A multi-modal non-inductive monitoring scenario-based intelligent health cockpit system and method

Abstract

This invention provides a multimodal, non-intrusive monitoring-based scenario-based intelligent health cockpit system and method, belonging to the field of automotive electronics and intelligent cockpit technology. The system includes a perception layer, a health AI analysis layer, a scenario decision-making layer, and a service execution layer connected sequentially to form a closed loop. The perception layer periodically and non-intrusively collects user physiological, environmental, and scenario data; the health AI analysis layer integrates and analyzes the data through an AI model, outputting health risk levels and service strategy codes; the scenario decision-making layer generates service control commands based on the codes; and the service execution layer executes the commands to adjust the cockpit environment or perform human-machine interaction. This invention achieves a closed-loop end-to-end system from health monitoring and intelligent analysis to scenario-based proactive services and emergency rescue, improving the level of driving health and safety assurance, service continuity, and system reliability.

Description

Technical Field

[0001] This invention relates to the field of automotive electronics and smart cockpit technology, and in particular to an intelligent health monitoring and proactive service system and method that integrates biosensing, environmental perception, artificial intelligence and vehicle networking. Background Technology

[0002] In the wave of transformation of the automotive industry towards "intelligentization, connectivity, and greening," the cockpit has been upgraded from a traditional driving space to a "mobile third space," and users' demands for health protection, comfort, and emergency safety during travel are experiencing explosive growth. The penetration rate of health functions in domestic intelligent cockpits is increasing year by year, but user satisfaction is not ideal. The core pain points are concentrated in three major aspects: fragmented technology implementation, lack of service loop, and insufficient medical collaboration.

[0003] Currently, in the industry, technologies related to healthy cabins are mostly at the stage of single-point breakthroughs: some models only achieve manual detection of basic physiological indicators (such as heart rate) and lack linkage with the in-vehicle environment; a few high-end models have simple health modes (such as air conditioning adjustment), but they are not connected to medical resources and cannot cope with sudden health risks; at the same time, problems such as the difficulty in obtaining pathological data, the insufficient practicality of existing detection indicators, and the imperfect user privacy protection mechanism further restrict the commercialization of healthy cabins and the improvement of user experience.

[0004] Against this backdrop, the market urgently needs an intelligent health cabin solution that can achieve a closed-loop system across the entire chain of "seamless monitoring - intelligent analysis - scenario-based services - medical linkage - privacy protection," which can not only meet users' needs for daily health management, but also cope with emergency health conditions during travel, while taking into account technical feasibility and commercial compatibility. Summary of the Invention

[0005] In view of the technical defects and drawbacks existing in the prior art, the present invention provides a multimodal non-intrusive monitoring scenario-based intelligent health cockpit system and method to overcome the above problems or at least partially solve the above problems, the specific solution of which is as follows;

[0006] As a first aspect of the present invention, a scenario-based intelligent health cabin system with multimodal non-intrusive monitoring is provided, characterized in that it includes components sequentially connected to form a processing closed loop:

[0007] The perception layer is used to periodically collect physiological data of users inside the vehicle, data of the in-vehicle environment, and data of vehicle driving scenarios.

[0008] The health AI analysis layer is used to receive, integrate, and analyze the physiological data, environmental data, and scene data from the perception layer, and generate the user's health risk level and the corresponding service strategy code based on the analysis results.

[0009] The scenario decision layer is used to receive the health risk level and service policy code, and generate service control instructions that are adapted to the current health risk scenario based on preset mapping rules.

[0010] The service execution layer is communicatively connected to the scenario decision layer and is used to execute the service control commands to implement scenario-based health services by adjusting the vehicle cabin environment or initiating human-machine interaction.

[0011] In some embodiments, the sensing layer includes:

[0012] The driver monitoring system sensor submodule is configured with an RGB-NIR dual-light sensor to collect the driver's facial features, physiological indicators and driving behavior data;

[0013] The occupant monitoring system sensor submodule is configured to collect physiological indicators of occupants inside the vehicle based on RGB-IR sensors;

[0014] The environmental sensor submodule is configured to collect data on in-vehicle temperature, humidity, CO2 concentration, PM2.5 concentration, and negative ion concentration.

[0015] The seat sensor submodule is embedded in the vehicle seat and is configured to collect the user's sitting posture data and contact physiological signals.

[0016] The perception layer is configured to perform data collection according to a preset cycle, wherein the collection cycle for the driver is a first duration and the collection cycle for the occupants of the vehicle is a second duration, the second duration being longer than the first duration; and within each collection cycle, invalid data frames caused by user face obstruction, abnormal lighting, or sensor malfunction are automatically filtered out, and if no valid data is collected within a cycle, the calculation of the next collection cycle is postponed until valid data is collected.

[0017] In some embodiments, the system further includes a data preprocessing layer;

[0018] The data preprocessing layer, connected between the perception layer and the health AI analysis layer, is used to purify and securely transmit raw data from each sensor submodule in the perception layer, and includes:

[0019] The data purification module is used to filter out outliers in the original data using an adaptive threshold algorithm and convert the filtered heterogeneous data into standardized data with a predefined JSON structure.

[0020] An encrypted transmission module is used to encrypt the standardized data and transmit it to the health AI analysis layer using both HTTPS protocol and AES-256 encryption algorithm.

[0021] The standardized data includes at least processed physiological data, environmental data, scene data, data validity status identifiers, and data encryption identifiers.

[0022] In some embodiments, the health AI analysis layer includes:

[0023] A multimodal data fusion submodule is used to receive standardized data from the data preprocessing layer;

[0024] The AI ​​model engine, connected to the multimodal data fusion submodule, is a dynamic decision-making model based on a fuzzy neural network, and is configured as follows:

[0025] The standardized data includes physiological, environmental, and scenario indicators as input.

[0026] Feature extraction and nonlinear transformation of input data are performed through hidden layers containing multiple neurons;

[0027] The user’s health risk level and the corresponding service policy code are output in parallel.

[0028] The health status assessment submodule is used to determine the user's health risk status based on the health risk level output by the AI ​​model engine or in combination with preset assessment logic.

[0029] The health risk levels include low risk, medium risk, and high risk, and the preset assessment logic includes:

[0030] When all physiological indicators are within the normal threshold range and environmental data are normal, the risk level is determined to be low.

[0031] When at least one physiological indicator is within the preset mild abnormality threshold range, or when environmental data is abnormal, it is judged as medium risk;

[0032] When any physiological indicator exceeds the preset threshold for severe abnormality, or when a user actively reports discomfort, the risk is determined to be high.

[0033] In some embodiments, the scenario decision layer pre-stores mapping rules between health risk levels and service strategies;

[0034] The scenario decision layer is configured to: based on the health risk level received from the health AI analysis layer, invoke and execute the mapping rule to generate corresponding service control instructions;

[0035] The mapping rules include:

[0036] When the health risk level is low, a service control command is generated to activate the comfort experience mode. The control command for the comfort experience mode includes one or more of the following: activating seat massage, playing preset soothing audio, controlling the air conditioning system to switch to external circulation mode, and increasing the concentration of negative ion generator.

[0037] When the health risk level is medium risk, a service control instruction is generated to initiate the active intervention and consultation mode. The control instruction for the active intervention and consultation mode includes one or more of the following: adjusting in-vehicle environmental parameters, initiating a health inquiry through a voice assistant, shortening the subsequent detection cycle for the user, and recording the current abnormal health event.

[0038] When the health risk level is high, a service control command is generated to activate the emergency response and rescue mode. The control command for the emergency response and rescue mode includes one or more of the following: automatically triggering the vehicle to pull over, automatically activating the emergency call system, sending a notification to a preset emergency contact, and continuously uploading real-time health data.

[0039] In some embodiments, the system further includes a medical collaboration layer;

[0040] The medical collaboration layer is communicatively connected to the scenario decision layer and is used to provide tiered medical services based on service control instructions generated by the scenario decision layer.

[0041] The medical collaboration layer includes:

[0042] The mild health service submodule is invoked when a service control instruction corresponding to a medium-risk scenario is received. It provides one or more services, including intelligent hospital and department recommendations based on user location and symptoms, video consultations connected to a telemedicine platform, and synchronizing the current health data to the user's health record.

[0043] The emergency rescue service submodule is automatically triggered upon receiving service control commands corresponding to high-risk scenarios and executes the following rescue process:

[0044] Automatically triggers the vehicle's emergency call system, establishing a communication link with the rescue center;

[0045] Upload the rescue data package containing the user's real-time health data, the vehicle's precise location, and vehicle status information to the rescue center;

[0046] Based on the vehicle location information, the system will coordinate with the emergency department of the nearest hospital and push the user's health information in advance.

[0047] The vehicle's voice system broadcasts the progress of the rescue operation to the user.

[0048] In some embodiments, to enable commercial deployment of the system, the system adopts a modular and customizable architecture, including:

[0049] A configurable software module package, configured to provide at least two deployment options:

[0050] The first deployment scheme includes a software module package that includes functional modules for implementing the perception layer, the data preprocessing layer, and basic adjustments to the air conditioner and seats.

[0051] The second deployment scheme, in addition to including all the functional modules of the first deployment scheme, further adds functional modules to realize the health AI analysis layer, the scenario decision-making layer, the medical collaboration layer and the feedback optimization layer;

[0052] The integrated hardware carrier comprises a driver monitoring system sensor submodule integrated into the rearview mirror or center console area of ​​the perception layer, and an occupant monitoring system sensor submodule integrated into the roof area. Furthermore, the sensor submodules are integrated with the cabin interior components in their respective areas using a one-piece molding process.

[0053] In some embodiments, the system further includes a feedback optimization layer;

[0054] The feedback optimization layer is communicatively connected to the service execution layer and the health AI analysis layer, and is used to achieve continuous system optimization. It is configured to perform the following steps:

[0055] After each scenario-based health service is completed by the service execution layer, user satisfaction evaluation information for the service performed is collected through the in-vehicle human-machine interface.

[0056] The satisfaction evaluation information is associated with the health risk level that triggered this service, the service strategy code, and the corresponding scenario data, and stored as an optimization sample.

