installed in the limiting slide groove 511 and can slide smoothly in the limiting slide groove 511 along the X-axis or Y-axis direction, but cannot move in the direction perpendicular to the plane of the fixed seat 51.

[0040] The fixed base 51 is also equipped with a limit guide rail 512.

[0041] The lateral adjustment device 53 includes a first adjustment component 531 and a second adjustment component 532. The first adjustment component 531 and the second adjustment component 532 are disposed on both sides of the slide 52. The first adjustment component 531 and the second adjustment component 532 are also connected to the driver 55. The driver 55 can synchronously drive the first adjustment component 531 and the second adjustment component 532 to move laterally to one side (that is, move along the positive direction of the X-axis), thereby realizing that the first adjustment component 531 controls the slide 52 to move laterally to the other side (that is, move along the negative direction of the X-axis), and the second adjustment component 532 controls the slide 52 to move laterally to the other side (that is, move along the negative direction of the X-axis). The two work together to realize the lateral micro-adjustment of the slide 52 on the fixed base 51, thereby realizing the lateral micro-adjustment of the PPG sensor 2 in the plane.

[0042] The first control component 531 and the second control component 532 are symmetrically arranged on both sides of the slide block 52, which can effectively counteract the skew torque caused by unilateral drive. This ensures that the slide block 52 slides smoothly in the limiting slide groove 511 and applies a balanced support force to the slide block 52, reducing jamming or displacement deviation caused by uneven force, and further improving the positioning accuracy and long-term stability of the PPG sensor 2.

[0043] like Figure 6 and Figure 7As shown, the first control component 531 includes a push plate 5311, a drive screw 5312, a drive lead screw 5313, a transmission gear sleeve 5314, and a power shaft 5315. The push plate 5311 is mounted on the limiting guide rail 512 and can slide along the limiting guide rail 512. The drive screw 5312, drive lead screw 5313, transmission gear sleeve 5314, and power shaft 5315 are all mounted on the fixed base 51 and are rotatably connected to the fixed base 51. One end of the drive screw 5312 is threadedly connected to the push plate 5311, and the other end is provided with a first transmission gear, which meshes with one end of the drive lead screw 5313 through the first gear. The other end of the drive lead screw 5313 is provided with a second transmission gear, which meshes with the transmission gear sleeve 5314 through the second transmission gear. One end of the power shaft 5315 is connected to the driver 55, and the other end... The end is equipped with a third transmission gear, which meshes with the transmission sleeve 5314. This enables the drive shaft 5315 to rotate via the driver 55, which in turn drives the transmission sleeve 5314 to rotate via the third transmission gear, which then drives the drive screw 5313 to rotate via the second transmission gear. Subsequently, the drive screw 5312 is driven to rotate via the first transmission gear, ultimately converting the rotational motion into linear displacement of the push plate 5311 along the limit guide rail 512. This allows for precise micro-adjustment of the slide 52 along the positive X-axis, thereby enabling micro-adjustment of the PPG sensor 2 in the positive X-axis direction. Furthermore, the first transmission gear and the drive screw 5313 constitute a worm gear mechanism with self-locking characteristics, effectively preventing the push plate 5311 from retracting due to external force when power is cut off or the driver stops, ensuring the long-term reliability of the PPG sensor 2's positioning.

[0044] The structure of the second control component 532 is the same as that of the first control component 531. The push plate 5311 of the second control component 532 can be moved synchronously by the driver 55.

[0045] The longitudinal adjustment device 54 includes a third adjustment component 541 and a fourth adjustment component 542, which are disposed on both sides of the slide 52. Both the third adjustment component 541 and the fourth adjustment component 542 are connected to the driver 55. The driver 55 can synchronously drive the third adjustment component 541 and the fourth adjustment component 542 to move longitudinally to one side (i.e., move along the positive direction of the Y-axis), and the fourth adjustment component 542 controls the slide 52 to move longitudinally to the other side (i.e., move along the negative direction of the Y-axis). The two work together to achieve lateral micro-adjustment of the slide 52 on the fixed base 51, thereby achieving lateral micro-adjustment of the PPG sensor 2 in the plane.

