Anti-carsickness method and system based on passenger carsickness feature recognition

Abstract

The invention discloses a carsickness prevention method and system based on passenger carsickness feature recognition. The method comprises the steps that S1, multi-mode signals of passengers are collected in real time; s2, performing timestamp alignment and motion artifact compensation on the acquired multi-mode signals; s3, carrying out feature extraction on the preprocessed multi-modal signal, and carrying out feature fusion on extracted feature vectors to form a comprehensive feature vector; s4, when the passenger takes the car for the first time, identifying the carsickness sensitivity grade of the passenger according to the multi-mode signal, and establishing a binding relationship between the facial features of the passenger and the carsickness sensitivity grade to form a personalized identification database; s5, inputting the comprehensive feature vector into a pre-trained machine learning model, calling parameters of the carsickness sensitivity grade dynamic adjustment machine learning model from the personalized recognition database when the passenger gets on the vehicle, and recognizing the real-time carsickness grade of the passenger; and S6, according to the identified real-time carsickness level, generating a corresponding intervention instruction according to a preset intervention strategy.

Description

Technical Field

[0001] This invention relates to the field of motion sickness identification and prevention technology, specifically to a method and system for preventing motion sickness based on the identification of occupant motion sickness characteristics. Background Technology

[0002] Motion sickness, commonly known as kinetosis, is a common physiological reaction caused by a conflict between the perception of motion and the actual movement of the vehicle, manifesting as a series of uncomfortable symptoms such as nausea, vomiting, and dizziness. With the increasing popularity of electric vehicles, their unique dynamic characteristics (such as rapid acceleration, rapid deceleration, rapid turning, and regenerative braking) have exacerbated the conflict between the vestibular system's perception and the actual motion of the vehicle, leading to a growing problem of motion sickness among passengers and seriously affecting their comfort and health.

[0003] Currently, the technology for monitoring and identifying motion sickness in passengers still has significant limitations, mainly in the following aspects:

[0004] 1. Subjective assessment methods are highly dependent on real-time objective data support. Currently widely used subjective methods such as motion sickness questionnaires (e.g., SSQ) rely on passengers' subjective descriptions and post-event reports, resulting in significant time lag. This method cannot achieve real-time, objective monitoring of motion sickness, making it difficult to provide timely and accurate data input for vehicle active intervention systems. Furthermore, individual subjective differences further affect the reliability and universality of the assessment results.

[0005] 2. Facial analysis technology has poor adaptability in dynamic in-vehicle environments. Although facial vision-based analysis technology enables non-contact monitoring, its effectiveness is easily affected by lighting conditions, occupant posture occlusion, privacy concerns, and limitations of in-vehicle computing resources. Its stability and reliability are difficult to guarantee in complex and ever-changing real-world driving environments.

[0006] 3. Skin contact devices specifically designed for mood and stress monitoring have inherent limitations. Existing wristband or finger clip-on skin contact devices are poorly suited to the natural hand movements of passengers in a vehicle (such as gripping armrests or operating mobile phones), making them prone to signal interruption or distortion due to device displacement or poor contact. Furthermore, the signal acquisition schemes and data processing algorithms of these devices are not optimized for the special conditions of unstable and high-vibration vehicles, lacking effective motion artifact recognition and removal mechanisms. Their built-in algorithm models are mostly general-purpose for mood or stress recognition, failing to model and optimize for the physiological response patterns specific to motion sickness induced by motion-perception conflict, resulting in insufficient specificity and sensitivity in motion sickness detection.

[0007] In summary, current technologies lack a solution for objectively monitoring and identifying motion sickness in passengers within a dynamic, non-laboratory in-vehicle environment. Therefore, there is an urgent need in this field for an innovative technological approach that can overcome these limitations and provide real-time, accurate information on passenger motion sickness, thereby laying the foundation for precise and personalized motion sickness prevention interventions. Summary of the Invention

[0008] The purpose of this invention is to provide a method and system for preventing motion sickness based on the identification of occupant motion sickness characteristics, so as to solve the problem that it is currently difficult to objectively monitor and identify the motion sickness status of occupants in a dynamic, non-laboratory vehicle environment.

[0009] To address the aforementioned technical problems, this invention provides a motion sickness prevention method based on passenger motion sickness feature recognition, comprising the following steps:

[0010] S1: Real-time acquisition of multimodal signals from occupants, including skin conductance signals, heart rate signals, and motion signals;

[0011] S2: Perform timestamp alignment and motion artifact compensation on the acquired multimodal signals;

[0012] S3: Extract features from the preprocessed multimodal signal and fuse the extracted feature vectors to form a comprehensive feature vector;

[0013] S4: When a passenger takes their first ride, the motion sickness sensitivity level of the passenger is identified based on multimodal signals. Then, facial recognition technology is used to establish a binding relationship between the passenger's facial features and the motion sickness sensitivity level, forming a personalized recognition database.