[0057] When the number of stored optimized samples reaches a preset threshold, the parameters of the AI ​​model engine in the health AI analysis layer are automatically fine-tuned based on the accumulated optimized sample set.

[0058] The goal of the parameter fine-tuning is to optimize the accuracy of the AI ​​model engine in judging health risk levels and the adaptability of its output to service strategy coding.

[0059] In some embodiments, the system further includes a privacy protection module integrated at various levels of the system to provide security protection throughout the entire data lifecycle, comprising:

[0060] The data security encryption / decryption unit provides dual encryption protection at both the hardware and software levels, including:

[0061] Hardware-level encryption is achieved through a dedicated encryption chip integrated into the vehicle's infotainment system, used to encrypt and store user health data stored locally.

[0062] Software-level encryption uses the HTTPS transport protocol and the AES-256 algorithm to encrypt data transmitted between different levels of the system.

[0063] The user access control unit is used to provide front-end data acquisition control, including:

[0064] A settings interface is provided, allowing users to independently enable or disable the data acquisition function of different sensor sub-modules in the perception layer;

[0065] A guest operation mode is provided, in which the system only provides a single health check function and does not retain any personal health data;

[0066] The data lifecycle management unit is used to execute backend data retention and cleanup strategies, including:

[0067] Local retention strategy: Automatically perform de-identification and deletion operations on user health data stored locally after the preset retention period has expired;

[0068] User-initiated cleanup: Responds to user-triggered data deletion commands and executes the operation of deleting all health data of the corresponding user stored locally and in the cloud.

[0069] As a second aspect of the present invention, a method for implementing a scenario-based intelligent health cockpit with multimodal non-intrusive monitoring is provided, applied to the system described in any of the above claims, the method comprising:

[0070] The sensory layer periodically collects physiological data of in-vehicle users, in-vehicle environmental data, and vehicle driving scenario data.

[0071] The health AI analysis layer receives and integrates the physiological data, environmental data, and scenario data, and generates the user's health risk level and the corresponding service strategy code based on the analysis results.

[0072] The scenario decision layer receives the health risk level and service policy code, and generates service control instructions adapted to the current health risk scenario based on preset mapping rules.

[0073] The service control commands are executed through the service execution layer to adjust the vehicle cabin environment or initiate human-machine interaction, thereby implementing scenario-based health services.

[0074] The present invention has the following beneficial effects:

[0075] 1. Fully automated closed loop: The system can autonomously complete the entire process from periodically and imperceptibly collecting physiological data (perception layer), to conducting health risk assessment based on multimodal data fusion (health AI analysis layer), to matching specific service strategies according to risk levels (scenario decision layer), and finally automatically executing vehicle control and human-machine interaction commands (service execution layer), without requiring manual intervention from the user, truly achieving "imperceptible" intelligent health management.

[0076] 2. Contextualized Intelligent Output: Unlike simple solutions that only provide numerical alarms, this system outputs executable service control commands. For example, instead of simply displaying "high heart rate," the system will automatically trigger a series of linked actions, such as lowering the air conditioning temperature or providing soothing voice suggestions. This shift from "information alerts" to "proactive services" significantly improves the immediacy and effectiveness of health protection, and reduces secondary risks caused by drivers being distracted when dealing with health alerts. Attached Figure Description

[0077] Figure 1 A schematic diagram of the framework of a scenario-based intelligent health cockpit system with multimodal non-contact monitoring provided in an embodiment of the present invention;

[0078] Figure 2 This is a schematic diagram of the non-intrusive health monitoring process provided in an embodiment of the present invention;

[0079] Figure 3 This is a schematic diagram of the health risk classification service process provided in an embodiment of the present invention;

[0080] Figure 4 A schematic diagram of a medical collaborative service process provided in an embodiment of the present invention;

[0081] Figure 5 This is a flowchart illustrating a scenario-based intelligent health cockpit implementation method for multimodal non-contact monitoring, as provided in an embodiment of the present invention. Detailed Implementation

[0082] To enable those skilled in the art to better understand the technical solutions of the present invention, exemplary embodiments of the present invention are described below in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0083] Where there is no conflict, the various embodiments of the present invention and the features thereof may be combined with each other.

[0084] As used herein, the term “and / or” includes any and all combinations of one or more related enumerated entries.

[0085] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used herein, the singular forms “a” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will also be understood that when the terms “comprising” and / or “made of” are used in this specification, the presence of the stated feature, integral, step, operation, element, and / or component is specified, but the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof is not excluded. Terms such as “connected” or “linked” are not limited to physical or mechanical connections but can include electrical connections, whether direct or indirect.

[0086] Unless otherwise specified, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art. It will also be understood that terms such as those defined in commonly used dictionaries should be interpreted as having the meaning consistent with their meaning in the context of the relevant art and the invention, and will not be interpreted as having an idealized or overly formal meaning unless expressly so defined herein.

[0087] In the technical solution of this invention, the collection, storage, use, processing, transmission, provision, and disclosure of user personal information all comply with relevant laws and regulations and do not violate public order and good morals. The use of user data in this technical solution follows relevant national laws and regulations (e.g., the "Information Security Technology - Personal Information Security Specification"). For example: appropriate measures are taken for personal information access control; restrictions are imposed on the display of personal information; the purpose of using personal information does not exceed the scope of direct or reasonable association; and explicit identity targeting is eliminated when using personal information to avoid precisely locating a specific individual.

[0088] To address at least one of the technical problems existing in the aforementioned related technologies, the present invention provides a scenario-based intelligent health cockpit system with multimodal non-intrusive monitoring. Figure 1 A flowchart illustrating a scenario-based intelligent health cockpit system with multimodal non-contact monitoring provided in an embodiment of the present invention includes:

[0089] The perception layer is used to periodically collect physiological data of users inside the vehicle, data of the in-vehicle environment, and data of vehicle driving scenarios.

[0090] The data preprocessing layer is used to fuse, encrypt, and standardize the data collected by the perception layer, and output structured data packets with a uniform format.

[0091] The health AI analysis layer is used to fuse and analyze the physiological data, environmental data and scenario data processed by the data preprocessing layer, and generate the user's health risk level and the corresponding service strategy code based on the analysis results.

[0092] The scenario decision layer is used to receive the health risk level and service policy code, and generate service control instructions that are adapted to the current health risk scenario based on preset mapping rules.

[0093] The service execution layer is communicatively connected to the scenario decision layer and is used to execute the service control commands to implement scenario-based health services by adjusting the vehicle cabin environment or initiating human-machine interaction.

[0094] The medical collaboration layer is used to automatically activate and establish a connection with the external medical rescue system when the health risk level reaches a preset high-risk threshold, and synchronize relevant health data and vehicle information.

[0095] The feedback optimization layer is used to collect service performance and external feedback information, and to iteratively optimize the corresponding model of the health AI analysis layer based on this.

[0096] This invention constructs a complete and adaptive intelligent health management closed loop, from automatic data collection to automatic service execution, realizing a paradigm shift from passive response to proactive intervention. This significantly improves the level of health protection and cabin intelligence during driving. For example, without any user intervention, the system can periodically capture multi-dimensional information such as the driver's heart rate, blood oxygen, and in-vehicle air quality (perception layer). Through a built-in AI model, this information is analyzed in real time to determine different levels of health status, such as "low-risk comfortable driving," "medium-risk mild fatigue," or "high-risk emergency situation" (health AI analysis layer). Based on this judgment, the system can automatically decide what service to provide, such as activating a massage or contacting roadside assistance (scenario decision layer). Finally, the system can directly control the vehicle's air conditioning, seats, audio system, and even the autonomous driving system to execute the service (service execution layer). This closed loop ensures full automation from problem detection to problem resolution, greatly reducing driving risks caused by sudden health problems of the driver.

[0097] In some embodiments, the sensing layer includes:

[0098] The driver monitoring system sensor submodule is configured with an RGB-NIR dual-light sensor to collect the driver's facial features, physiological indicators and driving behavior data;

[0099] The occupant monitoring system sensor submodule is configured to collect physiological indicators of occupants inside the vehicle based on RGB-IR sensors;

[0100] The environmental sensor submodule is configured to collect data on in-vehicle temperature, humidity, CO2 concentration, PM2.5 concentration, and negative ion concentration.

[0101] The seat sensor submodule is embedded in the vehicle seat and is configured to collect the user's sitting posture data and contact physiological signals.

[0102] The perception layer is configured to perform data collection according to a preset cycle, wherein the collection cycle for the driver is a first duration and the collection cycle for the occupants of the vehicle is a second duration, the second duration being longer than the first duration; and within each collection cycle, invalid data frames caused by user face obstruction, abnormal lighting, or sensor malfunction are automatically filtered out, and if no valid data is collected within a cycle, the calculation of the next collection cycle is postponed until valid data is collected.

[0103] The above embodiments, through the collaborative configuration of multi-source heterogeneous sensors and intelligent acquisition logic, achieve highly robust and continuous non-invasive monitoring of the physiological state of drivers and passengers in complex real-world driving environments, overcoming the limitations of single sensors being susceptible to environmental interference. For example, by configuring an RGB-NIR dual-light DMS sensor, the system can accurately capture the driver's facial features in low-light environments such as tunnels at night, ensuring continuous monitoring. By setting different detection cycles for drivers and passengers (e.g., 30 minutes for drivers and 60 minutes for passengers), the system resource allocation is optimized, focusing on drivers at higher risk. By fusing data from seat pressure sensors, when the driver is wearing a mask or sunglasses causing partial failure of the visual sensors, contact signals can be used to assist in judging heart rate and posture, ensuring uninterrupted monitoring in common scenarios such as facial occlusion, thus providing a stable and reliable data source for subsequent accurate analysis.

[0104] The following is a specific embodiment of a multimodal sensing layer provided by the present invention.

