Electric vehicle driver and passenger carsickness degree prediction method and system

Abstract

The invention discloses an electric vehicle driver carsickness degree prediction method and system, and relates to the field of driver carsickness prediction, and the method comprises the steps: obtaining historical driver physiological data and vehicle operation data, and carrying out the preprocessing of the physiological data and the vehicle operation data, performing feature extraction on the preprocessed physiological data and the preprocessed vehicle operation data to obtain feature parameters related to carsickness, and performing feature fusion on the feature parameters related to carsickness to obtain carsickness feature vectors; constructing a carsickness degree prediction model; and training the carsickness degree prediction model, and predicting the carsickness degree of the to-be-detected electric vehicle driver and passenger by using the trained carsickness degree prediction model to obtain a carsickness degree prediction result. And comprehensive and accurate prediction of the carsickness degree of the driver and passengers is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of vehicle sickness prediction, in particular to a method and system for predicting the degree of vehicle sickness of a driver or passenger of an electric vehicle. BACKGROUND

[0002] Vehicle sickness, as a common manifestation of motion sickness, is mainly caused by the conflict between the motion state perceived by the human body's inner ear balance receptors and the information received by the vision during vehicle driving, which leads to the disorder of the central nervous system and a series of discomfort symptoms such as dizziness, nausea, vomiting, and pale complexion. Vehicle sickness not only greatly affects the travel experience of the driver or passenger and reduces their satisfaction with the electric vehicle, but also may pose a threat to road safety in severe cases. With the continuous development and popularization of electric vehicles, their share in the travel field is increasing. While enjoying the many conveniences brought by electric vehicles, such as zero emissions, low noise, and lower operating costs, the phenomenon of vehicle sickness of the driver or passenger in the vehicle has gradually attracted widespread attention.

[0003] In the field of traditional fuel vehicles, there have been many studies and countermeasures on vehicle sickness, but due to the significant differences between electric vehicles and traditional fuel vehicles in driving characteristics, power system layout, and vehicle vibration characteristics, these differences may have different effects on the driver's or passenger's experience of vehicle sickness. For example, the power output of electric vehicles is more direct and rapid, and the vibration frequency and amplitude generated by the motor during operation are different from those of the engine, which may change the human body's motion perception and balance state in the vehicle. In addition, the center of gravity distribution, suspension system design, and interior space layout of electric vehicles may also change due to factors such as the installation location of the battery pack, further affecting the comfort and vehicle sickness tendency of the driver or passenger.

[0004] Currently, existing methods for predicting the degree of vehicle sickness often rely mainly on the subjective feedback of the driver or passenger, such as collecting passengers' descriptions of their symptoms and self-perception scores during the ride through questionnaires and other means. This method not only has a large subjective bias and is difficult to accurately quantify the degree of vehicle sickness, but also cannot deeply analyze the specific reasons for the occurrence of vehicle sickness and the interaction between the influencing factors. At the same time, some studies have attempted to use some physiological monitoring devices to predict the vehicle sickness state of the driver or passenger, such as heart rate variability monitors, electroencephalographs, etc., to indirectly reflect the vehicle sickness state. However, these devices are usually complex and costly, and data interpretation requires professional medical knowledge background, making it difficult to be widely popularized and applied in actual vehicle design, research and development, and use scenarios. SUMMARY

[0005] The purpose of the present application is to provide a method and system for predicting the degree of car sickness of an electric vehicle driver or passenger, which can comprehensively and accurately predict the degree of car sickness and avoid the problem of uncertain degree of car sickness caused by subjective factors of the driver or passenger.

[0006] To achieve the above-mentioned purpose, the present application provides the following solutions. In a first aspect, the present application provides a method for predicting the degree of car sickness of an electric vehicle driver or passenger, comprising: obtaining historical physiological data of the driver or passenger and vehicle operation data; preprocessing the physiological data and the vehicle operation data to obtain preprocessed physiological data and preprocessed vehicle operation data; extracting features from the preprocessed physiological data and the preprocessed vehicle operation data to obtain car sickness-related feature parameters, wherein the car sickness-related feature parameters include: Mean the average amplitude of an electroencephalogram signal sequence PSD the power spectral density of a frequency domain signal of the electroencephalogram signal sequence W the baseline level of a skin conductance signal sequence SCL the power spectral density of a frequency domain signal of the skin conductance signal sequence PSD S the wavelet-transformed skin conductance signal sequence W S the number of eye fixation points NOF the average eye fixation time MFD the blink frequency BR the distance between facial expression key points d the intensity of expression muscle activity MAI the variance of vehicle acceleration and the variance of vehicle angular velocity ; performing feature fusion on the car sickness-related feature parameters to obtain a car sickness feature vector; dividing the car sickness feature vector into a training set, a validation set, and a test set; constructing a car sickness degree prediction model; training the car sickness degree prediction model based on the training set to obtain a trained car sickness degree prediction model; using the trained car sickness degree prediction model to predict the degree of car sickness of a to-be-tested electric vehicle driver or passenger to obtain a car sickness degree prediction result.

[0007] In a second aspect, a system for predicting the degree of car sickness of an electric vehicle driver or passenger is provided, comprising: Wearable collection device and fixed collection device, the wearable collection device and the fixed collection device are used for acquiring historical driver physiological data and vehicle operation data; Motion sickness online evaluation platform, the motion sickness online evaluation platform is connected with the wearable collection device and the fixed collection device, is used for preprocessing the physiological data and the vehicle operation data, obtains the preprocessed physiological data and the preprocessed vehicle operation data, is also used for extracting features to the preprocessed physiological data and the preprocessed vehicle operation data, obtains the feature parameters related to motion sickness;The feature parameters related to motion sickness include: average amplitude of electroencephalogram sequence Mean Power spectral density of electroencephalogram sequence frequency domain signal PSD Wavelet transformed electroencephalogram sequence W Skin conductance signal sequence baseline level SCL Power spectral density of skin conductance signal sequence frequency domain signal PSD S Wavelet transformed skin conductance signal sequence W S Number of eye fixation points NOF Average eye fixation time MFD Blinking frequency BR Facial expression key point distance d Expression muscle activity intensity MAI Vehicle acceleration variance And vehicle angular velocity variance Also used for feature fusion to the feature parameters related to motion sickness, obtain the motion sickness feature vector;Also used for dividing the motion sickness feature vector into training set, validation set and test set;Also used for constructing motion sickness degree prediction model;Also used for training the motion sickness degree prediction model based on the training set, obtain the trained motion sickness degree prediction model;Also used for predicting the motion sickness degree of the electric vehicle driver to be measured by using the trained motion sickness degree prediction model, obtain the motion sickness degree prediction result.

[0008] According to the specific embodiments provided by the application, the application has the following technical effects: The application provides a method and system for predicting the degree of car sickness of an electric vehicle driver or passenger. BRIEF DESCRIPTION OF DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort.

[0010] Figure 1 A flowchart of a method for predicting the degree of car sickness of an electric vehicle driver or passenger according to an embodiment of the present application is provided. Figure 2 A functional module diagram of a system for predicting the degree of car sickness of an electric vehicle driver or passenger according to an embodiment of the present application is provided. DETAILED DESCRIPTION

[0011] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without any creative effort are within the scope of protection of the present application.

[0012] The present application aims to provide a method and system for predicting the degree of car sickness of an electric vehicle driver or passenger, and to achieve comprehensive and accurate prediction of the degree of car sickness of a driver or passenger.