[0046] The third control component 541 and the fourth control component 542 are symmetrically arranged on both sides of the slide block 52, which can effectively counteract the skew torque caused by unilateral drive. This ensures that the slide block 52 slides smoothly in the limiting slide groove 511 and applies a balanced support force to the slide block 52, reducing jamming or displacement deviation caused by uneven force, and further improving the positioning accuracy and long-term stability of the PPG sensor 2.

[0047] The structure of the third control component 541 is the same as that of the fourth control component 542, and the structure of the third control component 541 is also the same as that of the first control component 531.

[0048] The mounting plate 522 has an elastic buffer pad (not shown in the figure) on its surface, which can further ensure that the PPG sensor 2 is in close contact with the forehead skin after the slide 52 is accurately positioned, thereby further ensuring the contact stability between the PPG sensor 2 and the skin, effectively reducing the interference of motion artifacts on signal acquisition, and further improving the quality of cerebral blood oxygen saturation data acquisition.

[0049] Elastic cushioning pads can be made of air or silicone.

[0050] The driver 55 includes a drive gear 551, a drive shaft 552, and an adjusting sleeve 553. The drive gear 551 is fixedly mounted on the drive shaft 552, and the drive gear 551 can be synchronously connected to the power shaft 5315 in the first control assembly 531 and the second control assembly 532, or synchronously connected to the power shaft 5315 in the third control assembly 541 and the fourth control assembly 542. The drive shaft 552 passes through the adjusting sleeve 553 and is rotatably connected to the adjusting sleeve 553. The adjusting sleeve 553 is mounted on the fixed seat 51 and is threadedly connected to the fixed seat 51, thereby adjusting the axial position of the sleeve 553, and thus adjusting the drive gear 551. Specifically, by rotating the drive shaft 552, the drive gear 551 is driven to rotate, thereby enabling precise micro-adjustment of the PPG sensor 2 in the X and Y axes. By rotating the adjusting sleeve 553, the axial displacement of the drive gear 551 along the drive shaft 552 can be adjusted to switch the drive gear 551 to be synchronously connected with the power shaft 5315 in the first and second control components 531 and 532, or synchronously connected with the power shaft 5315 in the third and fourth control components 541 and 542, or not connected to any power shaft 5315, thereby achieving independent, precise, and interference-free adjustment of the PPG sensor 2 in the X and Y axes.

[0051] like Figure 8 As shown, the method for monitoring the head physiology of the pilot to determine in-flight hypoxia and syncope includes the following steps: Step S1, based on the real-time acquired data... The value and the predetermined driver's The baseline value is used to obtain the driver's blood oxygen decay ratio; based on the motion data and the pre-trained blood oxygen prediction model, a prediction result is obtained as to whether the driver is hypoxic. Step S2: Based on the blood oxygen attenuation ratio and the real-time acquired... The system determines whether the pilot is suspected of hypoxia; if suspected hypoxia, it determines whether the pilot has experienced in-flight hypoxia based on the predicted hypoxia result; if in-flight hypoxia occurs, it determines the pilot's condition based on the PI value and the PI attenuation coefficient obtained from the pre-determined PI baseline value of the pilot; if the pilot's condition is suspected of fainting, it determines whether the pilot has fainted based on the real-time acquired motion data, and if fainting occurs, the aircraft automatically levels off.

[0052] Prior to step S1, the blood oxygen saturation of each individual driver is predetermined. Benchmark value and PI benchmark value.

[0053] Acquire the pilot's head before the flight. The average PI value over the minutes is used as the driver's PI baseline value. ; Get the driver's head minutes The average value of the driver's benchmark value .

[0054] For example, Take 5 minutes, and modify it according to specific needs in actual application.

[0055] This invention sets up a system where, after the pilot wears a flight helmet, the PPG sensor automatically records the PI value over five minutes, calculates the average, and uses this average as the PI baseline value, denoted as [reference value]. .

[0056] PI value and blood oxygen saturation are obtained using PPG sensor 2. value.

[0057] For ease of description in this invention, Recorded as After the pilot puts on the flight helmet, the PPG sensor automatically records the pilot's data for five minutes. Value, calculate the mean, and use it as The baseline value is denoted as .