[0105] This sensing layer is not a simple superposition of single sensors, but a "sensor array" with systematic engineering design. The components complement each other in function and cross-verify data. The specific implementation process is as follows:

[0106] 1. Visual physiological monitoring subsystem: All-weather biosignal capture

[0107] Driver-specific module: Employs an RGB-NIR (near-infrared) dual-spectrum imaging system. The RGB sensor captures facial features and micro-expressions with high color fidelity under daytime or well-lit conditions; the NIR sensor (wavelength 850nm) is specifically designed for low-light or no-visible-light environments such as tunnels, nighttime, and backlighting, and can penetrate some obstructions (such as dark sunglasses) to continuously monitor changes in blood oxygen and blood flow in subcutaneous facial tissue. This module operates at 1920x1080 resolution and 30fps, integrated into the rearview mirror base or top of the center console to ensure comprehensive coverage of the driver's face. Its core outputs include heart rate, heart rate variability, and blood oxygen saturation extracted using photoplethysmography, as well as physiological parameters such as blood pressure trends, respiratory rate, and pupillary response estimated based on image analysis.

[0108] Occupant monitoring module: Employing an RGB-IR (infrared) sensor, operating at 1280x720 resolution and 25fps, it is typically embedded in the center of the roof console as a wide-angle lens, providing coverage of the entire rear seat and front passenger area. Its core function is to monitor the presence, posture (such as child safety seat status and body tilt), and basic physiological indicators of the occupants. This module is designed in a low-power sleep-wake mode, activating only when the presence of an occupant is detected.

[0109] 2. Cockpit Environment Perception Subsystem: Microclimate and Air Quality Quantification

[0110] This subsystem integrates a set of high-precision, miniaturized environmental sensors to quantify and define the "health and comfort" of the cabin in real time:

[0111] Thermal comfort unit: Integrated temperature and humidity sensors, measuring range covering extreme climates (-40℃ to 85℃), temperature accuracy ±0.5℃, humidity accuracy ±5% RH, providing accurate feedback for automatic air conditioning adjustment.

[0112] Air quality unit: Includes a CO2 sensor based on NDIR principle (range 0-5000ppm, accuracy ±50ppm) and a PM2.5 sensor based on laser scattering principle (range 0-1000μg / m³, accuracy ±10%), accurately monitoring air pollution caused by occupant respiration and external infiltration.

[0113] Active health factor unit: Equipped with a negative ion concentration sensor to monitor the working efficiency of the vehicle-mounted negative ion generator and the ambient concentration (unit: ions / cm³), quantifying its contribution to refreshing the mind and purifying the air.

[0114] 3. Contact-based sensing subsystem: anti-interference assistance and attitude fusion

[0115] Embedded fabric pressure sensor arrays and microvolt-level bioelectric sensing electrodes are used in the key support areas of the driver's and primary passenger's seats, including the seat cushions and backrests. This subsystem provides two key complementary pieces of information:

[0116] Contact-based physiological signals: When the visual sensor is temporarily malfunctioning due to the driver wearing a mask, sunglasses, or turning their head at a large angle, the electrocardiogram waveform and impedance are directly measured through contact electrodes, providing high signal-to-noise ratio heart rate and respiratory rate data that are unaffected by light or obstruction, ensuring the continuity of physiological monitoring.

[0117] Refined Posture and Pressure Distribution: The pressure array generates a "pressure cloud map" in real time, which is not only used to determine the presence of occupants and distinguish between adults / children / objects, but also to accurately identify poor posture (such as leaning to the side or curling up) and concentrated pressure points due to prolonged sitting, providing a direct basis for active seat adjustment and health reminders.

[0118] 4. The intelligent data acquisition logic and triggering strategy are as follows:

[0119] 4.1 Periodic Active Monitoring of Drivers

[0120] Baseline Cycle: The system sets a basic detection cycle of 30 minutes for the driver. This is the optimal empirical value that balances energy consumption, data continuity, and minimizing driving interference.

[0121] "High-Quality Data Window" Acquisition: Within each 30-minute cycle, the DMS and seat sensors do not operate at full power continuously. Instead, they collaboratively initiate a approximately 15-second "high-precision acquisition window." Within this window, the algorithm analyzes the video stream in real time, automatically selecting consecutive high-quality frames where the face is directly facing the subject, the lighting is uniform, and there are no obstructions. Simultaneously, it triggers the acquisition of seat contact signals and environmental data within that time slice. This logic ensures that each uploaded analysis data is of high quality, greatly improving the accuracy of subsequent AI analysis.

[0122] 4.2 Occupant monitoring as needed

[0123] Presence-based wake-up: The OMS sensor is in deep sleep mode by default to save energy. When a door open / close signal, seat pressure sensor, or cabin microwave radar detects that an occupant has entered the vehicle, the OMS will automatically wake up.

[0124] Low-frequency periodic monitoring: Physiological monitoring of occupants uses a longer cycle of 60 minutes, reflecting respect for the privacy of non-driving personnel and optimization of energy consumption. Its data collection logic is similar to that of the driver, but it focuses more on confirming basic vital signs and postural safety.

[0125] 4.3 Multi-dimensional triggering mechanism

[0126] Cycle trigger: The above 30 / 60 minute cycle is the baseline.

[0127] Scene adaptive triggering:

[0128] Long-duration driving: When the system detects that the continuous driving time exceeds 2 hours, the system can automatically temporarily increase the driver detection cycle to once every 15 minutes to enhance the monitoring of fatigue.

[0129] Environmental abrupt change: When environmental sensors detect a rapid increase in CO2 concentration above 1000 ppm or a sudden change in temperature, an additional synchronous snapshot acquisition across all sensing subsystems can be immediately triggered to assess the impact of environmental stress on the occupants.

[0130] Event Trigger: When events such as severe turbulence (airbag pre-trigger signal), sudden braking, or seatbelt pull-back are detected, high-frequency data acquisition is immediately initiated to record key physiological data before and after the event for potential health impact assessment or post-accident analysis.

[0131] User-triggered: Users can initiate an instant and comprehensive health check with one click via voice command ("Help me check my health status") or the vehicle's infotainment system interface to meet their proactive health management needs.

[0132] In some embodiments, such as Figure 2 As shown, this embodiment of the invention provides a process for non-intrusive health monitoring. The monitoring process begins with a preset detection cycle (e.g., 30 minutes for drivers). When a new detection cycle begins, the system first attempts to collect the user's physiological data through sensors in the perception layer (such as DMS). The specific process is as follows:

[0133] Data validity assessment: Within the period, the system will determine in real time whether "valid data" has been collected. Here, "valid data" refers to data frames that meet preset quality standards, such as data acquired when the user's face is not completely obscured, the lighting conditions are suitable, and the sensors are working properly.

[0134] Branch Path 1 (Valid Data Acquired): If valid data is successfully collected within the current detection cycle, the system will immediately package this data and send it to the subsequent data preprocessing layer and health AI analysis layer for processing. After completing this data processing and service response (if any), the system will immediately begin calculations and enter the next detection cycle, thereby achieving continuous, periodic, and regular monitoring.

[0135] Branch Path Two (No Valid Data Collected): If no valid data is collected by the end of the entire detection cycle (e.g., the driver wore sunglasses and a mask throughout the process, making facial features unrecognizable), the system will not immediately start the next cycle. The process will enter a "delayed" state. The system will continue to attempt to collect data until it successfully collects valid data, at which point the timer will begin for the next complete detection cycle.

[0136] The above embodiments address a key challenge in non-intrusive monitoring—monitoring interruptions caused by temporary interference. They avoid the resource waste and blind spots resulting from mechanically performing periodic checks when data is invalid, ensuring that the data used for each subsequent AI analysis is reliable. Simultaneously, the mechanism of "delaying the start of timing until valid data is collected" maximizes the continuity and coverage of health monitoring, preventing prolonged missed health status detections due to short-term obstruction or environmental changes, thereby enhancing the robustness of the entire system and user trust.

[0137] In some embodiments, the data preprocessing layer, connected between the perception layer and the health AI analysis layer, is used to purify and securely transmit raw data from each sensor submodule in the perception layer, and includes:

[0138] The data purification module is used to filter out outliers in the original data using an adaptive threshold algorithm and convert the filtered heterogeneous data into standardized data with a predefined JSON structure.

[0139] An encrypted transmission module is used to encrypt the standardized data and transmit it to the health AI analysis layer using both HTTPS protocol and AES-256 encryption algorithm.

[0140] The standardized data includes at least processed physiological data, environmental data, scene data, data validity status identifiers, and data encryption identifiers.

[0141] The above embodiments establish a standardized, clean, and secure data pipeline, transforming raw, messy, and multi-source sensor data into standardized information that can be efficiently and reliably analyzed by AI. During transmission, the security of user privacy data is ensured, laying a high-quality data foundation for upper-level intelligent decision-making. For example, heart rate data (in bpm) from the DMS and CO2 concentration data (in ppm) from environmental sensors have different formats. The preprocessing layer can uniformly convert them into a standardized JSON format and filter out obviously distorted outliers such as heart rates >200 bpm. Subsequently, these JSON data containing sensitive health information are doubly encrypted using HTTPS and AES-256 algorithms before being transmitted to the cloud or local AI analysis module. This process not only eliminates data "noise" and improves analytical accuracy but also prevents data from being stolen or tampered with during transmission through strong encryption, solving the core pain points of quality and security in health cabin data applications.

[0142] The following is a specific embodiment of the data preprocessing layer provided by this invention—a purification, unification, and secure channel for heterogeneous data. The specific implementation scheme and process are as follows.

[0143] Phase 1: Abnormal Data Filtering (Data Cleansing)

[0144] The raw data stream acquired by the perception layer inevitably contains outliers caused by momentary sensor malfunctions, strong environmental interference (such as overexposure / underexposure), user behavior (such as large head turns or facial occlusion), or electromagnetic noise. This layer employs an "adaptive threshold dynamic filtering algorithm" to clean the data stream in real time.

[0145] Algorithm principle: Instead of using a fixed threshold, the algorithm maintains a dynamic, reasonable range window for each physiological indicator, calculated based on recent historical data. For example, the heart rate threshold window is dynamically adjusted based on the user's mean and standard deviation of heart rate over the past 5 minutes.