[0013] The above objects, features and advantages of the present application will become more apparent from the following detailed description considered in conjunction with the accompanying drawings and specific embodiments.

[0014] As shown in the Figure 1 , the embodiment provides a method for predicting the degree of car sickness of a driver or passenger of an electric vehicle, comprising the following steps: Step 1: obtaining historical physiological data of a driver or passenger and vehicle operation data; the physiological data specifically includes electroencephalogram data, skin conductance data, eye movement data and facial expression data.

[0015] Step 2: preprocessing the physiological data and the vehicle operation data, including filtering, denoising and normalization, etc., to eliminate noise interference and dimension difference in the data, improve the quality and availability of the data, and obtain preprocessed physiological data and preprocessed vehicle operation data.

[0016] Step 3: extracting features from the preprocessed physiological data and the preprocessed vehicle operation data to obtain feature parameters related to car sickness.

[0017] Among them, the feature parameters related to car sickness include: the average amplitude of the electroencephalogram signal sequence Mean , the power spectral density of the frequency domain signal of the electroencephalogram signal sequence PSD , the electroencephalogram signal sequence after wavelet transform W , the baseline level of the skin conductance signal sequence SCL , the power spectral density of the frequency domain signal of the skin conductance signal sequence PSD S , the skin conductance signal sequence after wavelet transform W S , the number of eye fixation points NOF , the average eye fixation time MFD , the blink frequency BR , the facial expression key point distance d , the expression muscle activity intensity MAI , the vehicle acceleration variance and the vehicle angular velocity variance . Each feature parameter will be introduced one by one as follows, which specifically includes the following steps: The extraction process of the average amplitude of the electroencephalogram signal sequence Mean , the power spectral density of the frequency domain signal of the electroencephalogram signal sequence PSD and the electroencephalogram signal sequence after wavelet transform W is as follows: Step 301: obtaining the electroencephalogram signal sequence based on the preprocessed electroencephalogram data, and calculating the average amplitude of the electroencephalogram signal sequence Mean, power spectral density of frequency domain signal of EEG signal sequence PSD And the EEG signal sequence after wavelet transformation W .

[0018] The average amplitude of the EEG signal sequence Mean , calculated using the following formula: ; in, Indicates the first The amplitude of the EEG signal sampling point, N is the total number of EEG signal sampling points. This feature reflects the average amplitude level of the EEG signal over a period of time.

[0019] The power spectral density of the frequency domain signal of the EEG signal sequence PSD Calculated as follows: First, the EEG signal sequence is subjected to fast Fourier transform to convert it from the time domain to the frequency domain. FFT The result is a complex sequence that contains the frequency components of the signal and the corresponding amplitudes. The formula is as follows: ; in, x(n) is the time domain signal of the EEG signal sequence, is the frequency domain signal of the EEG signal sequence, is the total number of sampling points of the EEG signal, j is an imaginary unit, k is the frequency domain index, corresponding to the k spectral lines, , n is the time domain sampling point index, representing the first n sampling points, .

[0020] Secondly, based on the frequency domain signal of the EEG signal sequence, the power spectrum density of the frequency domain signal of the EEG signal sequence is calculated PSD , the formula is as follows: ; in, Express Modulus, Represents the power spectral density of the frequency domain signal of the EEG signal sequence, by calculating FFT The power spectrum density can be obtained by square the modulus of the result and dividing it by the number of sampling points. The power spectrum density of the frequency domain signal of the EEG signal sequence represents the power distribution of the signal at different frequencies, which can help identify the main frequency components in the EEG signal, such as α Wave (8-13Hz),β waves (13-30 Hz) and so on.

[0021] the wavelet-transformed electroencephalogram sequence W , which is calculated by the following formula: ; wherein, is the electroencephalogram sequence, is the wavelet basis function, is the scale parameter, is the translation parameter, denotes the conjugate. The result of the wavelet transform denotes the wavelet coefficients of the signal at different scales and positions, which can be used to analyze the time-frequency characteristics of the electroencephalogram sequence, such as detecting transient events and frequency changes in the electroencephalogram. Through wavelet transform, the present application can analyze the electroencephalogram in time domain and frequency domain at the same time, and extract the local features of the signal by convolving the electroencephalogram with wavelet functions at different scales and positions.

[0022] In addition to calculating the average amplitude of the electroencephalogram sequence Mean , the power spectral density of the frequency domain signal of the electroencephalogram sequence PSD and the wavelet-transformed electroencephalogram sequence W , the present embodiment also calculates the variance and standard deviation of the electroencephalogram sequence based on the electroencephalogram sequence.

[0023] The variance of the electroencephalogram sequence is calculated by the following formula: ; wherein, denotes the variance of the electroencephalogram sequence, which measures the dispersion degree of the amplitude of the electroencephalogram sequence, and a larger variance indicates that the signal amplitude fluctuates more greatly.

[0024] The standard deviation of the electroencephalogram sequence is calculated by the following formula: ; wherein, denotes the standard deviation of the electroencephalogram sequence, which is the square root of the variance and has the same dimension as the original signal, and more intuitively represents the fluctuation size of the electroencephalogram sequence.

[0025] This embodiment determines whether the EEG signal sequence has quality abnormalities by calculating the variance and standard deviation of the EEG signal sequence for subsequent process decision-making. When the variance or standard deviation is too small, it indicates that the signal may be a straight line due to the detachment or poor contact of the sensor collecting the EEG signal; when the variance or standard deviation is too large, it indicates that large myoelectric, electrooculographic artifacts or motion artifacts may be mixed in. Small or large variances or standard deviations will be marked as "abnormal signal quality". When the variance or standard deviation is not large, the EEG signal sequence is of mild quality abnormality, and the EEG signal sequence is labeled with a low-weight quality label, and the model automatically downgrades its weight during training or inference; when the variance or standard deviation is too large, the EEG signal sequence is of severe quality abnormality, and the entire segment is directly discarded or triggered to be re-collected.

[0026] The feature extraction of EEG signals in this embodiment involves time domain, frequency domain and time-frequency domain analysis of the EEG signal sequence, which can more comprehensively and accurately reflect the brain wave conditions of the driver and passengers, making the subsequent prediction of motion sickness degree more accurate.

[0027] Skin conductance signal sequence baseline level SCL , power spectral density of frequency domain signal of skin conductance signal sequence PSD S And the skin conductance signal sequence after wavelet transformation W S The extraction process is as follows: Step 302: Obtain a skin conductance signal sequence based on the preprocessed skin conductance data, and calculate the baseline level of the skin conductance signal sequence SCL , power spectral density of frequency domain signal of skin conductance signal sequence PSD S And the skin conductance signal sequence after wavelet transformation W S .

[0028] The skin conductance signal sequence baseline level SCL , calculated using the following formula: ; in, Indicates the first The amplitude of the skin conductance signal sampling point, is the total number of skin conductance signal sampling points. Reflects the baseline level of skin conductance and is often used for long-term trend analysis.

[0029] The power spectral density of the frequency domain signal of the skin conductance signal sequence PSD S , calculated as follows: First, perform fast Fourier transform on the skin conductance signal sequence, and the formula is as follows: ; in, y(n) is the time domain signal of the skin conductance signal sequence, is the frequency domain signal of the skin conductance signal sequence, is the total number of skin conductance signal sampling points, j is an imaginary unit, k is the frequency domain index, corresponding to the k spectral lines, , m is the time domain sampling point index, representing the first m sampling points, .