[0014] S5: Input the comprehensive feature vector into the pre-trained machine learning model, and dynamically adjust the parameters of the machine learning model by retrieving the motion sickness sensitivity level from the personalized identification database when the passenger gets on the vehicle, so as to identify the passenger's real-time motion sickness level.

[0015] S6: Based on the identified real-time motion sickness level, generate corresponding intervention instructions according to the preset intervention strategy.

[0016] Furthermore, step S2 specifically includes:

[0017] S21: Time-stamp align the skin conduction signal, heart rate signal, and motion signal to obtain a time-aligned sequence of skin conduction signal, heart rate signal, and motion signal.

[0018] S22: An adaptive noise cancellation algorithm is adopted to compensate for motion artifacts in the skin conductance signal and heart rate signal using motion signals, thereby eliminating noise in the skin conductance signal and heart rate signal.

[0019] Furthermore, the comprehensive feature vector is represented as:

[0020] ;

[0021] in: , These represent the mean and standard deviation of skin conductivity levels, respectively. This indicates the number of skin conductance responses per minute; Indicates the energy percentage of the target frequency band; This indicates the average heart rate; The standard deviation of all normal sinus intervals; This represents the root mean square of the difference between adjacent RR periods; Indicators representing heart rate variability; , and Let represent the dominant frequency, average power, and power spectral entropy of the motion signal, respectively, and T represent the transpose of the vector.

[0022] Furthermore, the multimodal signal identification of occupants' motion sickness sensitivity levels specifically includes:

[0023] Calculate the intensity of change in skin conductivity per unit time The calculation formula is as follows:

[0024] ;

[0025] Calculate the sensitivity coefficient of heart rate variability The calculation formula is as follows:

[0026] ;

[0027] in: This indicates the low-frequency heart rate of the occupants during dynamic driving conditions. With high heart rate The ratio, This indicates the low-frequency heart rate of the occupants at rest. With high heart rate The ratio;

[0028] Calculate exercise tolerance index The calculation formula is as follows:

[0029] ;

[0030] in: It represents the amplitude of motion perceived subjectively by the occupants or the response of their physiological systems. This represents the actual physical motion amplitude of the vehicle itself;

[0031] Based on skin conductance intensity index Heart rate variability sensitivity coefficient and stress sensitivity The fuzzy comprehensive evaluation method was used to classify occupants into three levels: low sensitivity, medium sensitivity, and high sensitivity.

[0032] Furthermore, a machine learning model is used to map the comprehensive feature vector to motion sickness levels, and the heart rate variability sensitivity coefficient is calculated. The output of the machine learning model is the probability distribution vector p for each motion sickness level, and the final motion sickness level prediction result is obtained. From the highest probability The decision, expressed as:

[0033] ;

[0034] .

[0035] Furthermore, the intervention strategy is as follows:

[0036] ;

[0037] Wherein: ϕ represents the control command vector; ϕdrive_smooth indicates a mild state, in which basic smoothness control commands are output; ϕdrive_smoother indicates a moderate state, in which enhanced smoothness control commands are output; ϕscent indicates the coordinated activation of the fragrance adjustment system, in which fragrance adjustment control commands are output; ϕdrive_ultra_smooth indicates a severe state, in which ultimate smoothness control commands are output; ϕalert indicates the simultaneous triggering of multiple alarm mechanisms, in which alarm control commands are output; and ϕsuggest indicates the activation of the intelligent suggestion system, in which intelligent suggestion control commands are output.

[0038] Secondly, the present invention provides a motion sickness prevention system based on passenger motion sickness feature recognition, including...

[0039] Sensing layer: Real-time acquisition of multimodal signals from occupants, including skin conductance signals, heart rate signals, and motion signals;

[0040] The recognition layer performs timestamp alignment and motion artifact compensation on the acquired multimodal signals; it extracts features from the preprocessed multimodal signals and fuses the extracted feature vectors to form a comprehensive feature vector; when a passenger takes their first ride, it identifies the passenger's motion sickness sensitivity level based on the multimodal signals, and then establishes a binding relationship between the passenger's facial features and motion sickness sensitivity level through facial recognition technology to form a personalized recognition database; it inputs the comprehensive feature vector into a pre-trained machine learning model, and dynamically adjusts the parameters of the machine learning model by retrieving the passenger's motion sickness sensitivity level from the personalized recognition database when the passenger boards the vehicle, thus identifying the passenger's real-time motion sickness level.

[0041] Decision-making level: Based on the identified real-time motion sickness level, generate corresponding intervention instructions according to the preset intervention strategy;

[0042] Execution layer: Controls the corresponding execution mechanism according to the intervention command.

[0043] Furthermore, the perception layer collects multimodal signals from occupants in real time through flexible fabric electrode gloves. The flexible fabric electrode gloves include: a skin conductance signal acquisition module, a heart rate signal acquisition module, a motion signal acquisition module, and a microcontroller, all of which are already installed on the glove body. The microcontroller transmits the skin conductance signals, heart rate signals, and motion signals collected by the skin conductance signal acquisition module, heart rate signal acquisition module, and motion signal acquisition module to the vehicle-mounted local processing terminal through a wireless communication module.