[0146] Filtering strategy:

[0147] Physical extreme value filtering: First, apply strict safety rules to instantly filter out data points that clearly violate physiological common sense, such as: heart rate > 200 bpm or < 30 bpm; blood oxygen saturation < 70%; respiratory rate > 60 breaths / min or < 6 breaths / min. Such data is usually caused by poor sensor contact or signal abrupt changes and is discarded directly.

[0148] Mutation point detection: By calculating the difference between adjacent data points, it identifies "peaks" or "troughs" with abnormally high rates of change within a short period of time. For example, if the heart rate spikes by 50 bpm in 1 second, after excluding scenarios where the user may be engaging in strenuous exercise, this mutation point will be marked as suspicious and temporarily stored, awaiting verification in subsequent data frames. If it is an isolated point, it will be filtered out.

[0149] Consistency verification: Cross-validation is performed using multi-sensor data. For example, if the visual sensor (DMS) temporarily fails due to overexposure in strong light, and the seat contact heart rate sensor can still provide a stable signal, then the latter is used, and the former data is marked as invalid; if both signals are abnormally contradictory, the entire frame of data is marked as "low confidence" and its weight is reduced or discarded in subsequent weighted fusion.

[0150] Output: After the above steps, the original data stream is purified into a continuous sequence of "valid data samples" that conform to physiological and logical changes.

[0151] Phase Two: Data Standardization and Encapsulation (Unified Format)

[0152] The filtered valid data comes from different types of sensors (optical, pressure, ambient gas, etc.), and their data formats, units, and precision vary. This layer performs "data standardization and structured encapsulation":

[0153] Unit unification and precision standardization: Convert all numerical data into standard units agreed upon by the system (such as heart rate: bpm; blood pressure: mmHg; CO2: ppm), and standardize the numerical precision to one decimal place to balance data accuracy and processing efficiency.

[0154] Structured JSON encapsulation: Processed data is encapsulated within a predefined JSON object. This structured design offers excellent readability, scalability, and facilitates network transmission and subsequent parsing. A standardized output format example is shown below:

[0155] json

[0156] {

[0157] "Metadata": {

[0158] User type: "Driver"

[0159] "Detection Time": "2024-07-20 14:35:22.000",

[0160] "Data Batch ID": "a1b2c3d4"

[0161] },

[0162] "Vital signs data": {

[0163] Heart Rate: {"Value": 82.3, "Unit": "bpm", "Reference Range": [60.0, 100.0], "Measurement Error": 5.0, "Confidence Level": 0.96},

[0164] "Blood Oxygen Saturation": {"Value": 96.5, "Unit": "%", "Reference Range": [95.0, 100.0], "Measurement Error": 2.0, "Confidence Level": 0.98},

[0165] Heart rate variability: {"Value": 78.5, "Unit": "ms", "Reference range": [50.0, 150.0], "Confidence level": 0.92},

[0166] Blood Pressure: {"Systolic Pressure": 125.6, "Diastolic Pressure": 82.3, "Unit": "mmHg", "Reference Range":{"Systolic Pressure": [90.0, 139.0], "Diastolic Pressure": [60.0, 89.0]}, "Measurement Error": 12.0}

[0167] },

[0168] "Environmental Data": {

[0169] Air quality: {"CO2": 580.0, "unit": "ppm"}, {"PM2.5": 12.3, "unit": "μg / m³"},

[0170] Thermal comfort: {"Temperature": 24.5, "Unit": "℃"}, {"Humidity": 58.2, "Unit": "%RH"},

[0171] "Active Health Factor": {"Negative Ion Concentration": 3500.0, "Unit": "ions / cm³"}

[0172] },

[0173] "Scene Data": {

[0174] "Driving Status": {"Duration": 45.0, "Unit": "minutes"}, {"Average Speed": 85.0, "Unit": "km / h"},

[0175] Road Condition Assessment: "Smooth"

[0176] "Lighting conditions": "Sunny day_Natural light",

[0177] Driver Behavior Classification: "Normal Attentive Driving"

[0178] },

[0179] "Data Quality Identifier": {

[0180] Status: "Valid"

[0181] Integrity score: 0.99

[0182] Preprocessing Notes: Filter 1 frame of optical occlusion data

[0183] },

[0184] "Safety and Tracking Labels": {

[0185] Encryption Algorithm: AES-256-GCM

[0186] "Data Signature": "eFgH...iJkL",

[0187] Serial Number: 20240720143522001

[0188] }

[0189] }

[0190] This JSON structure not only contains data values, but also embeds rich metadata, reference ranges, confidence levels, data quality identifiers, and security signatures, providing the AI ​​analysis layer with far richer contextual information than the original numerical values, enabling AI to make more intelligent judgments (for example, when the confidence level is low, decisions can be made more cautiously).

[0191] Phase 3: Encrypted Transmission and Secure Storage (Secure Channel)

[0192] The processed structured data contains sensitive personal health information, and its transmission and storage must meet the highest security standards. This layer implements an "end-to-end, multi-layered encryption and access control" strategy:

[0193] Transport layer encryption:

[0194] All JSON data packets transmitted from the preprocessing layer upwards (to the vehicle's central computing unit or the cloud) are encrypted via HTTPS / TLS 1.3 protocol to ensure that the data is not eavesdropped on or tampered with during transmission.

[0195] Within the HTTPS channel, an additional layer of AES-256 symmetric encryption is applied to the payload of the JSON data packets. The encryption key is dynamically managed by the vehicle-mounted security chip. This achieves "double encryption," ensuring that even if the transmission channel is theoretically compromised, the data content remains under strong cryptographic protection.

[0196] Storage layer encryption:

[0197] Local storage: Health data that needs to be temporarily cached or retained is written to an encrypted database within the vehicle's infotainment system. The database file itself is fully encrypted using a key provided by a hardware security module or a trusted execution environment. Access requires strict module permission verification, and unauthorized applications (such as the entertainment system) cannot read it.

[0198] Cloud storage: Data uploaded to the cloud is encrypted on the client side before leaving the vehicle. The data stored in the cloud is encrypted and can only be decrypted by a health analysis service holding a specific key, achieving end-to-end zero-trust security.

[0199] This data preprocessing layer embodiment systematically transforms the chaotic raw sensor data into high-quality, standardized, and highly secure intelligent analysis input through "adaptive threshold filtering → structured JSON encapsulation → double-encrypted transmission".

[0200] In some embodiments, the health AI analysis layer includes:

[0201] A multimodal data fusion submodule is used to receive standardized data from the data preprocessing layer;

[0202] The AI ​​model engine, connected to the multimodal data fusion submodule, is a dynamic decision-making model based on a fuzzy neural network, and is configured as follows:

[0203] The standardized data includes physiological, environmental, and scenario indicators as input.

[0204] Feature extraction and nonlinear transformation of input data are performed through hidden layers containing multiple neurons;

[0205] The user’s health risk level and the corresponding service policy code are output in parallel.

[0206] The health status assessment submodule is used to determine the user's health risk status based on the health risk level output by the AI ​​model engine or in combination with preset assessment logic.

[0207] The health risk levels include low risk, medium risk, and high risk, and the preset assessment logic includes:

[0208] When all physiological indicators are within the normal threshold range and environmental data are normal, the risk level is determined to be low.

[0209] When at least one physiological indicator is within the preset mild abnormality threshold range, or when environmental data is abnormal, it is judged as medium risk;

[0210] When any physiological indicator exceeds the preset threshold for severe abnormality, or when a user actively reports discomfort, the risk is determined to be high.

[0211] The above embodiments, by introducing multimodal data fusion and dynamic decision-making algorithms based on fuzzy neural networks, achieve accurate and quantitative assessment of complex health risks. This allows the system to make comprehensive judgments based on multiple indicators, much like a professional doctor, rather than relying on simple threshold alarms for a single indicator. This significantly reduces false alarm rates and provides actionable service guidance. For example, when the system simultaneously detects a slightly elevated driver's heart rate (105 bpm) and a high CO2 concentration in the vehicle (800 ppm), a traditional simple threshold alarm might not trigger any warning. However, the AI ​​analysis layer of this invention, through a trained model, integrates these two indicators with other scenario data such as driving time to comprehensively determine that the user is in a "medium-risk" state of mild discomfort and outputs a service strategy code of "environmental intervention + health consultation." This multi-dimensional fusion-based assessment method is more intelligent and realistic than single-indicator alarms, making the system's intervention measures more timely and appropriate.

[0212] The following is an embodiment of the present invention providing a multimodal health risk dynamic decision-making algorithm for intelligent cockpits. The core algorithm described in this embodiment is the brain that drives the entire intelligent health cockpit system to achieve "intelligent decision-making". Compared with traditional early warning algorithms based on single threshold comparison, this algorithm, for the first time, unifies and fuses three heterogeneous and asynchronous data streams—physiological signals, environmental parameters, and driving scenarios—in the cockpit environment, simulating the comprehensive judgment process of professional medical personnel through "observation, auscultation, inquiry, and palpation". The algorithm does not output a risk level in isolation, but simultaneously outputs a structured "service strategy code" that can be directly driven and executed, realizing the leap from "perceiving the problem" to "knowing what to do". Moreover, the algorithm has the ability to fine-tune online based on actual usage data, and can adapt to the physiological baseline and driving habits of different users to achieve personalized and precise services.

[0213] The specific implementation scheme of the algorithm is as follows:

[0214] 1. Algorithm Architecture: Hybrid Intelligent Model Based on Fuzzy Neural Networks

[0215] This algorithm employs a hybrid architecture of "fuzzy logic + deep neural network". The fuzzy logic front end handles uncertainties (such as vague concepts like "mild fatigue" and "fair air quality"), transforming expert experience into rules; the deep neural network back end automatically learns complex nonlinear mapping relationships from massive, high-dimensional fused data. This architecture combines the interpretability of rules with the powerful representational capabilities of deep learning.