[0030] Secondly, based on the frequency domain signal of the skin conductance signal sequence, the power spectral density of the frequency domain signal of the skin conductance signal sequence is calculated PSD S , the formula is as follows: ; in, For Modulus, is the power spectral density of the frequency domain signal of the skin conductance signal sequence PSD S .

[0031] The skin conductance signal sequence after wavelet transformation W S , calculated using the following formula: ; in, is the skin conductance signal sequence, (t) is the wavelet basis function, is the scale parameter, is the translation parameter, Indicates conjugation.

[0032] In addition to calculating the baseline level of the skin conductance signal sequence, this embodiment SCL , power spectral density of frequency domain signal of skin conductance signal sequence PSD S And the skin conductance signal sequence after wavelet transformation W S In addition, based on the skin conductance signal sequence, the skin conductance response of the skin conductance signal sequence is calculated. SCR and skin conductance response SCR frequency and amplitude.

[0033] The skin conductance response SCR is calculated using the following formula: ; reflects the instantaneous change of the skin conductance and is used for event-related analysis.

[0034] The frequency and amplitude of the skin conductance response SCR are calculated using the following formula: ; ; wherein, T is the length of the time window, represents the peak count of the skin conductance response SCR , which can reflect the number of stress responses of the skin conductance, i.e. how many independent sympathetic activation events, represents the frequency of the skin conductance response SCR , which reflects how fast the skin conductance events occur, represents the maximum value of the skin conductance response, represents the minimum value of the skin conductance response, represents the amplitude of the skin conductance response, which reflects the activation intensity of the skin conductance response.

[0035] When the amplitude of the skin conductance response is too large or too small, it indicates that the patch of the skin conductance acquisition device is off or saturated with sweat, and in this case, the entire skin conductance signal sequence data is marked as invalid; when the frequency of the skin conductance response is too high and the waveform is dense, it is suspected to be a motion artifact, and manual review or re-cleaning using the Independent Component Analysis (ICA) method is performed.

[0036] The embodiment calculates the frequency and amplitude of the skin conductance response SCR and the skin conductance response SCR of the skin conductance signal sequence, which can ensure that the samples entering the car sickness degree prediction model are all noise-free samples.

[0037] The feature extraction of the skin conductance signal in the embodiment involves time domain, frequency domain and time-frequency domain analysis of the skin conductance signal sequence, which can comprehensively reflect the physiological state changes of the driver or passenger during the driving of the electric vehicle and provide an important physiological basis for the evaluation of car sickness degree.

[0038] The number of eye fixation points of the eye movement signal sequence NOF , the average eye fixation time MFDand blink frequency BR The extraction process is as follows: Step 303: Obtain an eye movement signal sequence based on the preprocessed eye movement data, and calculate the number of eye fixation points of the eye movement signal sequence NOF , average eye fixation time MFD and blink frequency BR .

[0039] The number of eye fixation points NOF is calculated by the following formula: ; Wherein, is the total number of fixation points detected within a certain time. This feature reflects the number of eye fixations within a certain time period.

[0040] The average eye fixation time MFD is calculated by the following formula: ; Wherein, represents the duration of the th fixation point. The average fixation time can reflect the length of time the eye stays at each fixation point.

[0041] The blink frequency BR is calculated by the following formula: ; Wherein, is the number of blinks detected within a time , which can reflect the blink frequency of the eye. A higher blink frequency is related to the inattention or fatigue of the driver.

[0042] In addition to calculating the number of eye fixation points NOF , average eye fixation time MFD and blink frequency BR , this embodiment also extracts features from eye movement signal trajectories based on the eye movement signal sequence, and calculates eye movement trajectory length and eye movement trajectory curvature.

[0043] The eye movement trajectory length is calculated by the following formula: ; Wherein, represents the eye movement trajectory length, represents the coordinates of the th sampling point on the eye movement trajectory, is the total number of eye movement signal sampling points. This feature reflects the total length of the eye movement trajectory and can be used to evaluate the activity level of eye movement.

[0044] The curvature of the eye movement trajectory is calculated using the following formula: ; in, represents the curvature of the eye movement trajectory, is the angle change of the eye movement trajectory along a certain path, The curvature of the eye movement trajectory can reflect the degree of curvature of the eye movement trajectory. A higher curvature indicates a more complex eye movement trajectory.

[0045] If the eye movement track is too short, it indicates that the sensor collecting the eye movement data is blocked, the driver's eyes are closed, or other reasons have caused the eye movement signal to be lost. In this case, the entire eye movement signal sequence is discarded or downgraded. If the curvature of the eye movement track increases suddenly, it indicates suspected blink artifacts or head movement interference, which triggers manual review or artifact removal.

[0046] This embodiment calculates the length and curvature of the eye movement trajectory and uses them as hard thresholds in the data cleaning stage to ensure that noise-free samples are used for subsequent training or inference.

[0047] This embodiment can accurately reflect the eye movements of the driver and passengers through the special extraction of eye movement signals, making the subsequent prediction of the degree of motion sickness more accurate.

[0048] Facial expression keypoint distance of facial expression signal sequence d and facial muscle activity intensity MAI The extraction process is as follows: Step 304: Obtain a facial expression signal sequence based on the pre-processed facial expression data, and calculate the distance between the facial expression key points of the facial expression signal sequence. d and facial muscle activity intensity MAI .

[0049] The distance between the facial expression key points d , calculated as follows: First, get the coordinates of facial expression key points: ; in, Indicates the The coordinates of the key points of the expression, is the total number of facial expression key points, and the coordinates can be used to describe the geometry of facial expressions.

[0050] Secondly, based on the above facial expression key point coordinates, calculate the distance between facial expression key points d , using the following formula: ; in, Indicates the Facial expression key points and The distance between facial expression key points can be used to quantify the changes in facial expressions, such as the degree of mouth opening or eyebrow raising.

[0051] The intensity of facial muscle activity MAI , calculated using the following formula: ; in, Indicates the The intensity of the facial muscles is the total number of muscles, which reflects the overall activity level of the driver's facial muscles.

[0052] In addition to calculating the distance between the facial expression key points d and facial muscle activity intensity MAI In addition, the facial muscle activity change rate is calculated based on the facial expression signal sequence.

[0053] The facial muscle activity change rate is calculated using the following formula: ; in, Indicates the rate of change of facial muscle activity, Indicates the change in muscle activity intensity. The facial muscle activity change rate can reflect the dynamic change speed of facial expressions.

[0054] During the feature fusion stage, segments with high facial muscle activity change rates are assigned higher weights. By calculating the facial muscle activity change rates, this embodiment can ensure that when a driver or passenger suddenly feels unwell, the activity will not be diluted by the static mean.

[0055] In this embodiment, the feature extraction of facial expression signals involves the distance between key points of facial expression d and facial muscle activity intensity MAI , can accurately capture changes in the facial expressions of drivers and passengers, making subsequent predictions of motion sickness levels more accurate.

[0056] Vehicle acceleration variance and the vehicle angular velocity variance The extraction process is as follows: Step 305: Calculate the vehicle acceleration variance based on the pre-processed vehicle operation data and the vehicle angular velocity variance .