[0044] Furthermore, the skin electromyography (SEMG) signal acquisition module includes a skin surface electromyography sensor disposed on the glove body and the palm side of the hand, with the two electrode pads of the skin surface electromyography sensor disposed in the areas corresponding to the thenar and hypothenar muscles of the palm, respectively.

[0045] The heart rate signal module includes a photoplethysmography (PPG) detector located on the glove body and wrist area, and a PPG light source and photodetector positioned facing the wrist side.

[0046] The motion signal acquisition module includes a six-axis inertial measurement unit located on the back of the glove body.

[0047] Furthermore, when the microcontroller detects a continuous and regular low-frequency motion signal uploaded by the motion signal acquisition module, or begins to acquire a heart rate signal uploaded by the heart rate signal module, it automatically wakes up the entire system and puts it into working mode.

[0048] When the microcontroller detects that the skin conductance signal and heart rate signal have been in a stable state for more than a threshold and no vibration signal has been detected, or when it detects that the Bluetooth connection has been disconnected and the system has been in an idle state for more than a preset threshold, it enters a low-power sleep period.

[0049] The beneficial effects of this invention are as follows: By automatically collecting multimodal signals, including skin conductance signals, heart rate signals, and motion signals, as the data foundation, it overcomes the lag and subjectivity of traditional subjective questionnaires (such as SSQ) and the instability of image recognition technology, providing a real-time and objective data foundation for vehicle intervention systems; through signal alignment and motion artifact compensation, it significantly improves the signal-to-noise ratio and reliability of physiological signals in vibrating and non-stationary vehicle environments; by establishing personalized motion sickness sensitivity profiles and binding them to identity information, the system can "recognize" passengers and dynamically adjust the recognition model according to their individual physiological characteristics, thereby achieving individualized and accurate state recognition and strategy matching, greatly improving the accuracy of monitoring and intervention and the user experience. Attached Figure Description

[0050] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, use the same reference numerals to denote the same or similar parts. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0051] Figure 1 This is a flowchart of an embodiment of the motion sickness prevention method of the present invention;

[0052] Figure 2 This is a flowchart illustrating the multimodal feature extraction method according to an embodiment of the present invention.

[0053] Figure 3 This is a general framework diagram of an anti-motion sickness system according to an embodiment of the present invention;

[0054] Figure 4 This is a schematic diagram of the hardware layout on a flexible fabric electrode glove according to an embodiment of the present invention. Detailed Implementation

[0055] like Figure 1 The motion sickness prevention method based on passenger motion sickness feature recognition, as shown, includes the following steps:

[0056] S1: Real-time acquisition of multimodal signals from occupants, including skin conductance signals, heart rate signals, and motion signals;

[0057] S2: Perform timestamp alignment and motion artifact compensation on the acquired multimodal signals;

[0058] S3: Extract features from the preprocessed multimodal signal and fuse the extracted feature vectors to form a comprehensive feature vector;

[0059] S4: When a passenger takes their first ride, the motion sickness sensitivity level of the passenger is identified based on multimodal signals. Then, facial recognition technology is used to establish a binding relationship between the passenger's facial features and the motion sickness sensitivity level, forming a personalized recognition database.

[0060] S5: Input the comprehensive feature vector into the pre-trained machine learning model, and dynamically adjust the parameters of the machine learning model by retrieving the motion sickness sensitivity level from the personalized identification database when the passenger gets on the vehicle, so as to identify the passenger's real-time motion sickness level.

[0061] S6: Based on the identified real-time motion sickness level, generate corresponding intervention instructions according to the preset intervention strategy.

[0062] This invention overcomes the lag and subjectivity of traditional subjective questionnaires (such as the SSQ) and the instability of image recognition technology by automatically collecting multimodal signals, including skin conductance, heart rate, and motion signals, as a data foundation. This provides a real-time and objective data basis for vehicle intervention systems. Through signal alignment and motion artifact compensation, it significantly improves the signal-to-noise ratio and reliability of physiological signals in vibrating and unstable in-vehicle environments. By establishing personalized motion sickness sensitivity profiles and binding them to identity information, the system can "recognize" passengers and dynamically adjust the recognition model according to their individual physiological characteristics. This achieves individualized and accurate state recognition and strategy matching, greatly improving the accuracy of monitoring and intervention and the user experience.

[0063] According to one embodiment of this application, step S2 specifically includes:

[0064] S21: Time-stamp align the skin conduction signal, heart rate signal, and motion signal to obtain a time-aligned sequence of skin conduction signal, heart rate signal, and motion signal.

[0065] S22: An adaptive noise cancellation algorithm is adopted to compensate for motion artifacts in the skin conductance signal and heart rate signal using motion signals, thereby eliminating noise in the skin conductance signal and heart rate signal.