[0216] 2. Network Structure and Data Flow

[0217] The neural network part of the algorithm employs a carefully designed fully connected feedforward network structure, as follows:

[0218] Input layer:

[0219] Number of nodes: 15, corresponding to the standardized and normalized multimodal input vectors. The specific structure is as follows:

[0220] Physiological indicators (6 nodes): heart rate, heart rate variability, blood oxygen saturation, respiratory rate, facial skin temperature, and pupil diameter change rate.

[0221] Environmental indicator group (5 nodes): cabin CO2 concentration, PM2.5 concentration, temperature, humidity, and volatile organic compound index.

[0222] Scenario indicator group (4 nodes): continuous driving time, current road type (highway / city / rural road), time (influence of day and night rhythm), and recent steering wheel vibration variance (indirectly representing fatigue).

[0223] The input value of each node is Z-score standardized in the [0,1] interval to eliminate the influence of dimensions.

[0224] Hidden layer:

[0225] Structure: A 3-layer hidden layer is used, with 20 neurons in each layer. This "deep" structural design aims to abstract the 15-dimensional input layer by layer to extract high-order composite features such as "cardiovascular load index", "environmental stress coefficient", and "fatigue accumulation".

[0226] Activation function: All neurons in the hidden layer use the ReLU activation function to introduce nonlinearity and alleviate the gradient vanishing problem, thereby accelerating model convergence.

[0227] Regularization: During the training phase, Dropout technology (with a dropout rate of 0.2) is introduced after each hidden layer to randomly "mask" some neurons, forcing the network to learn redundant features, effectively preventing the model from overfitting on the training data and improving its generalization ability in unknown driving scenarios.

[0228] Output layer:

[0229] Nodes and functions: Two parallel and functionally decoupled output nodes are used instead of a single composite output.

[0230] Output node A (health risk level): The softmax activation function is used to output a three-dimensional probability vector, which corresponds to the probability of "low risk", "medium risk" and "high risk" respectively. The system takes the one with the highest probability as the final judgment level. This design can provide the judgment confidence and provide a reference for subsequent decision-making.

[0231] Output node B (Service Policy Encoding): Employs the Sigmoid activation function to output a multi-label encoding. Each bit or combination of bits in this encoding corresponds to a predefined, atomic service instruction in the scenario decision layer (e.g., "01" bit to start massage, "10" bit to switch to the outer loop, "11" bit to trigger ECall, etc.). Through learning, the model can directly output the optimal, combinatorial service encoding.

[0232] This parallel output structure of "one cause, multiple effects" is the key to this algorithm. It enables "risk diagnosis" and "prescription" to be completed simultaneously in a single forward propagation, achieving extremely low decision latency.

[0233] 3. Model training, validation, and dynamic iteration process

[0234] Training data construction:

[0235] To ensure the model's universality and authority, a large-scale, cross-domain training dataset was constructed, totaling over 180,000 labeled samples:

[0236] Medical-grade physiological baseline library: Contains more than 100,000 data points on multiple physiological parameters of healthy individuals at rest and during exercise from publicly available medical databases, ensuring that the model grasps the normal physiological range.

[0237] Real-world scenario database: Contains over 50,000 driving scenario data points collected in actual road tests, covering various weather conditions, road conditions, and time periods, and simultaneously records multimodal sensor data of the driver.

[0238] Clinical Collaboration Non-Pathological Database: In cooperation with tertiary hospitals, we have acquired more than 30,000 physiological change data of "non-patient" groups (such as physical examination participants) in outpatient examinations under specific stimuli (such as breath-holding or mild exercise) to model sub-health and mild abnormal states.

[0239] Model training:

[0240] Loss function: The weighted cross-entropy loss function is adopted, which imposes a higher penalty weight on misjudgments of the "high-risk" category, reflecting the high sensitivity to life safety.

[0241] Optimization process: An adaptive moment estimation optimizer was used and trained on a distributed computing cluster. After approximately 1000 training epochs, the model's performance on the independent validation set stabilized.

[0242] Performance metrics:

[0243] Overall accuracy: 92.3% in tasks that differentiate between low, medium and high risk levels.

[0244] High-risk recall rate: The recall rate (i.e. detection rate) for high-risk events is as high as 98.5%, minimizing underreporting.

[0245] Risk level misjudgment rate: The total proportion of medium-risk cases being classified as low-risk or high-risk is ≤ 5%.

[0246] Real-time iteration and personalized adaptation:

[0247] After deployment, the algorithm possesses the ability to "continuously learn," which is another major creative design of this embodiment.

[0248] Triggering mechanism: The system automatically triggers a model fine-tuning process every time it accumulates 1,000 complete closed-loop user usage data (including perception input, risk assessment, service execution and explicit / implicit user feedback).

[0249] Fine-tuning method: A transfer learning strategy is adopted. Based on the pre-trained global model, the parameters are fine-tuned with a small learning rate using newly collected localized data that has been cleaned and labeled by the feedback optimization layer.

[0250] Optimization Goal: Fine-tuning aims to better adapt the model to the baseline physiological characteristics, behavioral habits, and common local driving environments of the vehicle's primary users. For example, the model might learn that "User A's afternoon heart rate baseline is typically 5 bpm higher, which is within their normal range and should not be misjudged as medium risk."

[0251] Security Mechanism: All fine-tuning is performed in an encrypted sandbox environment and undergoes rigorous A / B testing. Only after the improvement is confirmed can it be updated to the online model, ensuring the security and stability of algorithm evolution.

[0252] The above algorithm implementation is not an isolated mathematical model, but a systems engineering project integrating expert knowledge, big data training, lightweight deployment, and online evolution capabilities. It successfully transforms the fuzzy cockpit health management problem into a quantifiable, computable, optimizable, and iterative intelligent decision-making task. It is the cornerstone of the "scenario-based intelligence" of this intelligent health cockpit system and the core barrier that distinguishes it from any existing single-point monitoring technology.

[0253] In some embodiments, the scenario decision layer pre-stores mapping rules between health risk levels and service strategies;

[0254] The scenario decision layer is configured to: based on the health risk level received from the health AI analysis layer, invoke and execute the mapping rule to generate corresponding service control instructions;

[0255] The mapping rules include:

[0256] When the health risk level is low, a service control command is generated to activate the comfort experience mode. The control command for the comfort experience mode includes one or more of the following: activating seat massage, playing preset soothing audio, controlling the air conditioning system to switch to external circulation mode, and increasing the concentration of negative ion generator.

[0257] When the health risk level is medium risk, a service control instruction is generated to initiate the active intervention and consultation mode. The control instruction for the active intervention and consultation mode includes one or more of the following: adjusting in-vehicle environmental parameters, initiating a health inquiry through a voice assistant, shortening the subsequent detection cycle for the user, and recording the current abnormal health event.

[0258] When the health risk level is high, a service control command is generated to activate the emergency response and rescue mode. The control command for the emergency response and rescue mode includes one or more of the following: automatically triggering the vehicle to pull over, automatically activating the emergency call system, sending a notification to a preset emergency contact, and continuously uploading real-time health data.

[0259] The above embodiments achieve a precise and dynamic mapping and triggering between health risk levels and specific cabin services, transforming the abstract "risk judgment" into a series of specific and executable vehicle control and interaction commands. This truly enables health services to be "scenario-based" and "systematic," providing a full spectrum of service capabilities from relief and intervention to rescue. For example, when the AI ​​analysis layer determines a risk level as "low risk," the scenario decision layer automatically maps and triggers a "comfort experience mode," generating a series of specific control commands such as "activate gentle seat massage for 15 minutes, play soothing music, and switch the air conditioning to external circulation." When a risk level is determined to be "high risk," it immediately maps to an "emergency response and rescue mode," generating a chain of commands such as "activate autonomous driving to pull over, trigger ECall emergency call, and notify emergency contacts." This hierarchical mapping mechanism ensures that the system can take the most appropriate and matched measures for different levels of health events, avoiding overreaction or underresponse.

[0260] Referring to the table below, this invention provides a method for matching risk levels and service codes based on AI analysis layer output with specific scenario-based service strategies to form a "risk level - scenario - service" mapping relationship:

[0261]

[0262] In some embodiments, such as Figure 3 As shown, it provides a flowchart of a health risk grading service, illustrating branch decision paths based on different risk levels:

[0263] Low-risk path: When the assessment result is "low-risk," the process enters the "low-risk service" stage. This stage corresponds to the mapping of the scenario decision layer and the operation of the service execution layer. The system will match and execute a service combination aimed at improving the comfort experience. For example, according to the documentation, the system may automatically execute "activate long-distance comfort mode," which specifically includes starting the seat's gentle massage, playing soothing music through the headrest speakers, switching the air conditioning to external circulation mode, and increasing the negative ion concentration, while providing friendly voice prompts.

[0264] Medium-Risk Path: When the assessment result is "medium risk," the process enters the "medium-risk service" stage. At this time, the system matches and executes a service combination mainly consisting of proactive intervention and health consultation. For example, the system will first perform "environmental intervention," such as automatically lowering the air conditioning temperature, slightly opening the windows for ventilation, and activating the CO2 purification function. Simultaneously, the voice assistant will proactively ask the user if they need health advice and provide professional relief solutions. In addition, the system will "shorten the detection cycle" (e.g., from 30 minutes to 15 minutes) for continuous tracking and record the abnormal data in the user's health record.

[0265] High-risk path: When the assessment result is "high-risk," the process enters the "high-risk service" stage. At this point, the system will trigger the highest level of emergency response. The service immediately splits into two parallel main lines:

[0266] Emergency vehicle control: The system first executes "emergency parking", which means that the automatic driving system (if supported) takes over the vehicle, controls the vehicle to decelerate smoothly, pull over to the side of the road and activate the hazard lights, and automatically opens the windows for ventilation.

[0267] Emergency Rescue: The system automatically triggers an ECall to dial emergency services and automatically uploads the user's key health data and precise vehicle location to the rescue center. Simultaneously, the system sends notification SMS messages to the user's pre-set emergency contacts. Throughout the process, the in-vehicle voice system will reassure the user and provide updates on the rescue progress.

[0268] After all branch paths have completed their service execution, the process finally "ends," completing a full service loop for a specific health risk scenario.