[0058] The PPG sensor 2 is a device used to measure pulse waves. Its name translates to "photoplethysmography," and it offers advantages such as being non-invasive, easy to operate, and providing stable measurement results. It works by detecting the periodic changes in blood volume with the heartbeat using photoelectric methods, thereby recording the pulse wave signal. The PPG sensor 2 can simultaneously obtain cerebral oxygen saturation. value, The measurement measures changes in cerebral blood oxygenation in the frontal cortex, reflecting the balance of oxygen supply to the brain and demonstrating changes in the driver's cognitive function.

[0059] The perfusion index (PI) value is obtained using PPG sensor 2. The PI value is a real-time output hemodynamic parameter that reflects the "quality" of blood flow in peripheral microvessels. The larger the value, the better the local arterial dilation and perfusion; the smaller the value, the more likely it is vasoconstriction or insufficient perfusion. PI is a dimensionless ratio (%) and there is no absolute normal value. It needs to be dynamically observed or combined with other perfusion indicators for comprehensive judgment.

[0060] The median PI value for healthy adults is approximately 1.4 to 3.9, while the median PI value for patients in shock is approximately 0.5 to 1.3.

[0061] Meanwhile, before step S1, it is necessary to acquire the physiological parameter data and motion data of the pilot's head in real time during the pilot's flight.

[0062] The physiological parameter data includes Value and PI value; The motion data includes head motion data, pitch angle, yaw angle, and roll angle; The head movement data includes time windows. Number of nods Nodding frequency Angle of looking down and the percentage of people looking down Among them, time window Nodding frequency within and the percentage of people looking down based on , Calculated.

[0063] The driver's head movement data is obtained using motion sensor 3.

[0064] Motion sensor 3 includes a three-axis gyroscope and a three-axis accelerometer to measure the rotation angles of the human body along three axes, including pitch angle. Yaw angle Roll angle The three axes are the sagittal axis, coronal axis, and vertical axis along the human body.

[0065] Motion sensor 3 obtains the driver's head movement data. Based on the head movement data, the data processing chip built into the flight cap calculates the nodding frequency and the percentage of head tilting.

[0066] Time window The number of nods within is recorded as ,based on The nodding frequency was obtained and denoted as . Previous studies have shown that when pilots experience hypoxia or extreme fatigue in the air (which is usually accompanied by cerebral hypoxia), The frequency of nodding will increase significantly, as shown in the calculation below: Formula (1) For example, the time window The value is 5s; The method of judgment is the pitch angle recorded by the motion sensor. And then back Each nod is recorded as one instance.

[0067] Under hypoxic conditions, in addition to nodding, drivers may also look down for extended periods. To describe this phenomenon, the percentage of drivers looking down was calculated. Within a time window of t, the time during which the head tilt angle is greater than 15° is recorded as... pitch angle The length of time, The calculation is as follows: Formula (2) During flight, real-time physiological parameters and motion data of the pilot's head are acquired to provide basic data for subsequent assessment of the pilot's hypoxia and syncope.

[0068] Step S1 includes steps S11-S13.

[0069] Step S11: Construct a blood oxygen prediction model to obtain a pre-trained blood oxygen prediction model.

[0070] like Figure 9 As shown, the blood oxygen prediction model includes an input layer, an LSTM temporal feature extraction layer, and a fully connected classification decision layer; the LSTM temporal feature extraction layer is used to extract temporal features; the fully connected classification decision layer makes classification decisions based on the temporal features. The pre-trained blood oxygen prediction model is obtained through the following process: Obtain physiological parameters and motion data of the heads of multiple drivers within a certain historical period to obtain corresponding... , , , and After preprocessing, the samples are used as training samples, and together with the corresponding samples labeled as hypoxic, they form a training sample set. The loss function is the binary cross-entropy loss between the prediction of hypoxia and the true label of the sample; the Adam optimizer is used for iterative training. If continuous The accuracy of the validation set in epochs is less than Or the F1 score on the validation set is greater than or equal to And the recall rate is greater than or equal to If the training stops, the well-trained blood oxygen prediction model is obtained. in, , and These are the accuracy threshold, F1 threshold, and recall threshold, respectively.

[0071] The blood oxygen prediction model balances the simplicity of capturing temporal features and making classification decisions.