[0066] The core of the adaptive noise cancellation algorithm is to use one or more reference inputs (motion signals) to estimate and eliminate noise in the main signals (electrical skin signal and heart rate signal). The formula for compensating for motion artifacts in the electrical skin signal applied to this system is as follows:

[0067]

[0068] in: Indicates the first The original electrodermal signal at each sampling time was contaminated by motion artifacts; This represents the reference noise signal derived from six-axis inertial measurement unit (IMU) data (such as acceleration magnitude or angular velocity), which is related to motion artifacts but not to the actual skin conductance signal; Indicates length is The adaptive filter in the first The first moment Each weighting coefficient; M represents the purified electrodermal signal estimate output after ANC algorithm processing; M-1 represents the upper bound index of the order or length of the adaptive filter; more simply: M represents the total number of "weights" or "tap" in this adaptive filter, because the index of the summation symbol ∑ starts from k=0, so to represent M weights, the index k needs to go from 0 to M-1; when k=0, it corresponds to the first weight. and reference signal When k=1, it corresponds to the second weight. and a reference signal delayed by one sampling point When k = M - 1, it corresponds to the Mth weight. and the reference signal delayed by (M-1) sampling points Filter weights The least mean square algorithm is typically used for online updates, and its update rules are as follows:

[0069] ;

[0070] in: This represents the step size parameter, which controls the convergence speed and stability. and These represent the weight vector and the reference input vector, respectively.

[0071] This embodiment employs an adaptive noise cancellation (ANC) algorithm, which actively uses motion signals to estimate and eliminate artifacts in physiological signals. This fundamentally solves the core problem of signal distortion caused by motion interference in the vehicle environment, laying the foundation for high-precision feature extraction.

[0072] According to one embodiment of this application, feature extraction is the step of transforming the purified raw signal into key indicators that can characterize motion sickness. The system extracts a series of representative features in the time and frequency domains from signals of three modalities: skin conductance, cardiovascular, and kinetic signals, forming a comprehensive feature vector. Represented as:

[0073] ;

[0074] in: , These represent the mean and standard deviation of skin conductivity levels, respectively. This indicates the number of skin conductance responses per minute; Indicates the energy percentage of the target frequency band; This indicates the average heart rate; The standard deviation of all normal sinus intervals; This represents the root mean square of the difference between adjacent RR periods; Indicators representing heart rate variability; , and Let represent the dominant frequency, average power, and power spectral entropy of the motion signal, respectively, and T represent the transpose of the vector.

[0075] This comprehensive feature vector It also includes information reflecting sympathetic nerve excitability (skin conductance characteristics), autonomic nerve balance (cardiovascular characteristics), and external physical stimulation (motor characteristics), providing a comprehensive and multi-dimensional quantitative basis for judging motion sickness. The fused multimodal features enable machine learning models to learn more complex and robust patterns, thereby significantly improving the discrimination and recognition accuracy of different levels of motion sickness.

[0076] Skin conductance characteristics primarily reflect the excitability of the sympathetic nervous system. In the time domain, the mean of skin conductance levels is calculated. and standard deviation The calculation formula is:

[0077] ;

[0078] ;

[0079] in, Let N represent the i-th electroreceptor signal within the window, and N represent the number of samples within the window.

[0080] Mean skin conductivity level It reflects the baseline excitation level, while the skin conductance level reflects its fluctuations. Additionally, the number of skin conductance responses per minute can be extracted. This is a direct manifestation of rapid sympathetic nerve stress.

[0081] In the frequency domain, wavelet packet transform is performed on the SCL sequence to extract the energy percentage (P) of the target frequency band (e.g., 0.05-0.2Hz) associated with sympathetic nerve activity. band This is a stable indicator for measuring mental stress. The energy percentage of the target frequency band. The calculation formula is:

[0082] ;

[0083] in: This indicates the extraction of target frequency band energy related to sympathetic nerve activity. This represents the total energy across the entire frequency band.

[0084] The core of the heart rate feature set is the heart rate variability index. First, the mean heart rate (HR) is calculated. Then, several indicators are calculated from the RR interval sequence, including: ① the standard deviation (SDNN) of all normal sinus intervals, reflecting overall heart rate variability; ② the root mean square of the difference between adjacent RR intervals. ① It mainly reflects the activity of the parasympathetic nervous system; ③ The LF / HF ratio, that is, the ratio of low-frequency power to high-frequency power, reflects the balance between the sympathetic and parasympathetic nervous systems.

[0085] Standard deviation of all normal sinus intervals The calculation formula is:

[0086] ;

[0087] Root mean square of the difference between adjacent RR intervals The calculation formula is:

[0088] ;

[0089] in: This represents the i-th RR interval within the window. This represents the mean of all RR intervals within the window. This represents the (i+1)th RR interval within the window.