[0269] In some embodiments, the medical collaboration layer is communicatively connected to the scenario decision layer, and is used to provide tiered medical services based on service control instructions generated by the scenario decision layer;

[0270] The medical collaboration layer includes:

[0271] The mild health service submodule is invoked when a service control instruction corresponding to a medium-risk scenario is received. It provides one or more services, including intelligent hospital and department recommendations based on user location and symptoms, video consultations connected to a telemedicine platform, and synchronizing the current health data to the user's health record.

[0272] The emergency rescue service submodule is automatically triggered upon receiving service control commands corresponding to high-risk scenarios and executes the following rescue process:

[0273] a) Automatically trigger the vehicle's emergency call system to establish a communication link with the rescue center;

[0274] b) Upload the rescue data package containing the user's real-time health data, the vehicle's precise location, and vehicle status information to the rescue center;

[0275] c) Based on the vehicle location information, coordinate with the emergency department of the nearest hospital and push the user's health information in advance;

[0276] d) Use the vehicle's voice system to report the progress of the rescue to the user.

[0277] In the above embodiments, the vehicle cabin is deeply and hierarchically integrated with external medical resources, bridging the "last mile" from in-vehicle health management to pre-hospital emergency care and professional medical services. This upgrades the cabin from an independent health monitoring space into an intelligent hub connecting users and medical services, significantly enhancing the system's ability to respond to serious health risks. For example, in the event of a user experiencing sudden cardiac discomfort (high risk), the system not only automatically stops and calls for help, but also, through the emergency rescue service submodule, packages and sends real-time health data such as the patient's electrocardiogram and blood oxygen trends from the critical five minutes before the onset of symptoms, along with the vehicle's precise location, to the 120 emergency center and the emergency department of the nearest hospital. This allows emergency physicians to understand the patient's condition before arriving at the scene, enabling them to prepare in advance and achieving "information before the patient," thus gaining valuable "golden time" for rescue. For mild discomfort, users can directly conduct remote consultations through the in-vehicle system to obtain professional advice.

[0278] In some embodiments, such as Figure 4As shown, a flowchart of a tiered medical collaborative service is provided. This flowchart demonstrates how the intelligent health cabin system establishes different levels of linkage mechanisms with external medical resources based on risk levels, achieving seamless integration from online health consultation to offline emergency rescue. The specific implementation flow is as follows:

[0279] like Figure 4 As shown, the process begins with "medical service triggering" and immediately proceeds to the critical decision point of "health risk level determination." This step, derived from the output of the health AI analysis layer, serves as the trigger for the entire medical collaboration process. Depending on the determination result, the process is divided into two core paths:

[0280] Medium-Risk Path (Mild Medical Services): When a user is assessed as "medium-risk," the process transitions to the "Mild Medical Services" branch. This branch aims to provide users with convenient online health management and consultation support, specifically including the following optional collaborative service steps:

[0281] Hospital / Department Recommendation: Based on vehicle GPS positioning and user voice description of symptoms (such as "headache"), the system intelligently recommends suitable nearby hospitals and corresponding departments (such as neurology) through an integrated medical knowledge graph, and displays navigation routes and estimated waiting times.

[0282] Remote consultation: Users can initiate video consultations with doctors from internet hospitals via the vehicle's large screen. During the consultation, the system can authorize doctors to access abnormal health data from the trip (such as elevated blood pressure) to assist in remote diagnosis and provide medication advice. Consultation records can be archived.

[0283] Health record synchronization: All health data detected this time (including abnormal indicators) can be encrypted and synchronized to the user's personal mobile health APP or the electronic health record of the cooperating hospital after authorization, forming a continuous health record.

[0284] High-risk path (severe / emergency rescue services): When determined to be "high-risk," the process immediately enters the "severe rescue services" branch. This branch initiates the offline emergency rescue closed loop with the highest priority, and the steps are strict and rapid:

[0285] Automatic ECall Trigger: The system automatically activates the built-in emergency call system and connects to rescue centers such as 120.

[0286] Data Packaging and Upload: At the same time the call is connected, the system automatically uploads a standardized data packet containing nearly 5 minutes of high-frequency physiological data (such as electrocardiogram trend), high-precision vehicle location (GPS + Beidou dual-mode positioning), and current vehicle status (such as having pulled over) to the rescue center dispatch platform via the communication network.

[0287] Linking with emergency departments: While dispatching an ambulance, the rescue center can, based on the received vehicle location information, notify and coordinate with the nearest hospital emergency department that has the corresponding treatment capabilities, and push the patient's health data to the emergency doctor's terminal in advance, realizing "information arrives before the patient arrives".

[0288] Full-process status tracking and broadcasting: The system broadcasts the rescue progress to the user in real time through the vehicle's voice system, such as "The rescue team has set off and is 2 kilometers away from you," until the rescue team arrives at the scene and completes the handover.

[0289] In some embodiments, the system further includes a feedback optimization layer;

[0290] The feedback optimization layer is communicatively connected to the service execution layer and the health AI analysis layer, and is used to achieve continuous system optimization. It is configured to perform the following steps:

[0291] a) After each scenario-based health service is completed by the service execution layer, the user's satisfaction evaluation information for the service performed is collected through the in-vehicle human-machine interface;

[0292] b) Link the satisfaction evaluation information with the health risk level that triggered this service, the service strategy code and the corresponding scenario data, and store it as an optimization sample;

[0293] c) When the number of stored optimized samples reaches a preset threshold, the parameters of the AI ​​model engine in the health AI analysis layer are automatically fine-tuned based on the accumulated optimized sample set.

[0294] The goal of the parameter fine-tuning is to optimize the accuracy of the AI ​​model engine in judging health risk levels and the adaptability of its output to service strategy coding.

[0295] The above embodiments endow the system with continuous self-learning and optimization capabilities, enabling it to iterate its core AI model and service strategies based on real user feedback and performance data. This allows the system to become increasingly "intelligent" and "personalized" with use, effectively improving long-term user satisfaction and the system's lifecycle value. For example, after suggesting "play soothing music" to a user with a high heart rate, the system will proactively ask the user, "Did this suggestion make you feel more relaxed?" If a large number of users report that the music is not effective in this scenario, the feedback optimization layer will automatically trigger fine-tuning of the AI ​​model when such feedback data accumulates to a certain amount (e.g., 1000). In the future, when the system detects a similar high heart rate scenario again, it may adjust its suggestions to prioritize "slightly open the car window for ventilation" or "activate the fragrance to refresh the mind," making the system's service recommendations increasingly personalized and effective.

[0296] In some embodiments, the system further includes a privacy protection module integrated at various levels of the system to provide security protection throughout the entire data lifecycle, comprising:

[0297] The data security encryption / decryption unit provides dual encryption protection at both the hardware and software levels, including:

[0298] Hardware-level encryption is achieved through a dedicated encryption chip integrated into the vehicle's infotainment system, used to encrypt and store user health data stored locally.

[0299] Software-level encryption uses the HTTPS transport protocol and the AES-256 algorithm to encrypt data transmitted between different levels of the system.

[0300] The user access control unit is used to provide front-end data acquisition control, including:

[0301] A settings interface is provided, allowing users to independently enable or disable the data acquisition function of different sensor sub-modules in the perception layer;

[0302] A guest operation mode is provided, in which the system only provides a single health check function and does not retain any personal health data;

[0303] The data lifecycle management unit is used to execute backend data retention and cleanup strategies, including:

[0304] Local retention strategy: Automatically perform de-identification and deletion operations on user health data stored locally after the preset retention period has expired;

[0305] User-initiated cleanup: Responds to user-triggered data deletion commands and executes the operation of deleting all health data of the corresponding user stored locally and in the cloud.

[0306] In the above embodiments, by constructing a multi-layered privacy and security protection system that combines hardware and software to cover the entire data lifecycle, the system fully leverages the value of health data while maximizing the protection of users' personal privacy and data self-determination rights. This addresses the privacy and trust issues that users are most concerned about in the promotion of smart health cabins, making the technical solution compliant with legal requirements and more commercially feasible. For example, the system uses a dedicated encryption chip in the vehicle's infotainment system to perform hardware-level encryption on locally stored health data, making the data difficult to read even if the system is illegally disassembled. Users can manage permissions for rear-seat occupants through the vehicle's infotainment system settings, similar to managing permissions for a mobile app. The system automatically deletes locally identifiable personal health data by default after 3 months, and users can also immediately clear all cloud and local data using the "one-click deletion" function. This combination of "encrypted storage (to prevent cracking), controllable permissions (to prevent abuse), and time-limited deletion (to prevent hoarding)" gives users a strong sense of control over their own data and establishes a solid foundation of trust.

[0307] In some embodiments, to enable commercial deployment of the system, the system adopts a modular and customizable architecture, including:

[0308] A configurable software module package, configured to provide at least two deployment options:

[0309] The first deployment scheme includes a software module package that includes functional modules for implementing the perception layer, the data preprocessing layer, and basic adjustments to the air conditioner and seats.

[0310] The second deployment scheme, in addition to including all the functional modules of the first deployment scheme, further adds functional modules to realize the health AI analysis layer, the scenario decision-making layer, the medical collaboration layer and the feedback optimization layer;

[0311] The integrated hardware carrier includes a driver monitoring system sensor submodule integrated into the rearview mirror or center console area of ​​the perception layer, an occupant monitoring system sensor submodule integrated into the roof area, and the sensor submodule and the cabin interior components in the area are combined with a one-piece molding process.

[0312] The above embodiments propose a highly flexible and cost-controllable modular deployment scheme, enabling the same advanced intelligent health cockpit technology to be "tailored" and adapted to different vehicle positioning and cost budgets. This greatly enhances the market universality and commercialization capability of the technology, while reducing mass production costs and improving interior aesthetics through integrated hardware processes. For example, for economy models, a "first deployment scheme" can be adopted, deploying only a software package including DMS and OMS sensors, basic data preprocessing, and air conditioning / seat adjustment functions to achieve basic driver status monitoring and comfort adjustment at a controllable cost. For high-end models, a "second deployment scheme" can be adopted, deploying a complete software package including all AI analysis, medical collaboration, and other advanced functions. In terms of hardware, sensors are seamlessly integrated into the rearview mirror housing and roof interior, avoiding exposed "patchwork" installations, which is not only aesthetically pleasing but also reduces manufacturing costs by reducing individual assembly steps.