[0072] Head physiological parameter data and motion data of multiple pilots within a certain historical period were acquired, with training samples derived from historical flight data of multiple civil aviation pilots; the sample data were preprocessed as follows: The pilot's historical flight data is organized into a three-dimensional input format: [number of samples, time step, number of features], where the number of samples is the number of samples in a training batch; the number of features = 5 (corresponding to...). , , , and (5 indicators) The time step needs to be determined based on the actual data collection frequency (e.g., collect data once every 10 seconds, take 30 consecutive time points as a sample, and the time step = 30).

[0073] Sample label coding: Convert the binary label of "hypoxia / non-hypoxia" into numerical labels of 0 (non-hypoxia) and 1 (hypoxia).

[0074] (1) Input layer, used to receive input time-series samples.

[0075] A time series sample The shape is (seq_len, 5); seq-len is the time step, for example, a 10-second time window; Five input features: , , , and ; The input batch is used during training and has a shape of (batch_size, seq_len, 5). The value of batch_size is determined by considering hardware conditions and dataset size. For example, in this invention, the value is 64, and it is dynamically adjusted based on training efficiency, convergence stability, and hardware load.

[0076] The time series partitioning method was adopted, and the training sample set was divided into training set, validation set and test set in a ratio of 7:2:1. The first 70% of the time series was used as training set, the middle 20% as validation set and the last 10% as test set, so as to preserve the time series correlation.

[0077] (2) LSTM temporal feature extraction layer, which processes the input temporal sample sequence time by time to capture the long-term dependency relationship and temporal dynamic pattern between temporal sample features; Output: Shaped (seq_len, h), including the encoded information for each time step, where h is the hidden state dimension of the LSTM temporal feature extraction layer, i.e., the dimension of the feature vector output at each time step.

[0078] (3) Fully connected classification decision layer, including fully connected layer, Dropout layer and output layer, is used to map the temporal features extracted by LSTM temporal feature extraction layer to the final classification decision.

[0079] Fully connected layers: There are one or more of them, with the activation function being ReLU, used to further combine high-level features. The number of neurons, m, is adjustable, for example, 32 or 64.

[0080] Dropout layer: Drops out a portion of neurons at any time during training to prevent overfitting. The ratio (r) is the probability of keeping neurons; for example, r is 0.6.

[0081] Output layer: 1 neuron, used for binary classification decision tasks; activation function Sigmoid, compresses the output to the range (0, 1), representing the predicted probability of whether the pilot experienced hypoxia during flight. .

[0082] (4) Training and optimization Loss function: Binary cross-entropy, which measures the probability of prediction. With real labels The difference between (0 or 1); The optimizer Adam is used to update all parameters of the LSTM temporal feature extraction layer and the fully connected classification decision layer in the model based on the loss gradient.

[0083] The binary cross-entropy (BCE) loss is used to output the probability value of a single neuron, where 0 ≤ 0. ≤1, The predicted probability of hypoxia is perfectly suited to the binary cross-entropy loss. The core idea of ​​the loss function is to measure the probability distribution predicted by the model. With real labels The differences are as follows: Formula (3) This loss function is stable for gradient calculation in binary classification tasks and converges quickly.

[0084] For example, regarding the training termination condition, The value is 10. Values , The value is 0.95. The value is 0.93.

[0085] Stop when any of the following conditions are met: (1) The accuracy of the validation set for 10 consecutive epochs is less than ; (2) The F1 value of the validation set is ≥0.95 and the recall rate is ≥0.93.

[0086] This step yields a pre-trained blood oxygen prediction model.

[0087] Step S12, based on the real-time acquired data... The value and the predetermined driver's The baseline value is used to obtain the driver's blood oxygen saturation ratio.

[0088] The data will be monitored in real time every 1 second. The average of the values ​​is denoted as The percentage of blood oxygen saturation is calculated as follows: Formula (4) in, This represents the percentage of blood oxygen depletion. for Benchmark value.

[0089] Step S13: Based on the motion data and the pre-trained blood oxygen prediction model, obtain the prediction result of whether the driver is hypoxic.

[0090] During flight, under hypoxic conditions, the human body will experience a decline in consciousness, and the standard deviation of the head attitude angle will decrease significantly. The standard deviation of the attitude angle of the motion sensor will be calculated separately.