[0057] The vehicle acceleration variance , by the following way: First, based on the pre-processed vehicle operation data, a vehicle acceleration signal sequence is obtained, and based on the vehicle acceleration signal sequence, a vehicle acceleration mean value is calculated, which is calculated by the following formula: ; wherein, represents the vehicle acceleration mean value, represents the acceleration value of the i-th vehicle acceleration signal sampling point in the vehicle acceleration signal sequence, is the total number of vehicle acceleration signal sampling points, and the feature reflects the average acceleration level of the vehicle in a period of time. Secondly, based on the above vehicle acceleration mean value, the vehicle acceleration variance

[0058] is calculated, using the following formula: ; In this embodiment, the vehicle acceleration variance can measure the fluctuation degree of the vehicle acceleration, and a larger variance indicates that the vehicle acceleration changes greatly, which may be related to the frequent acceleration and deceleration of the vehicle.

[0059] The vehicle angular velocity variance is calculated by the following way: First, based on the pre-processed vehicle operation data, a vehicle angular velocity signal sequence is obtained, and based on the vehicle angular velocity signal sequence, a vehicle angular velocity mean value is calculated, which is calculated by the following formula: ; wherein, represents the vehicle angular velocity mean value, represents the angular velocity value of the i-th vehicle angular velocity signal sampling point in the vehicle angular velocity signal sequence, is the total number of vehicle angular velocity signal sampling points, and the feature reflects the average rotation speed of the vehicle in a period of time. Secondly, based on the above vehicle angular velocity mean value, the vehicle angular velocity variance

[0060] is calculated, using the following formula: ; In this embodiment, the vehicle angular velocity variance can measure the fluctuation degree of the vehicle angular velocity, and a larger variance indicates that the rotation motion of the vehicle changes greatly, which may be related to the frequent steering of the vehicle.

[0061] In addition to calculating the above vehicle acceleration variance​​ and vehicle angular velocity variance Further, based on the pre-processed vehicle operation data, the maximum value of a vehicle acceleration signal sequence is calculated and the minimum value , the maximum value of a vehicle angular velocity signal sequence and the minimum value , the vehicle average speed , the vehicle speed variance and the maximum value of a vehicle speed signal sequence and the minimum value .

[0062] The maximum value of the vehicle acceleration signal sequence and the minimum value is calculated by the following formula: ; ; The maximum value of the vehicle acceleration signal sequence and the minimum value may reflect the extreme acceleration and deceleration of the vehicle during driving.

[0063] The maximum value of the vehicle angular velocity signal sequence and the minimum value is calculated by the following formula: ; ; The maximum value of the vehicle angular velocity signal sequence and the minimum value may reflect the extreme rotational speed of the vehicle during driving.

[0064] Based on the pre-processed vehicle operation data, a vehicle speed signal sequence is obtained, and the vehicle average speed is calculated by the following formula: ; wherein, denotes the speed value of the i-th vehicle speed signal sampling point in the vehicle speed signal sequence, is the total number of vehicle speed signal sampling points, and this feature reflects the average speed level of the vehicle in a period of time. The vehicle speed variance

[0065] is calculated by the following formula: ; The vehicle speed variance​ The degree of fluctuation of the vehicle speed is measured, and a larger variance indicates that the vehicle speed changes greatly, which can be related to frequent acceleration and deceleration of the vehicle.

[0066] The maximum value of the vehicle speed signal sequence and the minimum value are calculated by the following formula: ; ; The maximum value of the vehicle speed signal sequence and the minimum value can reflect the extreme speed conditions of the vehicle during driving.

[0067] In this embodiment, the maximum value of the vehicle acceleration signal sequence and the minimum value , the maximum value of the vehicle angular velocity signal sequence and the minimum value , and the maximum value of the vehicle speed signal sequence and the minimum value are calculated, so that the time stamp of strong stimulation events such as “sudden acceleration and deceleration” and “sudden turning” can be marked; and by delaying alignment with the physiological signal (SCR amplitude, facial muscle activity change rate), the lag of physical stimulation to physiological response can be quantified, helping the model to learn the correct causal window.

[0068] In the feature fusion stage, the average vehicle speed and the vehicle speed variance are calculated, so that the physiological features in the period with high variance are given higher weights, and the invalid data of the low-speed smooth road section is avoided to dilute the model attention.

[0069] In this embodiment, the extraction of vehicle motion features involves vehicle acceleration, angular velocity and speed, which can accurately reflect the state of the vehicle used by the driver and passenger, making the subsequent car sickness degree prediction more accurate.

[0070] Step 4: Feature fusion is performed on the car sickness related feature parameters to obtain a car sickness feature vector, specifically including the following steps: Step 401: The average amplitude of the EEG signal sequence Mean , the power spectral density of the frequency domain signal of the EEG signal sequence PSD , and the wavelet transformed EEG signal sequence W are weighted and fitted to obtain an EEG feature vector.

[0071] The average amplitude of the EEG signal sequence Mean , the power spectral density of the frequency domain signal of the EEG signal sequence PSDand the wavelet transformed electroencephalogram sequence W , weighting and fitting are performed, and the following formula is used: ; wherein, is the average amplitude of the electroencephalogram sequence Mean , the power spectral density of the electroencephalogram sequence in the frequency domain PSD and the wavelet transformed electroencephalogram sequence W the electroencephalogram feature after weighting and fitting, , , , is a linear weighting coefficient, which is obtained by linear least squares estimation, A 1 ，K 1 ，B 1 is a nonlinear mapping parameter, which is obtained by quadratic fitting of the training set by nonlinear least squares.

[0072] The electroencephalogram feature vector formula is as follows: ; wherein, is the electroencephalogram feature vector, n is the number of historical drivers and passengers collecting electroencephalogram data, is the electroencephalogram feature of the first driver or passenger, is the electroencephalogram feature of the second driver or passenger, is the electroencephalogram feature of the n driver or passenger.

[0073] Step 402: weighting and fitting are performed on the skin conductance signal sequence baseline level SCL , the power spectral density of the skin conductance signal sequence in the frequency domain PSD S and the wavelet transformed skin conductance signal sequence W S , and the skin conductance feature vector is obtained.

[0074] Weighting and fitting are performed on the skin conductance signal sequence baseline level SCL , the power spectral density of the skin conductance signal sequence in the frequency domain PSD S and the wavelet transformed skin conductance signal sequence W S , and the following formula is used: ; wherein, is the skin conductance signal sequence baseline levelSCL , power spectral density of frequency domain signal of skin conductance signal sequence PSD S And the skin conductance signal sequence after wavelet transformation W S Weighted and fitted EEG features, b 1 、 b 2 、 b 3 、 b 4 is the linear weighting coefficient, obtained by linear least squares estimation, A 2 ，K 2 ，B 2 is the nonlinear mapping parameter, which is obtained by quadratic fitting of the training set through nonlinear least squares.

[0075] The skin conductance eigenvector formula is as follows: ; in, is the skin conductance eigenvector, m The number of historical drivers and passengers for whom skin conductance data was collected, is the skin conductance characteristic of the first driver and passenger, is the skin conductance characteristic of the second driver and passenger, For the m Skin conductance characteristics of the drivers and passengers.

[0076] Step 403: Calculate the number of eye gaze points NOF , average eye gaze time MFD and blink frequency BR , perform weighting and fitting to obtain the eye movement feature vector.

[0077] Number of eye fixations NOF , average eye gaze time MFD and blink frequency BR , weighted and fitted, using the following formula: ; in, is the number of eye fixation points NOF , average eye gaze time MFD and blink frequency BR Weighted and fitted EEG features, c 1 、 c 2 、 c3 are linear weighting coefficients, estimated by linear least squares, A 3 ，K 3 ，B 3 ，C 3 are nonlinear mapping parameters, obtained by quadratic fitting of the training set by nonlinear least squares.