[0090] The motion feature set directly quantifies the degree of body sway. Power spectral entropy (H_spectral) is a key feature that quantifies the degree of motion disorder. The calculation formula is:

[0091] ;

[0092] in: Indicates the power spectrum at frequency The normalized energy value at that location. High entropy indicates complex and irregular motion, while low entropy indicates stable and regular motion.

[0093] According to one embodiment of this application, the multimodal signal identification of a passenger's motion sickness sensitivity level specifically includes:

[0094] ① Calculate the intensity of change in skin conductivity per unit time. The calculation formula is as follows:

[0095] ;

[0096] ② Calculate the sensitivity coefficient of heart rate variability The calculation formula is as follows:

[0097] ;

[0098] in: This indicates the low-frequency heart rate of the occupants during dynamic driving conditions. With high heart rate The ratio, This indicates the low-frequency heart rate of the occupants at rest. With high heart rate The ratio;

[0099] ③ Calculate exercise tolerance index The calculation formula is as follows:

[0100] ;

[0101] in: It represents the amplitude of motion perceived subjectively by the occupant or responded to by the physiological system; it is a quantitative value of the intensity of the occupant's physiological response. It represents the actual physical motion amplitude of the vehicle itself, which represents the intensity of the motion stimulus applied to the occupants;

[0102] ④ Based on the skin conductance intensity index Heart rate variability sensitivity coefficient and stress sensitivity The fuzzy comprehensive evaluation method was used to classify occupants into three levels: low sensitivity, medium sensitivity, and high sensitivity.

[0103] By calculating the intensity of changes in skin conductance, heart rate variability sensitivity coefficient, and exercise tolerance index of an individual per unit time, the inherent physiological differences between individuals are eliminated, and the assessment of "personal sensitivity" rather than "absolute physiological value" is truly realized.

[0104] After identifying a passenger's motion sickness sensitivity level, and with the passenger's consent, the system collects the passenger's facial information and establishes a binding relationship between the passenger's facial features and motion sickness sensitivity level through facial recognition technology, forming a personalized identification database that can be quickly retrieved.

[0105] When a passenger enters the vehicle, the system initiates a real-time facial recognition process with the passenger's consent. By calculating the Euclidean distance between the current facial features and features stored in the database, identity matching is completed within 200 milliseconds. Upon successful matching, the system immediately retrieves the corresponding individual sensitivity profile and preset parameters of the machine learning model, including feature weight configuration and recognition thresholds.

[0106] According to one embodiment of this application, a machine learning model is used to map a comprehensive feature vector to a motion sickness level. The output of the machine learning model is the probability distribution vector p for each motion sickness level, and the final motion sickness level prediction result is obtained. From the highest probability The decision, expressed as:

[0107] ;

[0108] .

[0109] The machine learning model is trained offline using labeled data collected in a laboratory or real-world vehicle environment (i.e., data that simultaneously records physiological data and subjective motion sickness levels). The machine learning model can employ existing gradient boosting decision tree models (such as LightGBM / XGBoost) or Long Short-Term Memory (LSTM) network models. Gradient boosting decision tree models learn by synthesizing feature vectors. It can detect the non-linear relationship between these features and motion sickness levels, is good at handling heterogeneous features, and has a fast training speed.

[0110] Long Short-Term Memory (LSTM) network models are recurrent neural networks with sophisticated gating mechanisms that allow them to learn long-term dependencies. An input to an LSTM network model is a time sequence, such as the feature vector sequence of the past 30 seconds: [ (t-29), (t-28), ..., [(t)], LSTM processes this sequence step by step, and the "memory cells" inside the model are continuously updated. It integrates the information of the entire time window to make a judgment on the current moment. For example, it observes the trend of continuous increase in skin conductance and continuous decrease in heart rate variability. Even if the current instantaneous feature does not reach the threshold, it can identify motion sickness tendency in advance. It also captures the dynamic temporal evolution of motion sickness, which is beneficial for early identification.

[0111] During training, model parameters can be optimized by minimizing a loss function such as multi-class cross-entropy.

[0112]

[0113] in: Representing sample data The true label, This represents the probability predicted by the model.

[0114] According to one embodiment of this application, the system's decision core is based on a formalized condition-triggered intervention strategy, which is:

[0115] ;

[0116] Wherein: ϕ represents the control command vector; ϕdrive_smooth indicates a mild state, in which basic smoothness control commands are output; ϕdrive_smoother indicates a moderate state, in which enhanced smoothness control commands are output; ϕscent indicates the coordinated activation of the fragrance adjustment system, in which fragrance adjustment control commands are output; ϕdrive_ultra_smooth indicates a severe state, in which ultimate smoothness control commands are output; ϕalert indicates the simultaneous triggering of multiple alarm mechanisms, in which alarm control commands are output; and ϕsuggest indicates the activation of the intelligent suggestion system, in which intelligent suggestion control commands are output.