[0091] The standard deviation of the pitch angle is calculated based on the pitch angle, yaw angle, and roll angle, respectively. Yaw angle standard deviation and roll angle standard deviation ; Will , , , and After preprocessing, the data is input into a pre-trained blood oxygen prediction model to obtain a prediction result of whether the driver is hypoxic.

[0092] Get in real time , , , and ,Will , , , and After timestamp alignment, the pre-trained blood oxygen prediction model is input, and the prediction result of whether the driver is hypoxic is output.

[0093] The purpose of step S1 is to calculate the blood oxygen decay ratio based on real-time acquired physiological parameters, and to obtain the prediction result of whether the pilot is hypoxic during flight by constructing a pre-trained blood oxygen prediction model and combining it with motion data.

[0094] Step S2 includes steps S21-S24.

[0095] Step S21: Based on the blood oxygen attenuation ratio , and the real-time acquisition of the The value is used to determine whether the driver is suspected of having hypoxia; like or If so, it is determined that the driver is suspected of having hypoxia; among them, This represents the percentage of blood oxygen depletion. The threshold for the proportion of blood oxygen decay. For every The real-time monitoring of the second The average value for Threshold.

[0096] For example, It is 15%. It is 55%. The value is 1 second, but in practical applications, it can be modified according to specific needs.

[0097] when 15% or At 55%, it is believed that the pilot has experienced suspected hypoxia. At this point, the result information of "suspected in-flight hypoxia" is sent to the airborne terminal. The airborne terminal sends the result information to the civil aviation ground control tower through the data link between the aircraft and the ground. Ground control personnel continue to verify and intervene in a timely manner based on the real-time situation.

[0098] Step S22: If hypoxia is suspected, then combine the hypoxia prediction results output by the blood oxygen prediction model to determine whether the pilot has experienced in-flight hypoxia. If it is determined that the pilot is suspected of hypoxia, and the prediction result of hypoxia is hypoxia, then it is determined that the pilot is experiencing in-flight hypoxia. Step S23: If it is determined that the pilot has experienced in-flight hypoxia, the PI attenuation coefficient is calculated based on the PI value and the pre-determined PI baseline value of the pilot to determine the pilot's condition.

[0099] Based on PI attenuation coefficient To determine the driver's condition.

[0100] The PI value reflects the level of blood perfusion to the head. When syncope occurs in mid-air, the PI value drops significantly.

[0101] The PI value is collected every 1 second in the air and recorded as follows. ,based on Based on the PI reference value, the PI attenuation coefficient is calculated as follows: Formula (5) in, This is the PI attenuation coefficient.

[0102] Assess the driver's condition as follows: like If so, the driver's condition is judged to be suspected of fainting; like If so, the driver's condition is determined to be extremely low infusion. like If so, the driver's condition is determined to be low perfusion; like If so, the driver's status is determined to be normal (PI). in, , and These are the threshold values ​​for the first, second, and third PI attenuation coefficients, respectively. ; For example, The values ​​are 0.6, 0.4, and 0.2.

[0103] Real-time attenuation coefficient The value is used to make a judgment, and the mapping judgment formula is used as follows: Formula (5) in, Based on attenuation coefficient The state judgment function is based on The value of is output as a predefined blood perfusion level.

[0104] Formula (5) can be used to classify the blood perfusion level of the pilot's head in the air into four categories (normal PI, low perfusion, extremely low perfusion, suspected syncope), and ground command personnel continue to verify and intervene in a timely manner according to the situation.

[0105] Step S24: If the pilot's state is suspected of fainting, then determine whether the pilot is fainting based on the real-time acquired motion data. If fainting is detected, the aircraft will automatically level off.

[0106] If the result indicates suspected syncope, further verification of the result is required.

[0107] If the driver's condition is suspected to be syncope, the determination of whether the driver is syncope is based on the pitch angle in the real-time acquired motion data.

[0108] If the absolute value of the pitch angle in the real-time acquired motion data is greater than the preset pitch angle threshold, then the driver is confirmed to have fainted.

[0109] The preset pitch angle threshold is 20°.