[0078] The facial expression feature vector is calculated as follows: ; wherein, is the facial expression feature vector, p is the number of historical drivers and passengers whose eye movement data are collected, is the eye movement feature of the first driver or passenger, is the eye movement feature of the second driver or passenger, is the eye movement feature of the nth driver or passenger. p

[0079] Step 404: weighting and fitting the facial expression key point distance d and the expression muscle activity intensity MAI to obtain a facial expression feature vector.

[0080] The facial expression key point distance d and the expression muscle activity intensity MAI are weighted and fitted, and the following formula is used: ; wherein, is the facial expression feature after weighting and fitting the facial expression key point distance d and the expression muscle activity intensity MAI , e 1 , e 2 , e 3 are linear weighting coefficients, estimated by linear least squares; A 4 , K 4 , B 4 are nonlinear mapping parameters, obtained by quadratic fitting of the training set by nonlinear least squares.

[0081] The facial expression feature vector is calculated as follows: ; wherein, ​facial expression feature vector of the first driver, q number of historical drivers for collecting facial expression data, facial expression feature of the first driver, facial expression feature of the second driver, facial expression feature of the first driver. q facial expression feature of the second driver.

[0082] Step 405: Weighting and fitting the vehicle acceleration variance and vehicle angular velocity variance to obtain a vehicle motion sickness feature vector.

[0083] The vehicle motion sickness feature vector is as follows: ; wherein, is the vehicle motion sickness feature vector, r is the number of historical drivers for collecting vehicle operation data, is the vehicle motion sickness feature of the first driver, is the vehicle motion sickness feature of the first driver. r is the vehicle motion sickness feature of the second driver.

[0084] Step 406: Weighting and fitting the electroencephalogram feature vector, the skin conductance feature vector, the eye movement feature vector, the facial expression feature vector, and the vehicle motion sickness feature vector to obtain a comprehensive feature vector F .

[0085] The comprehensive feature vector F is as follows: .

[0086] Step 407: Normalizing the elements in the comprehensive feature vector F to eliminate the influence on the fusion result caused by the large differences in the dimensions and numerical ranges of different features. The normalization method can adopt Min-Max normalization to map the feature values to the interval [0, 1]. For a feature , the normalization formula is: ; wherein, and respectively represent the minimum value and the maximum value of the feature in all samples, represents the normalized feature value.

[0087] Step 408: Determine the sensitivity and contribution of the driver's motion sickness condition based on the EEG feature vector, the skin conductance feature vector, the eye movement feature vector, the facial expression feature vector, and the vehicle motion sickness feature vector. Weights can be determined using correlation analysis or machine learning methods, such as principal component analysis (PCA).

[0088] Step 409: Determine the feature weights of the EEG feature vector, the skin conductance feature vector, the eye movement feature vector, the facial expression feature vector, and the vehicle motion sickness feature vector based on the sensitivity and contribution of the driver or passenger to the motion sickness state.

[0089] The formulas for the feature weights of the EEG feature vector, the skin conductance feature vector, the eye movement feature vector, the facial expression feature vector, and the vehicle motion sickness feature vector are as follows: ; ; ; ; ; in, is the feature weight of the EEG feature vector, is the feature weight of the skin conductance feature vector, is the feature weight of the eye movement feature vector, is the feature weight of the facial expression feature vector, is the feature weight of the vehicle motion sickness feature vector, For the n The feature weights assigned to the EEG features of the drivers and passengers are For the m The feature weights assigned to the skin conductance features of the drivers and passengers are For the p The feature weights assigned to the eye movement features of the drivers and passengers are For the q The feature weights assigned to the facial expression features of the drivers and passengers are For the r The feature weights assigned to the vehicle motion sickness features of the driver and passengers in the vehicle.

[0090] Step 410: multiplying the normalized comprehensive feature vector with the feature weight of the electroencephalogram feature vector, the feature weight of the skin conductance feature vector, the feature weight of the eye movement feature vector, the feature weight of the facial expression feature vector, and the feature weight of the vehicle motion sickness feature vector to obtain a weighted feature vector F weighted , which is a motion sickness feature vector.

[0091] The weighted feature vector F weighted The formula is as follows: ; Wherein, F weighted The weighted feature vector is represented by, and the element-wise multiplication is represented by, The normalized electroencephalogram feature vector is represented by, The normalized skin conductance feature vector is represented by, The normalized eye movement feature vector is represented by, The normalized facial expression feature vector is represented by, The normalized vehicle motion sickness feature vector is represented by.

[0092] Through the above feature fusion process, the embodiment obtains a feature vector that comprehensively considers physiological signals and vehicle running states, which can more comprehensively reflect the motion sickness state of the driver or passenger during the driving of the electric vehicle, and provides strong data support for subsequent motion sickness degree evaluation and cause analysis.

[0093] Step 5: dividing the motion sickness feature vector into a training set, a validation set, and a test set; Step 6: constructing a motion sickness degree prediction model, the motion sickness degree prediction model comprising: An input layer for receiving the motion sickness feature vector.

[0094] A convolution layer connected to the input layer, the convolution layer comprising convolution kernels, each convolution kernel being used to extract local features of the motion sickness feature vector, and through convolution operation, the model can capture the spatial correlation in the feature vector.

[0095] A pooling layer connected to the convolution layer, the pooling layer being used to reduce the dimension of the motion sickness feature vector, reduce the calculation amount, and retain important feature information, the pooling operation in the application comprising maximum pooling and average pooling.

[0096] An LSTM layer connected to the pooling layer, the LSTM layer being used to learn the change of the motion sickness feature over time, capable of processing time series data and capturing the time dependence in the data.

[0097] a fully connected layer connected to the LSTM layer, configured to weight and process the features output by the LSTM layer and map the processed features to a classification space of car sickness levels.

[0098] an output layer connected to the fully connected layer, configured to convert the features output by the fully connected layer into a probability distribution using a softmax activation function, representing the probabilities of the occupant being in different car sickness level categories.

[0099] The car sickness level prediction model in this embodiment is a hybrid neural network model based on deep learning, specifically a convolutional neural network (CNN) combined with a long short-term memory network (LSTM). This model can effectively process time series data and capture spatial and temporal features in the data, thereby achieving accurate evaluation of car sickness levels.

[0100] The car sickness level prediction model in this embodiment can also be constructed using other machine learning algorithms or data mining algorithms, such as support vector machines, decision trees, etc.

[0101] Step 7: Train the car sickness level prediction model based on the training set to obtain a trained car sickness level prediction model. The training process specifically includes: Step 701: Initialize the car sickness level prediction model constructed in step 6: initialize the parameters of the model, including convolution kernel weights, LSTM unit parameters, and fully connected layer weights, etc.

[0102] Step 702: Input the training set data into the model, and perform forward propagation through the convolutional layer, pooling layer, LSTM layer, and fully connected layer to calculate the output of the model.

[0103] Step 703: Use the cross-entropy loss function to calculate the difference between the model output and the true label. The cross-entropy loss function can measure the similarity between probability distributions and is suitable for multi-classification problems.

[0104] Step 704: According to the value of the loss function, calculate the gradient of the model parameters through the backpropagation algorithm, and update the model parameters using the Adam optimization algorithm.