[0117] The system translates intervention strategies into specific sets of control commands. Each command includes fields such as target system, action type, intensity parameters, and duration. The system distributes these commands to the vehicle control system via an onboard bus (such as CAN or Ethernet), ensuring coordinated operation to avoid conflicts; for example, smoothing acceleration / deceleration curves, limiting rapid acceleration / braking, and receiving adjustment commands for suspension stiffness / steering assist, thereby performing corresponding ride comfort control. The fragrance control system receives fragrance adjustment control commands and releases fragrance to improve the in-vehicle environment.

[0118] The system employs a central coordinator responsible for monitoring the response status of each subsystem to commands. When multiple commands compete for resources, the coordinator arbitrates based on the severity of motion sickness and predefined priorities, ensuring that high-priority commands (such as the ultimate smoothness control for severe motion sickness) are executed first. The system continuously monitors changes in the physiological characteristics of passengers after intervention, forming a feedback loop. If the motion sickness level does not decrease or even increases within a preset time window after intervention, the system will trigger a command optimization process, such as increasing the intensity of intervention or switching the intervention mode. Furthermore, the system records the execution effect of each command for offline optimization of the command parameter library, continuously improving the accuracy and effectiveness of anti-motion sickness interventions.

[0119] This system achieves adaptive optimization of the strategy through a closed-loop feedback mechanism. It is based on the changes in physiological characteristics before and after the intervention. Construct an intervention effectiveness index evaluation model:

[0120] ;

[0121] in, The weighting coefficients of each physiological characteristic satisfy the following conditions: Significant and effective intervention criteria include: a decrease in skin conductance response rate exceeding 15% ( Heart rate variability (LF / HF ratio) decreased by more than 20% ), while the RMSSD value increased by more than 10% ( ).

[0122] Secondly, this invention discloses a motion sickness prevention system based on passenger motion sickness feature recognition, including...

[0123] Sensing layer: Real-time acquisition of multimodal signals from occupants, including skin conductance signals, heart rate signals, and motion signals;

[0124] The recognition layer performs timestamp alignment and motion artifact compensation on the acquired multimodal signals; it extracts features from the preprocessed multimodal signals and fuses the extracted feature vectors to form a comprehensive feature vector; when a passenger takes their first ride, it identifies the passenger's motion sickness sensitivity level based on the multimodal signals, and then establishes a binding relationship between the passenger's facial features and motion sickness sensitivity level through facial recognition technology to form a personalized recognition database; it inputs the comprehensive feature vector into a pre-trained machine learning model, and dynamically adjusts the parameters of the machine learning model by retrieving the passenger's motion sickness sensitivity level from the personalized recognition database when the passenger boards the vehicle, thus identifying the passenger's real-time motion sickness level.

[0125] Decision-making level: Based on the identified real-time motion sickness level, generate corresponding intervention instructions according to the preset intervention strategy;

[0126] Execution layer: Controls the corresponding execution mechanism according to the intervention command.

[0127] According to one embodiment of this application, the sensing layer acquires multimodal signals from the occupant in real time using a flexible fabric electrode glove. The flexible fabric electrode glove includes a skin conductance signal acquisition module, a heart rate signal acquisition module, a motion signal acquisition module, and a microcontroller, all mounted on the hand. The microcontroller transmits the skin conductance signals, heart rate signals, and motion signals acquired by the skin conductance signal acquisition module, heart rate signal acquisition module, and motion signal acquisition module, respectively, to the vehicle-mounted local processing terminal via a wireless communication module. The flexible fabric electrode glove uses an elastic fabric matrix as its body, ensuring comfort and stability during extended wear and effectively avoiding displacement and signal interruption caused by hand movements in traditional wristbands and finger clip devices. The wires connecting the controller and each module can be embedded in the glove body using a mixed-weave method.

[0128] According to one embodiment of this application, the electromyography (EMG) signal acquisition module includes a surface EMG sensor disposed on the glove body and one side of the palm. The two electrode pads of the surface EMG sensor are respectively disposed in areas corresponding to the thenar and hypothenar muscles of the palm. The surface EMG sensor is a sensing device used for non-invasive detection of muscle electrical activity on the human skin surface. It captures the weak voltage (typically in the range of ±1mV) generated during muscle contraction through metal electrodes attached to the skin. After processing by pre-amplification, filtering, and notch filtering circuits, it outputs an amplified signal for real-time display, recording, or further analysis by the back-end system.

[0129] The heart rate signal module includes a photoplethysmography (PPG) detector located on the glove body and wrist area, and a PPG light source and photodetector positioned facing the front of the wrist. PPG is a non-invasive measurement technique that uses photoelectric methods to detect changes in blood volume in living tissue. When the light source illuminates the skin, the fluctuations in blood volume caused by the heartbeat result in periodic fluctuations in the intensity of light absorption and scattering over time. The photodetector converts these light intensity changes into electrical signals, thus obtaining waveforms reflecting changes in blood flow and blood volume.