[0110] If at this time, the absolute value of the pitch angle obtained by the motion sensor If the system detects that the pilot has fainted and lost consciousness, causing their head to droop naturally, it sends a "confirmed fainting" message to the onboard terminal, and the aircraft enters automatic leveling mode. Once the pilot regains consciousness, they can resume normal operation and control is returned to the pilot.

[0111] If at this time, the motion sensor measures If the pilot faints, a message indicating "unconfirmed fainting" is sent to the onboard terminal, which in turn sends it to the ground civil aviation dispatch center. Ground control personnel closely monitor the pilot's condition and decide on subsequent intervention measures.

[0112] Step S2 is used to determine whether the pilot is suspected of hypoxia based on the blood oxygen decay ratio and real-time value, to confirm whether hypoxia has occurred in the air based on the hypoxia prediction result, and then to determine the pilot's status based on the PI decay coefficient. If fainting is suspected, the flight is automatically leveled based on motion data to ensure flight safety.

[0113] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0114] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A mid-air syncope early warning and protection system based on multimodal data, characterized in that, Includes a multimodal physiological sensing module and a main control module; The multimodal physiological sensing module is used to acquire multimodal data, including physiological parameter data and motion data, of civil aviation pilots during flight in real time; The main control module is used to receive the physiological parameter data and motion data and perform preprocessing. The determination of whether the pilot experienced in-flight syncope is based on preprocessed physiological parameter data and motion data. If the pilot faints, a fainting warning will be issued and the aircraft will be automatically leveled to ensure the pilot's flight safety.

2. The in-flight syncope early warning and protection system based on multimodal data according to claim 1, characterized in that, The main control module is installed on the rear side of the pilot's flight helmet and connected to the multimodal physiological sensing module, including a data processing chip, a storage module, a Bluetooth communication device, and a battery; The data processing chip is used to preprocess the physiological parameter data and motion data, and to execute a head physiological monitoring method for judging the pilot's in-flight hypoxia and syncope based on the preprocessed physiological parameter data and motion data, thereby determining whether the pilot has experienced in-flight syncope. The storage module is used to store the physiological parameter data and motion data, as well as the result of whether fainting is determined by the data processing module based on the preprocessed physiological parameter data and motion data. The Bluetooth communication device is used to send the preprocessed physiological parameter data and motion data, as well as the corresponding result of whether or not the patient has fainted, to the airborne terminal. The battery is used to power the multimodal physiological sensing module and the main control module.

3. The in-flight syncope early warning and protection system based on multimodal data according to claim 2, characterized in that, The multimodal physiological sensing module includes a PPG sensor and a motion sensor; The PPG sensor is installed on the inner forehead area of ​​the pilot's flight helmet and is adjusted to fit snugly against the wearer's forehead for real-time acquisition of the pilot's physiological parameter data; wherein, the physiological parameter data includes Value and PI value; The motion sensor is installed inside the pilot's flight helmet at the center of the top of the head to acquire the pilot's motion data in real time; wherein, the motion data includes head motion data, pitch angle, yaw angle, and roll angle; the head motion data includes a time window. Number of nods Nodding frequency Angle of looking down and the percentage of people looking down Among them, time window Nodding frequency within and the percentage of people looking down based on , Calculated.

4. The in-flight syncope early warning and protection system based on multimodal data according to claim 3, characterized in that, The system also includes an airborne terminal and a civil aviation ground control tower; The airborne terminal is used to, based on the received result of whether the pilot has fainted, issue a warning voice broadcast if fainting occurs, and automatically level the aircraft; and to send the physiological parameter data and motion data, the preprocessed physiological parameter data and motion data, the result of whether fainting occurs, and the result of whether the aircraft has automatically leveled the aircraft to the civil aviation ground control via a data link. The civil aviation ground tower is used to display the trends of the pilot's physiological parameters and motion status in real time based on the received data, and to classify and store the received data to provide data support for subsequent flight safety assessments and event retrospectives. If the pilot experiences in-flight fainting, an emergency command will be sent to the onboard terminal via the data link.