[0105] In addition to training the car sickness level prediction model constructed in step 6, the trained car sickness level prediction model is also validated and tested based on the validation set and test set divided in step 5. Specifically, the performance of the model is evaluated on the validation set, and the best model parameters are selected. The generalization ability of the model is tested on the test set to ensure that the model can accurately evaluate the car sickness level on unseen data.

[0106] In this embodiment, in order to further improve the accuracy and generalization ability of the car sickness degree prediction model, the following methods can be used for model evaluation and optimization: ①Cross-validation: use cross-validation method to evaluate the car sickness degree prediction model, to ensure the stability and consistency of the model on different data sets.

[0107] ②Hyperparameter optimization: optimize the hyperparameters of the car sickness degree prediction model, such as learning rate, batch size, convolution kernel size, etc., through grid search or random search, to improve the performance of the model.

[0108] ③Data augmentation: perform augmentation processing on the training data, such as adding noise, data translation, etc., to increase the diversity of the data and improve the robustness of the car sickness degree prediction model.

[0109] ④Model integration: use Boosting model integration method to fuse the prediction results of multiple models to improve the overall prediction performance of the car sickness degree prediction model.

[0110] Step 8: use the trained car sickness degree prediction model to predict the car sickness degree of the test electric vehicle driver, and obtain the car sickness degree prediction result.

[0111] In this embodiment, the car sickness degree prediction result can be divided into no car sickness, mild car sickness, moderate car sickness, severe car sickness, etc.

[0112] In this embodiment, the car sickness feature vector after feature extraction and fusion is input into the trained neural network model, and the model calculates the probability distribution of the driver being in different car sickness degree levels through forward propagation. According to the class corresponding to the maximum value of the probability distribution, the current car sickness degree level of the driver is determined. For example, if the probability distribution output by the model is [0.1, 0.3, 0.5, 0.1], it means that the driver is most likely to be in moderate car sickness, so the car sickness degree is determined to be moderate car sickness.

[0113] Through the above electric vehicle driver car sickness degree prediction method, this embodiment can accurately and objectively predict the car sickness degree of the driver, and provide strong support for subsequent car sickness cause analysis and preventive measures.

[0114] According to the electric vehicle driver car sickness degree prediction method, the present application also provides an electric vehicle driver car sickness degree prediction system, which is shown in Figure 2 , the electric vehicle driver car sickness degree prediction system comprises: Wearable collection device and fixed collection device, the wearable collection device and fixed collection device are used to obtain historical driver physiological data and vehicle operation data.

[0115] The wearable collection device includes an electroencephalograph and a skin electricity collector; the electroencephalograph is used to collect electroencephalogram data, can detect the electrical activity change of the brain, and indirectly reflects the brain nerve activity of the human body in the car sickness state by analyzing the characteristics of the electroencephalogram, such as frequency, amplitude, etc.; the skin electricity collector is used to collect skin conductance data, can measure the change of the skin conductance of the human body, and the index is closely related to the activity of the autonomic nervous system of the human body. In the process of car sickness, with the generation of emotions such as tension and anxiety, the skin conductance will change accordingly, thereby providing a physiological basis for car sickness evaluation.

[0116] The fixed collection device includes an eye tracker, a facial expression collector, and a gyroscope; the eye tracker is used to collect eye movement data, can accurately record the eye movement trajectory, gaze time, blink frequency and other eye movement characteristics of the driver or passenger, and since the balance sense and visual system of the human body are disturbed in car sickness, the eye movement pattern will change, and these eye movement data are helpful for analyzing the influence of car sickness on visual perception and eye movement; the facial expression collector is used to collect facial expression data, and the facial expression collector can capture the subtle changes of facial expressions, and judge whether the driver or passenger appears car sickness related discomfort expressions such as frowning and opening mouth by recognizing and analyzing the facial muscle movement and expression characteristics; the gyroscope is used to collect vehicle running data, and in the embodiment, the gyroscope is mainly used to measure the rotational motion parameters of the vehicle in the driving process, including angular velocity, angular acceleration, etc., and these data can be used to analyze the influence of the dynamic attitude change of the vehicle on the car sickness feeling of the driver or passenger.

[0117] The wearable collection device further includes functional near-infrared spectroscopy (fNIRS), which uses near-infrared light to non-invasively monitor the change of cerebral cortex blood oxygen, and reflects the local neural activity level in real time, thereby providing an objective physiological index of brain function state for car sickness prediction.

[0118] The change of cerebral cortex blood oxygen includes the change of oxygenated hemoglobin concentration HbO and the change of deoxygenated hemoglobin concentration HbR.

[0119] The car sickness data collection control terminal is connected with the wearable collection device and the fixed collection device, is used to perform communication basic operations such as formatting, packaging and encapsulating, and time stamping on the data collected by the wearable collection device and the fixed collection device, and the data collection control terminal establishes a communication connection with the wearable collection device and the fixed collection device through a local area network and a 485 bus.

[0120] The motion sickness online evaluation platform is connected with the motion sickness data collection control terminal, and is used for pre-processing physiological data and vehicle operation data after performing a basic communication operation, obtaining pre-processed physiological data and pre-processed vehicle operation data, and is also used for extracting features from the pre-processed physiological data and pre-processed vehicle operation data, obtaining feature parameters related to motion sickness; the feature parameters related to motion sickness include: average amplitude of an electroencephalogram sequence Mean , power spectral density of a frequency domain signal of an electroencephalogram sequence PSD , electroencephalogram sequence after wavelet transform W , baseline level of a skin conductance signal sequence SCL , power spectral density of a frequency domain signal of a skin conductance signal sequence PSD S , skin conductance signal sequence after wavelet transform W S , number of eye fixation points NOF , average eye fixation time MFD , blink frequency BR , facial expression key point distance d , expression muscle activity intensity MAI , vehicle acceleration variance , and vehicle angular velocity variance ; the feature parameters related to motion sickness are also used for feature fusion to obtain a motion sickness feature vector; the motion sickness feature vector is also divided into a training set, a verification set and a test set; a motion sickness degree prediction model is also constructed; the motion sickness degree prediction model is trained based on the training set to obtain a trained motion sickness degree prediction model; the trained motion sickness degree prediction model is used to predict the motion sickness degree of a to-be-tested electric vehicle driver or passenger to obtain a motion sickness degree prediction result.

[0121] The motion sickness online evaluation platform is connected with the motion sickness data collection control terminal through Ethernet, mobile network and the like, is responsible for receiving motion sickness related data from each terminal, processes and analyzes the data by using an algorithm, generates a motion sickness prediction result, and can feed back the result to the corresponding terminal device, and provides an intuitive motion sickness condition evaluation for vehicle manufacturers, researchers and drivers or passengers.

[0122] The motion sickness evaluation terminal APP is connected with the motion sickness online evaluation platform through 5G cellular communication, used for receiving and displaying the motion sickness degree prediction result, installed on the mobile terminal device of the driver, such as a smart phone, a tablet computer and the like, so as to facilitate the driver to record the motion sickness feeling in real time during the driving of the electric vehicle, including but not limited to the type, degree, occurrence time and the like of the motion sickness symptoms, and upload the subjective feeling data to the motion sickness online evaluation platform, to provide personalized motion sickness prevention suggestions and measures for the driver.