[0313] Based on the same inventive concept, this invention also provides a method for implementing a scenario-based intelligent health cockpit with multimodal non-contact monitoring, applicable to any of the systems described above, with reference to... Figure 5 As shown, the method includes:

[0314] S1. Periodically collect physiological data of users in the vehicle, in-vehicle environmental data, and vehicle driving scenario data through the perception layer;

[0315] S2. The health AI analysis layer receives and integrates the physiological data, environmental data and scenario data, and generates the user's health risk level and the corresponding service strategy code based on the analysis results.

[0316] S3. Receive the health risk level and service strategy code through the scenario decision layer, and generate service control instructions adapted to the current health risk scenario based on preset mapping rules.

[0317] S4. Execute the service control instructions through the service execution layer to adjust the vehicle cabin environment or initiate human-machine interaction, thereby implementing scenario-based health services.

[0318] The following is a complete embodiment provided by the present invention:

[0319] Assume a user (driver) is driving continuously for a long period (>2 hours) on a highway. Their vehicle is equipped with a "multimodal, non-intrusive monitoring scenario-based intelligent health cockpit system" based on this patented solution. This embodiment will fully demonstrate how the system seamlessly switches from "routine health maintenance" to "emergency response to acute illnesses," and ultimately completes a closed-loop "post-event optimization," vividly illustrating how the seven core levels of this invention work collaboratively. The specific process is as follows:

[0320] Phase 1: Seamless Monitoring and Routine Health Maintenance

[0321] Triggers and perceptions:

[0322] The system automatically starts a new round of monitoring according to the preset driver-specific detection cycle (e.g., 30 minutes), without requiring any manual operation from the driver, embodying the core concept of "seamless".

[0323] Multimodal sensing layer collaborative operation:

[0324] DMS sensor: Based on RGB-NIR dual-light imaging, it can still stably capture clear facial video streams of drivers even in the alternating light and dark environment of highway tunnels, for subsequent extraction of physiological indicators.

[0325] Environmental sensors detected that the CO2 concentration inside the vehicle has risen to 800 ppm (indicating slightly turbid air), while the PM2.5 concentration is normal.

[0326] Seat sensors: The pressure distribution map shows that the driver's posture is stable, but the contact sensors detect weak muscle tension signals.

[0327] All raw data was collected synchronously.

[0328] Data processing and intelligent decision-making:

[0329] Data preprocessing layer: Standardizes and cleans the heterogeneous data. For example, it filters out an invalid heart rate data frame caused by vehicle bumps, and converts heart rate (82 bpm), blood oxygen (96%), CO2 concentration (800 ppm), etc., into standard JSON format and encrypts them with AES-256.

[0330] Health AI Analysis Layer: Encrypted data packets are fed into a dynamic decision-making model based on a fuzzy neural network. The model takes 15 standardized indicators (heart rate, blood oxygen, respiratory rate, HRV, systolic blood pressure, diastolic blood pressure, temperature, humidity, CO2, PM2.5, negative ions, driving duration, road conditions, light intensity, and user status) as input. After calculation by the hidden layer, the model outputs in parallel:

[0331] Health risk level: Low risk (all core physiological indicators are within normal thresholds).

[0332] Service policy code: COMFORT_LONG_TRIP (corresponds to long-distance comfort mode).

[0333] Contextualized service execution:

[0334] Scenario decision layer: Upon receiving the low-risk and COMFORT_LONG_TRIP codes, immediately match the specific "comfort experience service combination" from the pre-set mapping rule base.

[0335] Service execution layer: Precisely executes combined instructions to form "scenario-based" services.

[0336] Tactile adjustment: The driver's seat activates a 15-minute "wave-like" gentle massage to relieve muscle fatigue.

[0337] Hearing adjustment: The headrest speaker plays alpha wave soothing music independently, and the volume automatically adapts to the ambient noise so as not to affect the driver's ability to hear the navigation.

[0338] Environmental regulation: The air conditioning system automatically switches to external circulation mode (fan speed level 1) to introduce fresh air and reduce CO2 concentration; at the same time, the negative ion generator power is increased to make the negative ion concentration in the cabin reach 5000 ions / cm³, which helps to refresh the mind.

[0339] Voice interaction: The system uses synthesized voice to provide contextual prompts: "We have detected that you have been driving continuously for 2 hours. We have activated the long-distance comfort mode for you. We suggest you take a short break at the next service area."

[0340] Phase Two: Instantaneous Detection and Accurate Assessment of Sudden Acute Risks

[0341] Sudden anomaly capture:

[0342] After running in comfort mode for about 25 minutes, the driver suddenly experienced chest pain and shortness of breath. The DMS's infrared (NIR) module accurately captured the momentary pain micro-expressions and pale skin on the driver's face, while the RGB module analyzed the rapid changes in facial blood oxygenation signals.

[0343] Almost simultaneously, the seat-contact physiological sensors detected a sudden muscle stiffness and abnormal vibration in the driver's torso.

[0344] High-speed analysis and risk escalation:

[0345] Data preprocessing layer: Marks the suddenly changing raw data (heart rate: 165 bpm, blood oxygen: 88%, shortness of breath) as "emergency data stream", processes it first and transmits it with encryption.

[0346] Health AI Analysis Layer:

[0347] The dynamic decision-making model receives this abrupt data stream. Taking into account the increased workload factor of "driving time (>2.5 hours)," the model completes recalculation within milliseconds.

[0348] Core creative judgment: The model does not alarm simply because "heart rate > 120 bpm", but comprehensively evaluates the combination of "extremely high heart rate + extremely low blood oxygen + respiratory disturbance + painful expression", which has a very high weight in emergency medicine models such as myocardial infarction and pulmonary embolism.

[0349] The model output immediately switches to:

[0350] Health risk level: High risk.

[0351] Phase 3: Fully automated, multi-threaded emergency response and rescue coordination

[0352] The system has been upgraded from "detection-alarm" to a fully automated closed loop of "decision-handling-rescue".

[0353] Level 1 Response: Automated vehicle control (ensuring immediate safety)

[0354] The scenario decision layer generates the highest priority "emergency vehicle control instruction package" based on the high risk and EMERGENCY_CARDIC encoding.

[0355] Service execution layer takes over vehicles:

[0356] Safe takeover: If the vehicle has L2+ level autonomous driving capabilities, the system will immediately issue a control request and smoothly take over the lateral and longitudinal control of the vehicle while ensuring safety.

[0357] Emergency pullover: Control the vehicle to slow down gradually, activate the right turn signal, automatically move into the emergency lane and come to a complete stop.

[0358] Warning and ventilation: The vehicle will automatically turn on the hazard warning lights and lower the driver's side window by 10cm to ensure ventilation in the passenger compartment.

[0359] Level 2 Response: Intelligent Medical Rescue Coordination (Activation of Professional Rescue Channels)

[0360] The "Emergency Rescue Service Submodule" of the medical collaboration layer is activated simultaneously, executing standardized rescue protocols:

[0361] One-click trigger: Automatically activates the vehicle's ECall (emergency call) system, directly connecting to the local 120 emergency center, without requiring the driver to operate their mobile phone.

[0362] Information Synchronization: Simultaneously with the establishment of a voice call, the system encrypts and uploads a structured rescue data packet to the emergency center's dispatch platform via the data network. This data packet contains:

[0363] Key vital signs: High-frequency heart rate, blood oxygen, and respiratory waveform trend charts in the 5 minutes prior to the incident, providing doctors with information on the disease progression.

[0364] Precise location: GPS + Beidou dual-mode positioning coordinates, automatically converted into a readable location description of "GXX Expressway, southbound, K123+500 meters", with an accuracy of ≤5 meters.

[0365] Vehicle and patient status: "The vehicle has automatically pulled over and the hazard lights are on. The driver appears to be experiencing acute cardiac discomfort and is unresponsive."

[0366] Pre-hospital coordination: While dispatching an ambulance, the emergency center dispatcher can use the system's location data to instantly connect with the nearest hospital's emergency department that is qualified as a chest pain center. The patient's vital signs trend chart is already displayed on a screen in the emergency room.

[0367] Family Notification: The system automatically sends a text message to the driver's pre-set emergency contact (spouse): "Your family member [relative] has experienced a sudden medical emergency at [specific location], triggering automatic rescue. An ambulance has been dispatched. Please keep your phone accessible."

[0368] Level 3 Response: Full Human-Computer Interaction and Status Maintenance

[0369] Emergency script for activating the voice interaction module in the service execution layer:

[0370] Reassure the driver in a calm and clear voice: "(Driver's name), the system has detected that you require emergency medical assistance. The vehicle has been safely brought to a stop, and rescue is on its way. Please remain calm and breathe slowly and deeply. Rescue personnel will arrive in approximately 8 minutes."

[0371] Subsequently, the system can broadcast an estimate of the rescue progress every minute to alleviate anxiety.

[0372] Phase 4: Post-event data closure and system evolution

[0373] Rescue handover: Upon arrival of the ambulance, medical personnel, using data transmitted in advance by the vehicle's onboard system, have a preliminary assessment of the patient's condition and can quickly provide on-site treatment and transfer the patient.

[0374] Privacy and Data Management:

[0375] All data from this incident (videos, physiological signals, vehicle control logs, and communication records) are stored in a vehicle-mounted encryption chip with high-strength encryption.

[0376] According to the privacy protection module's rules, this data is retained locally for three months by default for user review. After the expiration date, it will be automatically de-identified and personally identifiable information will be deleted. Users can also use the "one-click delete" function to immediately and completely delete all related data.

[0377] System self-optimization:

[0378] The data from this successful emergency response, after being anonymized, forms a valuable "high-risk handling optimization sample," containing complete time-series data and final result labels from the appearance of abnormal signals to AI judgment, vehicle control execution, and rescue coordination.