5. The in-flight syncope early warning and protection system based on multimodal data according to claim 2, characterized in that, The method for monitoring the head physiology of the pilot to determine in-flight hypoxia and syncope includes: Based on real-time acquisition The value and the predetermined driver's The baseline value is used to obtain the driver's blood oxygen decay ratio; based on the motion data and the pre-trained blood oxygen prediction model, a prediction result is obtained as to whether the driver is hypoxic. Based on the blood oxygen attenuation ratio and the real-time acquired... The value is used to determine whether the driver is suspected of having hypoxia; If hypoxia is suspected, the pilot's hypoxia status will be determined based on the predicted hypoxia results. If hypoxia occurs in the air, the driver's condition is determined by the PI attenuation coefficient obtained based on the PI value and the pre-determined PI reference value of the driver. If the pilot's condition is suspected of being unconscious, the system will determine whether the pilot is unconscious based on real-time motion data. If the pilot is unconscious, the aircraft will automatically level off.

6. The in-flight syncope early warning and protection system based on multimodal data according to claim 5, characterized in that, The standard deviation of the pitch angle is calculated based on the pitch angle, yaw angle, and roll angle, respectively. Yaw angle standard deviation and roll angle standard deviation ; Will , , , and After preprocessing, the data is input into a pre-trained blood oxygen prediction model to obtain a prediction result of whether the driver is hypoxic.

7. The in-flight syncope early warning and protection system based on multimodal data according to claim 5, characterized in that, like or If so, it is determined that the driver is suspected of having hypoxia; among them, This represents the percentage of blood oxygen depletion. The threshold for the proportion of blood oxygen decay. For every The real-time monitoring of the second The average value for Threshold.

8. The in-flight syncope early warning and protection system based on multimodal data according to claim 7, characterized in that, If it is determined that the pilot is suspected of hypoxia, and the prediction result of hypoxia is hypoxia, then it is determined that the pilot is experiencing in-flight hypoxia. Based on PI attenuation coefficient To determine the driver's condition; like If so, the driver's condition is judged to be suspected of fainting; like If so, the driver's condition is determined to be extremely low infusion. like If so, the driver's condition is determined to be low perfusion; like If so, the driver's status is determined to be normal (PI). in, , and These are the threshold values ​​for the first, second, and third PI attenuation coefficients, respectively. ; If the driver's condition is suspected to be syncope, the system will determine whether the driver is syncope based on the pitch angle in the real-time motion data. If the absolute value of the pitch angle in the real-time motion data is greater than the preset pitch angle threshold, then the driver is confirmed to be syncope.

9. The in-flight syncope early warning and protection system based on multimodal data according to claim 8, characterized in that, The adjustment mechanism is installed in the forehead area of ​​the pilot's flight cap and is used to adjust the fit position of the PPG sensor with the forehead based on the wearer's hairline position to ensure that the PPG sensor can fit in close contact with the pilot's forehead skin. The adjustment mechanism includes a fixed base, a sliding base, a lateral adjustment device, a longitudinal adjustment device, and a driver; The mounting base is installed on the forehead area of ​​the pilot's flight helmet; the slide is mounted on the mounting base and can slide on the mounting base, the slide being used to mount the PPG sensor; The lateral adjustment device is used to control the slide to move laterally in the plane; The longitudinal adjustment device is used to control the slide to move longitudinally in the plane; The driver is used to control the operation of the lateral adjustment device and the longitudinal adjustment device respectively.

10. The in-flight syncope early warning and protection system based on multimodal data according to claim 5, characterized in that, The blood oxygen prediction model includes an input layer, an LSTM temporal feature extraction layer, and a fully connected classification decision layer; the LSTM layer is used to extract temporal features; the fully connected layer makes classification decisions based on the temporal features. The pre-trained blood oxygen prediction model is obtained through the following process: Obtain physiological parameters and motion data of the heads of multiple drivers within a certain historical period to obtain corresponding... , , , and After preprocessing, the samples are used as training samples, and together with the corresponding samples labeled as hypoxic, they form a training sample set. The loss function is the binary cross-entropy loss between the prediction of hypoxia and the true label of the sample; the Adam optimizer is used for iterative training. If continuous The accuracy of the validation set for each epoch is less than [percentage missing]. Or the F1 score on the validation set is greater than or equal to And the recall rate is greater than or equal to If the training stops, the well-trained blood oxygen prediction model is obtained. in, , and These are the accuracy threshold, F1 threshold, and recall threshold, respectively.

Images