[0123] The motion sickness degree prediction system for the driver of the electric vehicle relates to the present embodiment. In actual application, first, fixed collection devices such as an eye tracker, a facial expression collector, a gyroscope and the like are installed and deployed in the electric vehicle, and it is ensured that the fixed collection devices are connected with the motion sickness data collection control terminal through a local area network or a 485 bus to establish a stable communication connection. At the same time, the driver wears wearable collection devices such as an electroencephalograph, a skin electricity collector and the like, and the wearable collection devices are also connected with the motion sickness data collection control terminal. When the electric vehicle starts to drive, the fixed collection devices and the wearable collection devices start the data collection work synchronously, respectively acquire the running state data of the vehicle and the physiological characteristic data of the driver, and transmit the data to the motion sickness data collection control terminal in real time.

[0124] After the motion sickness data collection control terminal processes and operates the received data, the data is transmitted to the motion sickness online evaluation platform through Ethernet or a mobile network. At the same time, the driver can open the motion sickness evaluation terminal App at any time during the driving to record the motion sickness symptoms such as the degree of dizziness, whether there is a feeling of nausea and the like, and submit to the motion sickness online evaluation platform.

[0125] After the motion sickness online evaluation platform receives the various types of data, the data is processed according to a preset algorithm. First, the data is preprocessed to remove noise and interference components in the data. Then, the characteristic parameters related to the motion sickness are extracted, such as the power spectral density and the characteristic frequency extracted from the electroencephalogram data, the skin conductance level change rate extracted from the skin electricity data, and the eye movement trajectory stability parameter extracted from the eye movement data. Then, the characteristic parameters are fused to construct a motion sickness characteristic vector, and input into a machine learning model trained in advance to obtain the current motion sickness degree evaluation result of the driver, for example, to determine that the driver is in mild motion sickness. At the same time, the change relationship between the vehicle running data and the physiological characteristic parameters is analyzed to analyze that the motion sickness is probably caused by the frequent jolt of the vehicle at a certain section, which leads to the imbalance of the human body balance and further causes the motion sickness symptoms.

[0126] The platform feeds back the motion sickness degree evaluation result and the motion sickness cause analysis report to the motion sickness evaluation terminal App, and the driver can check the specific situation and the cause of the motion sickness through the motion sickness evaluation terminal App, so as to take corresponding preventive measures, such as adjusting the sitting posture, avoiding reading in the bumpy road section, etc. In addition, the vehicle manufacturer can also collect a large amount of motion sickness data of the driver through accessing the motion sickness online evaluation platform, and deeply analyze the motion sickness situation of different vehicle models and different driving conditions, so as to optimize the design of the electric vehicle, such as improving the suspension system and adjusting the power output characteristics, so as to reduce the probability of the driver's motion sickness, and improve the comfort and market competitiveness of the electric vehicle.

[0127] The application provides a method and system for predicting the degree of motion sickness of the driver of an electric vehicle, which comprehensively considers the unique properties of the electric vehicle, accurately and objectively evaluates the degree of motion sickness of the driver, and deeply analyzes the causes of motion sickness and the related influencing factors, thereby providing strong technical support for the vehicle manufacturers to optimize the vehicle design, improve the driving comfort and ensure the driving safety, and has application value and market prospect.

[0128] The technical features of the above embodiments can be combined arbitrarily, and to make the description concise, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combinations of the technical features do not exist contradictions, they should be considered as the scope of the present application.

[0129] The principles and implementation modes of the application are described by using specific examples in the present application, and the above embodiment is only used to help understand the method and the core idea of the application; meanwhile, according to the idea of the application, the specific implementation mode and the application range can be changed by the person skilled in the art. In conclusion, the content of the present application should not be understood as the limitation of the application.

Claims

1. A method for predicting the degree of motion sickness of electric vehicle drivers and passengers, characterized in that: The method for predicting the degree of motion sickness of electric vehicle drivers and passengers comprises: Obtain historical driver and passenger physiological data and vehicle operation data; Preprocessing the physiological data and the vehicle operation data to obtain preprocessed physiological data and preprocessed vehicle operation data; Feature extraction is performed on the pre-processed physiological data and the pre-processed vehicle operation data to obtain feature parameters related to motion sickness; the feature parameters related to motion sickness include: the average amplitude of the EEG signal sequence Mean , power spectral density of frequency domain signal of EEG signal sequence PSD , EEG signal sequence after wavelet transform W , skin conductance signal sequence baseline level SCL , power spectral density of frequency domain signal of skin conductance signal sequence PSD S , skin conductance signal sequence after wavelet transform W S , number of eye fixations NOF , average eye gaze time MFD , blink frequency BR , facial expression key point distance d , intensity of facial muscle activity MAI , vehicle acceleration variance and the vehicle angular velocity variance ; Performing feature fusion on the feature parameters related to motion sickness to obtain a motion sickness feature vector; Dividing the motion sickness feature vector into a training set, a validation set, and a test set; Build a motion sickness prediction model; Training the motion sickness degree prediction model based on the training set to obtain a trained motion sickness degree prediction model; The trained motion sickness prediction model is used to predict the motion sickness degree of the electric vehicle driver and passenger to obtain the motion sickness degree prediction result.

2. The method for predicting the degree of motion sickness of electric vehicle drivers and passengers according to claim 1, characterized in that: The physiological data specifically includes: EEG data, skin conductance data, eye movement data and facial expression data.

3. The method for predicting the degree of motion sickness of electric vehicle drivers and passengers according to claim 1, characterized in that: The physiological data and the vehicle operation data are preprocessed, specifically: filtering, denoising and normalizing the physiological data and the vehicle operation data.

4. The method for predicting the degree of motion sickness of electric vehicle drivers and passengers according to claim 2, characterized in that: Feature extraction is performed on the preprocessed physiological data and the preprocessed vehicle operation data to obtain feature parameters related to motion sickness, specifically comprising the following steps: Based on the preprocessed EEG data, calculate the average amplitude of the EEG signal sequence Mean , power spectral density of frequency domain signal of EEG signal sequence PSD And the EEG signal sequence after wavelet transformation W ; Calculate the baseline level of the skin conductance signal sequence based on the preprocessed skin conductance data SCL , power spectral density of frequency domain signal of skin conductance signal sequence PSD S And the skin conductance signal sequence after wavelet transformation W S ; Calculate the number of eye fixations based on preprocessed eye movement data NOF , average eye gaze time MFD and blink frequency BR ; Calculate the distance between facial expression key points based on the preprocessed facial expression data d and facial muscle activity intensity MAI ; Calculate the vehicle acceleration variance based on the pre-processed vehicle operation data and the vehicle angular velocity variance .