[0130] The motion signal acquisition module includes a six-axis inertial measurement unit (IMU) located on the back of the glove body. The six-axis IMU is an inertial sensor composed of a three-axis accelerometer and a three-axis gyroscope, which can measure the linear acceleration and angular velocity of an object in three-dimensional space in real time and provide 6-DOF attitude information.

[0131] The low-power microcontroller and signal conditioning circuit are located on the back of the glove body. The microcontroller and signal conditioning circuit perform preliminary amplification, filtering and analog-to-digital conversion on the original signal, and finally wirelessly transmit the pre-processed data to the local processing terminal through the Bluetooth Low Energy module.

[0132] To ensure the system's portability and extended working capacity, the gloves employ an integrated, dedicated power supply solution. The core of the system is powered by a rechargeable micro lithium polymer battery, whose slim design is cleverly integrated into the wrist sleeve, providing power without compromising wearing comfort or freedom of movement.

[0133] The system can also integrate intelligent sensing and control logic, enabling it to automatically start and stop power supply under specific conditions.

[0134] Automatic power-on: When the system detects continuous and regular low-frequency vibrations (typically after a vehicle engine starts) through the integrated motion signal acquisition module, or detects human physiological signals consistent with the wearing status through the wrist heart rate signal acquisition module, the power management unit will determine that the user has entered the vehicle environment and is wearing gloves, and then automatically wake up and power on the entire system, entering standby and data acquisition state.

[0135] Automatic power cut-off: When the motion signal acquisition module and heart rate signal acquisition module jointly determine that the hand has been in a static state for more than a preset threshold and no vehicle vibration characteristics are detected, or the Bluetooth connection is disconnected and the system has been idle for a period of time, the power management unit will determine that the trip is over or the user has taken off the gloves, and then automatically cut off the power supply to the main system, retaining only the key sensor detection circuits and entering an extremely low power sleep state to maximize energy saving.

[0136] During active operation, the efficient power management unit can perform high-precision conversion and distribution of battery voltage for modules operating at different voltages within the system. For charging, a convenient magnetic contact charging design allows users to quickly replenish energy.

[0137] Furthermore, the system integrates a dynamic power management strategy at the firmware level. The microcontroller can dynamically adjust the processor core frequency and peripheral circuit power consumption according to the real-time task load, and automatically enter a low-power sleep mode during idle periods of data acquisition and transmission. Through this intelligent power supply strategy, the system optimizes the overall battery life to more than 8 hours, which is sufficient to meet the continuous monitoring needs of a single long-distance trip.

[0138] The instruction manual states that before this motion sickness prevention method and system can be operated, individual consent from the passenger must be obtained before any data collection and subsequent related work can be carried out.

[0139] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for preventing motion sickness based on identifying motion sickness characteristics of a passenger, characterized in that, The method comprises the steps of: S1: collecting multi-modal signals of the passenger in real time, the multi-modal signals comprising a skin electricity signal, a heart rate signal and a motion signal; S2: performing timestamp alignment and motion artifact compensation on the collected multi-modal signals; S3: performing feature extraction on the preprocessed multi-modal signals, and performing feature fusion on the extracted feature vectors to form a comprehensive feature vector; S4: when the passenger first gets on the vehicle, identifying the car sickness sensitivity level of the passenger according to the multi-modal signals, and then establishing a binding relationship between the facial features of the passenger and the car sickness sensitivity level through face recognition technology to form a personalized identification database; S5: inputting the comprehensive feature vector into a pre-trained machine learning model, and dynamically adjusting the parameters of the machine learning model according to the car sickness sensitivity level of the passenger retrieved from the personalized identification database when the passenger gets on the vehicle to identify the real-time car sickness level of the passenger; S6: generating corresponding intervention instructions according to the identified real-time car sickness level according to a preset intervention strategy.

2. The anti-motion sickness method based on occupant motion sickness feature recognition according to claim 1, characterized in that, The step S2 specifically comprises: S21: performing timestamp alignment on the skin electricity signal, the heart rate signal and the motion signal to obtain a skin electricity signal sequence, a heart rate signal sequence and a motion signal sequence that are completely aligned in time; S22: using an adaptive noise cancellation algorithm to compensate for motion artifacts in the skin electricity signal and the heart rate signal using the motion signal to eliminate noise in the skin electricity signal and the heart rate signal. 3.The anti-motion sickness method based on the passenger motion sickness feature recognition of claim 1, wherein, The comprehensive feature vector is represented as: ； wherein: , respectively represent the mean and standard deviation of the skin conductance level; represents the number of skin conductance responses per minute; represents the energy ratio of the target frequency band; represents the mean heart rate; represents the standard deviation of all normal sinus intervals; represents the root mean square of adjacent RR intervals; represents the heart rate variability index; , and respectively represent the dominant frequency, mean power and power spectrum entropy of the motion signal, and T represents the transpose of the vector. 4.The anti-motion sickness method based on the passenger motion sickness feature recognition of claim 1, wherein, The multi-modal signals identify the car sickness sensitivity level of the passenger specifically comprising: The change intensity of the skin conductance level in a unit of time is calculated The formula is: ； Computing heart rate variability sensitivity factor with the formula: ； wherein: represents the ratio of the low frequency heart rate of the occupant in a dynamic driving state to the high frequency heart rate represents the ratio of the low frequency heart rate of the occupant in a resting state to the high frequency heart rate ;​ Calculating exercise tolerance indicators with the formula: ； wherein: represents the magnitude of motion perceived subjectively by the occupant or responded to by the physiological system, represents the true physical motion magnitude of the vehicle itself; based on the galvanic skin response intensity index , heart rate variability sensitivity coefficient and stress sensitivity , the passengers are divided into three levels of low sensitivity, medium sensitivity and high sensitivity by using fuzzy comprehensive evaluation method.