[0379] The feedback optimization layer stores the sample in the database. When such high-quality samples accumulate to a preset number (e.g., 1000), an offline fine-tuning of the dynamic decision-making model in the health AI analysis layer will be automatically triggered.

[0380] Model Evolution: Through training with a large number of real-world samples, the model will be more sensitive and accurate in recognizing patterns such as "sudden heart rate spikes accompanied by a drop in blood oxygen" in the future. It will also be more able to distinguish between "cardiac" and "strenuous exercise" scenarios, thereby continuously reducing the false alarm rate and improving response accuracy.

[0381] Example embodiments have been disclosed herein, and while specific terminology has been used, it is for illustrative purposes only and should be construed as such, and is not intended to be limiting. In some instances, it will be apparent to those skilled in the art that features, characteristics, and / or elements described in conjunction with particular embodiments may be used alone, or in combination with features, characteristics, and / or elements described in conjunction with other embodiments, unless otherwise expressly indicated. Therefore, those skilled in the art will understand that various changes in form and detail may be made without departing from the scope of the invention as set forth in the appended claims.

Claims

1. A scenario-based intelligent health cockpit system with multimodal non-intrusive monitoring, characterized in that, include: The perception layer is used to periodically collect physiological data of users inside the vehicle, data of the in-vehicle environment, and data of vehicle driving scenarios. The health AI analysis layer is used to receive, integrate, and analyze the physiological data, environmental data, and scene data from the perception layer, and generate the user's health risk level and the corresponding service strategy code based on the analysis results. The scenario decision layer is used to receive the health risk level and service policy code, and generate service control instructions that are adapted to the current health risk scenario based on preset mapping rules. The service execution layer is communicatively connected to the scenario decision layer and is used to execute the service control commands to implement scenario-based health services by adjusting the vehicle cabin environment or initiating human-machine interaction.

2. The system according to claim 1, characterized in that, The sensing layer includes: The driver monitoring system sensor submodule is configured with an RGB-NIR dual-light sensor to collect the driver's facial features, physiological indicators and driving behavior data; The occupant monitoring system sensor submodule is configured to collect physiological indicators of occupants inside the vehicle based on RGB-IR sensors; The environmental sensor submodule is configured to collect data on in-vehicle temperature, humidity, CO2 concentration, PM2.5 concentration, and negative ion concentration. The seat sensor submodule is embedded in the vehicle seat and is configured to collect the user's sitting posture data and contact physiological signals. The perception layer is configured to perform data collection according to a preset cycle, wherein the collection cycle for the driver is a first duration and the collection cycle for the occupants of the vehicle is a second duration, the second duration being longer than the first duration; and within each collection cycle, invalid data frames caused by user face obstruction, abnormal lighting, or sensor malfunction are automatically filtered out, and if no valid data is collected within a cycle, the calculation of the next collection cycle is postponed until valid data is collected.

3. The system according to claim 2, characterized in that, The system also includes a data preprocessing layer; The data preprocessing layer, connected between the perception layer and the health AI analysis layer, is used to purify and securely transmit raw data from each sensor submodule in the perception layer, and includes: The data purification module is used to filter out outliers in the original data using an adaptive threshold algorithm and convert the filtered heterogeneous data into standardized data with a predefined JSON structure. An encrypted transmission module is used to encrypt the standardized data and transmit it to the health AI analysis layer using both HTTPS protocol and AES-256 encryption algorithm. The standardized data includes at least processed physiological data, environmental data, scene data, data validity status identifiers, and data encryption identifiers.

4. The system according to claim 3, characterized in that, The health AI analysis layer includes: A multimodal data fusion submodule is used to receive standardized data from the data preprocessing layer; The AI ​​model engine, connected to the multimodal data fusion submodule, is a dynamic decision-making model based on a fuzzy neural network, and is configured as follows: The standardized data includes physiological, environmental, and scenario indicators as input. Feature extraction and nonlinear transformation of input data are performed through hidden layers containing multiple neurons; The user’s health risk level and the corresponding service policy code are output in parallel. The health status assessment submodule is used to determine the user's health risk status based on the health risk level output by the AI ​​model engine or in combination with preset assessment logic. The health risk levels include low risk, medium risk, and high risk, and the preset assessment logic includes: When all physiological indicators are within the normal threshold range and environmental data are normal, the risk level is determined to be low. When at least one physiological indicator is within the preset mild abnormality threshold range, or when environmental data is abnormal, it is judged as medium risk; When any physiological indicator exceeds the preset threshold for severe abnormality, or when a user actively reports discomfort, the risk is determined to be high.

5. The system according to claim 4, characterized in that, The scenario decision-making layer pre-stores mapping rules between health risk levels and service strategies; The scenario decision layer is configured to: based on the health risk level received from the health AI analysis layer, invoke and execute the mapping rule to generate corresponding service control instructions; The mapping rules include: When the health risk level is low, a service control command is generated to activate the comfort experience mode. The control command for the comfort experience mode includes one or more of the following: activating seat massage, playing preset soothing audio, controlling the air conditioning system to switch to external circulation mode, and increasing the concentration of negative ion generator. When the health risk level is medium risk, a service control instruction is generated to initiate the active intervention and consultation mode. The control instruction for the active intervention and consultation mode includes one or more of the following: adjusting in-vehicle environmental parameters, initiating a health inquiry through a voice assistant, shortening the subsequent detection cycle for the user, and recording the current abnormal health event. When the health risk level is high, a service control command is generated to activate the emergency response and rescue mode. The control command for the emergency response and rescue mode includes one or more of the following: automatically triggering the vehicle to pull over, automatically activating the emergency call system, sending a notification to a preset emergency contact, and continuously uploading real-time health data.

6. The system according to claim 5, characterized in that, The system also includes a medical collaboration layer; The medical collaboration layer is communicatively connected to the scenario decision layer and is used to provide tiered medical services based on service control instructions generated by the scenario decision layer. The medical collaboration layer includes: The mild health service submodule is invoked when a service control command corresponding to a medium-risk scenario is received. It provides one or more services, including intelligent hospital and department recommendations based on user location and symptoms, video consultations connected to a telemedicine platform, and synchronizing the current health data to the user's health record. The emergency rescue service submodule is automatically triggered upon receiving service control commands corresponding to high-risk scenarios and executes the following rescue process: Automatically triggers the vehicle's emergency call system, establishing a communication link with the rescue center; The rescue data package containing the user's real-time health data, the vehicle's precise location, and vehicle status information is uploaded to the rescue center. Based on the vehicle location information, the system will coordinate with the emergency department of the nearest hospital and push the user's health information in advance. The vehicle's voice system broadcasts the progress of the rescue operation to the user.

7. The system according to claim 6, characterized in that, To enable the commercial deployment of the system, the system adopts a modular and customizable architecture, including: A configurable software module package, configured to provide at least two deployment options: The first deployment scheme includes a software module package that includes functional modules for implementing the perception layer, the data preprocessing layer, and basic adjustments to the air conditioner and seats. The second deployment scheme, in addition to including all the functional modules of the first deployment scheme, further adds functional modules to realize the health AI analysis layer, the scenario decision-making layer, the medical collaboration layer and the feedback optimization layer; The integrated hardware carrier includes a driver monitoring system sensor submodule integrated into the rearview mirror or center console area of ​​the perception layer, an occupant monitoring system sensor submodule integrated into the roof area, and the sensor submodule and the cabin interior components in the area are combined with a one-piece molding process.

8. The system according to claim 1, characterized in that, The system also includes a feedback optimization layer; The feedback optimization layer is communicatively connected to the service execution layer and the health AI analysis layer, and is used to achieve continuous system optimization. It is configured to perform the following steps: After each scenario-based health service is completed by the service execution layer, user satisfaction evaluation information for the service performed is collected through the in-vehicle human-machine interface. The satisfaction evaluation information is associated with the health risk level that triggered this service, the service strategy code, and the corresponding scenario data, and stored as an optimization sample. When the number of stored optimized samples reaches a preset threshold, the parameters of the AI ​​model engine in the health AI analysis layer are automatically fine-tuned based on the accumulated optimized sample set. The goal of the parameter fine-tuning is to optimize the accuracy of the AI ​​model engine in judging health risk levels and the adaptability of its output to service strategy coding.

9. The system according to claim 1, characterized in that, The system also includes a privacy protection module, which is integrated into all levels of the system to provide security protection throughout the entire data lifecycle. This privacy protection module includes: The data security encryption / decryption unit provides dual encryption protection at both the hardware and software levels, including: Hardware-level encryption is achieved through a dedicated encryption chip integrated into the vehicle's infotainment system, used to encrypt and store user health data stored locally. Software-level encryption uses the HTTPS transport protocol and the AES-256 algorithm to encrypt data transmitted between different levels of the system. The user access control unit is used to provide front-end data acquisition control, including: A settings interface is provided, allowing users to independently enable or disable the data acquisition function of different sensor sub-modules in the perception layer; A guest operation mode is provided, in which the system only provides a single health check function and does not retain any personal health data; The data lifecycle management unit is used to execute backend data retention and cleanup strategies, including: Local retention strategy: Automatically perform de-identification and deletion operations on user health data stored locally after the preset retention period has expired; User-initiated cleanup: Responds to user-triggered data deletion commands and executes the operation of deleting all health data of the corresponding user stored locally and in the cloud.

10. A method for implementing a scenario-based intelligent health cockpit with multimodal non-contact monitoring, characterized in that, Applied to the system as described in any one of claims 1-9, the method comprises: The sensory layer periodically collects physiological data of in-vehicle users, in-vehicle environmental data, and vehicle driving scenario data. The health AI analysis layer receives and integrates the physiological data, environmental data, and scenario data, and generates the user's health risk level and the corresponding service strategy code based on the analysis results. The scenario decision layer receives the health risk level and service policy code, and generates service control instructions adapted to the current health risk scenario based on preset mapping rules. The service control commands are executed through the service execution layer to adjust the vehicle cabin environment or initiate human-machine interaction, thereby implementing scenario-based health services.

Images