5. The method for predicting the degree of motion sickness of electric vehicle drivers and passengers according to claim 1, characterized in that: Performing feature fusion on the feature parameters related to motion sickness to obtain a motion sickness feature vector specifically includes the following steps: The average amplitude of the EEG signal sequence Mean , power spectral density of frequency domain signal of EEG signal sequence PSD And the EEG signal sequence after wavelet transformation W , perform weighting and fitting to obtain the EEG feature vector; Baseline level of the skin conductance signal sequence SCL , power spectral density of frequency domain signal of skin conductance signal sequence PSD S And the skin conductance signal sequence after wavelet transformation W S , perform weighting and fitting to obtain the skin conductance feature vector; Number of eye fixations NOF , average eye gaze time MFD and blink frequency BR , perform weighting and fitting to obtain the eye movement feature vector; The distance between the facial expression key points d and facial muscle activity intensity MAI , perform weighting and fitting to obtain the facial expression feature vector; The vehicle acceleration variance and the vehicle angular velocity variance Perform weighting and fitting to obtain the vehicle motion sickness feature vector; The EEG feature vector, the skin conductance feature vector, the eye movement feature vector, the facial expression feature vector and the vehicle motion sickness feature vector are weighted and fitted to obtain a comprehensive feature vector F ; The comprehensive feature vector F Normalize the elements in ; determining the sensitivity and contribution of the driver's motion sickness state according to the EEG feature vector, the skin conductance feature vector, the eye movement feature vector, the facial expression feature vector, and the vehicle motion sickness feature vector; determining, based on the sensitivity and contribution of the driver and passenger to motion sickness, a feature weight of the EEG feature vector, a feature weight of the skin conductance feature vector, a feature weight of the eye movement feature vector, a feature weight of the facial expression feature vector, and a feature weight of the vehicle motion sickness feature vector; The normalized comprehensive feature vector is multiplied by the feature weight of the EEG feature vector, the feature weight of the skin conductance feature vector, the feature weight of the eye movement feature vector, the feature weight of the facial expression feature vector, and the feature weight of the vehicle motion sickness feature vector to obtain a weighted feature vector F weighted , the weighted feature vector is the motion sickness feature vector.

6. The method for predicting the degree of motion sickness of electric vehicle drivers and passengers according to claim 5, characterized in that: The EEG eigenvector formula is as follows: ； in, is the EEG feature vector, n The number of historical drivers and passengers for whom EEG data was collected, is the EEG characteristic of the first driver and passenger, is the EEG characteristic of the second driver and passenger, For the n EEG characteristics of the drivers and passengers; The skin conductance eigenvector formula is as follows: ； in, is the skin conductance eigenvector, m The number of historical drivers and passengers for whom skin conductance data was collected, is the skin conductance characteristic of the first driver and passenger, is the skin conductance characteristic of the second driver and passenger, For the m Skin conductance characteristics of the drivers and passengers; The eye movement feature vector formula is as follows: ； in, is the eye movement feature vector, p The number of historical drivers and passengers for collecting eye movement data, is the eye movement characteristic of the first driver and passenger, is the eye movement characteristics of the second driver and passenger, For the p Eye movement characteristics of drivers and passengers; The facial expression feature vector formula is as follows: ； in, is the facial expression feature vector, q The number of historical drivers and passengers for collecting facial expression data, is the facial expression feature of the first driver and passenger, is the facial expression feature of the second driver, For the q Facial expression characteristics of the driver and passengers; The vehicle motion sickness eigenvector formula is as follows: ； in, is the vehicle motion sickness feature vector, r To collect the historical number of drivers and passengers in the vehicle operation data, is the vehicle motion sickness characteristic of the first driver and passenger in the vehicle, For the r Vehicle motion sickness characteristics of drivers and passengers in vehicles; The comprehensive feature vector F The formula is as follows: ； The formulas for the feature weights of the EEG feature vector, the skin conductance feature vector, the eye movement feature vector, the facial expression feature vector, and the vehicle motion sickness feature vector are as follows: ； ； ； ； ； in, is the feature weight of the EEG feature vector, is the feature weight of the skin conductance feature vector, is the feature weight of the eye movement feature vector, is the feature weight of the facial expression feature vector, is the feature weight of the vehicle motion sickness feature vector, For the n The feature weights assigned to the EEG features of the drivers and passengers are For the m The feature weights assigned to the skin conductance features of the drivers and passengers are For the p The feature weights assigned to the eye movement features of the drivers and passengers are For the q The feature weights assigned to the facial expression features of the drivers and passengers are For the r The feature weights assigned to the vehicle motion sickness features of the vehicle in which the driver and passengers are riding; The weighted feature vector F weighted The formula is as follows: ； in, F weighted represents the weighted eigenvector, ⊙ represents element-wise multiplication, represents the normalized EEG feature vector, represents the normalized skin conductance feature vector, represents the normalized eye movement feature vector, represents the normalized facial expression feature vector, Represents the normalized vehicle motion sickness feature vector.

7. The method for predicting the degree of motion sickness of electric vehicle drivers and passengers according to claim 1, characterized in that: The motion sickness degree prediction model includes: An input layer, configured to receive the motion sickness feature vector; a convolution layer connected to the input layer, the convolution layer comprising convolution kernels, each convolution kernel being used to extract local features of the motion sickness feature vector; A pooling layer connected to the convolutional layer, wherein the pooling layer is used to reduce the dimension of the motion sickness feature vector; An LSTM layer, connected to the pooling layer, for learning how motion sickness features change over time; A fully connected layer, connected to the LSTM layer, for weighting and processing the features output by the LSTM layer, and mapping the processed features to a classification space of motion sickness degree; The output layer is connected to the fully connected layer, and the output layer uses a softmax activation function to convert the features output by the fully connected layer into a probability distribution.

8. A system for predicting the degree of motion sickness of electric vehicle drivers and passengers, characterized in that: The electric vehicle driver and passenger motion sickness prediction system comprises: Wearable data collection equipment and fixed data collection equipment, which are used to obtain historical driver and passenger physiological data and vehicle operation data; An online motion sickness evaluation platform is connected to the wearable data acquisition device and the fixed data acquisition device, and is used to pre-process the physiological data and the vehicle operation data to obtain pre-processed physiological data and pre-processed vehicle operation data, and is also used to extract features from the pre-processed physiological data and pre-processed vehicle operation data to obtain feature parameters related to motion sickness; the feature parameters related to motion sickness include: the average amplitude of the EEG signal sequence Mean , power spectral density of frequency domain signal of EEG signal sequence PSD , EEG signal sequence after wavelet transform W , skin conductance signal sequence baseline level SCL , power spectral density of frequency domain signal of skin conductance signal sequence PSD S , skin conductance signal sequence after wavelet transform W S , number of eye fixations NOF , average eye gaze time MFD , blink frequency BR , facial expression key point distance d , intensity of facial muscle activity MAI , vehicle acceleration variance and the vehicle angular velocity variance ; It is also used to perform feature fusion on the feature parameters related to motion sickness to obtain a motion sickness feature vector; It is also used to divide the motion sickness feature vector into a training set, a validation set and a test set; It is also used to construct a motion sickness degree prediction model; It is also used to train the motion sickness degree prediction model based on the training set to obtain a trained motion sickness degree prediction model; It is also used to use the trained motion sickness degree prediction model to predict the motion sickness degree of the electric vehicle driver and passenger to be tested to obtain a motion sickness degree prediction result.

9. The system for predicting the degree of motion sickness of electric vehicle drivers and passengers according to claim 8, characterized in that: The electric vehicle driver and passenger motion sickness degree prediction system further includes: a motion sickness evaluation terminal APP, which is connected to the motion sickness online evaluation platform and is used to receive and display the motion sickness degree prediction result.

10. The system for predicting the degree of motion sickness of electric vehicle drivers and passengers according to claim 8, characterized in that: The wearable data acquisition device includes: an electroencephalogram (EEG) device and a skin charge collector; The electroencephalogram (EEG) device is used to collect EEG data; The skin conductance collector is used to collect skin conductance data; The fixed acquisition equipment includes: an eye tracker, a facial expression collector and a gyroscope; The eye tracker is used to collect eye movement data; The facial expression collector is used to collect facial expression data; The gyroscope is used to collect vehicle operation data.

Images