5. The anti-motion sickness method based on occupant motion sickness feature recognition according to claim 1 or 3, characterized in that, The machine learning model is used to map the comprehensive feature vector to a car sickness level, and the output of the machine learning model for calculating the heart rate variability sensitivity coefficient is a probability distribution vector p of each car sickness level, and a final car sickness level prediction result is determined by the maximum probability , and the expression is: ； 。 6.The anti-motion sickness method based on the passenger motion sickness feature recognition according to claim 5, characterized in that, The intervention strategy is: ； Wherein: represents a control instruction vector; represents a mild state, and then a basic smoothness control instruction is output; represents a moderate state, and then an enhanced smoothness control instruction is output; represents a collaborative activation of a fragrance adjustment system, and then a fragrance adjustment control instruction is output; represents a severe state, and then an extreme smoothness control instruction is output; represents a synchronous triggering of a multiple alarm mechanism, and then an alarm control instruction is output; represents an intelligent suggestion system, and then an intelligent suggestion control instruction is output.

7. A motion sickness prevention system based on identification of motion sickness characteristics of an occupant, characterized by, The method comprises A perception layer: collecting multi-modal signals of the passenger in real time, the multi-modal signals comprising a skin electricity signal, a heart rate signal and a motion signal; An identification layer: performing timestamp alignment and motion artifact compensation on the collected multi-modal signals; performing feature extraction on the preprocessed multi-modal signals, and performing feature fusion on the extracted feature vectors to form a comprehensive feature vector; when the passenger first gets on the vehicle, identifying the car sickness sensitivity level of the passenger according to the multi-modal signals, and then establishing a binding relationship between the facial features of the passenger and the car sickness sensitivity level through face recognition technology to form a personalized identification database; and inputting the comprehensive feature vector into a pre-trained machine learning model, and dynamically adjusting the parameters of the machine learning model according to the car sickness sensitivity level of the passenger retrieved from the personalized identification database when the passenger gets on the vehicle to identify the real-time car sickness level of the passenger; The decision layer generates corresponding intervention instructions according to the preset intervention strategy according to the identified real-time car sickness level; The execution layer controls the corresponding execution mechanism according to the intervention instructions.

8. The anti-motion sickness system based on occupant motion sickness feature recognition of claim 7, wherein, The perception layer collects multi-modal signals of the passenger in real time through a flexible fabric electrode glove, and the flexible fabric electrode glove includes a glove body, a skin electricity signal collection module, a heart rate signal module, a motion signal collection module and a microcontroller installed on the hand, the microcontroller sends the skin electricity signal, the heart rate signal and the motion signal collected by the skin electricity signal collection module, the heart rate signal collection module and the motion signal collection module to the vehicle-mounted local processing terminal through a wireless communication module.

9. The anti-motion sickness system based on occupant motion sickness feature recognition of claim 8, wherein, The skin electricity signal collection module includes a skin surface electromyography sensor arranged on one side of the palm of the glove body and the palm, and two electrode pieces of the skin surface electromyography sensor are arranged in regions corresponding to the thenar muscle and the hypothenar muscle of the palm respectively; The heart rate signal module includes a light source and a photoelectric detector of an optical volume pulse wave detector arranged on the glove body and the wrist region, and the light source and the photoelectric detector face the wrist side; The motion signal collection module includes a six-axis inertial measurement unit arranged on the back of the glove body.

10. The anti-motion sickness system based on occupant motion sickness feature recognition of claim 8 or 9, wherein, When the microcontroller detects a continuous and regular low-frequency motion signal uploaded by the motion signal collection module, or starts to obtain a heart rate signal uploaded by the heart rate signal module, the whole system is automatically woken up to enter a working mode; When the microcontroller detects that the skin electricity signal and the heart rate signal are in a stable state for more than a threshold value and does not detect a vibration signal, or detects that the Bluetooth is in a disconnected state and the system is in an idle state for more than a preset threshold value, the system enters a low-power sleep period.

Images