Sound-light warning interaction method considering driving style influence

Abstract

The application provides a sound-light warning interaction method considering the influence of driving style, which is based on the electromyographic signals and vehicle signals of the driver in a simulated driving environment, adopts a K-means clustering recognition algorithm to mark the driving style as conservative, normal and aggressive, and combines an entropy weight method and a TOPSIS method to score the driving safety. The trust and acceptance of the driver to the warning strategy are obtained through a two-way MANCOVA analysis, the parameters of the sound-light warning strategy are adjusted based on the trust and acceptance, and finally, experimental verification is carried out. The results show that the sound-light warning strategy considering the driving style can effectively improve the driving safety of drivers with different styles.

Description

Technical Field

[0001] This invention belongs to the field of driving behavior and driving style recognition, and specifically designs an audio-visual warning interaction method that takes into account the influence of driving style. Background Technology

[0002] With the rapid development of the automotive industry, frequent traffic accidents have become a serious problem for countries worldwide. Existing research indicates that traffic accidents are primarily caused by human error. This necessitates efficient driver warning systems to help drivers avoid accidents. A driver's driving style is highly correlated with driving safety, and drivers with different driving styles experience warning systems differently. Current warning strategies largely fail to meet the needs of drivers with different styles, leading to numerous errors in alerting to dangerous driving behaviors and reducing public trust in warning systems. Therefore, researching driver warning strategies that consider the influence of driving style is essential. Summary of the Invention

[0003] To address the aforementioned problems, this invention overcomes the shortcomings of existing methods and proposes an audio-visual warning interaction method that considers driving style. This method collects drivers' electromyographic signals and driving data through driving simulation experiments, uses the K-means clustering algorithm to categorize driving styles as conservative, normal, and average, and combines entropy weighting and TOPSIS methods to conduct a comprehensive safety assessment of the driving behavior of drivers with different styles. Finally, it proposes corresponding audio-visual warning strategies for different driving styles.

[0004] To achieve the above-mentioned technical features, the objective of this invention is as follows: a sound-light warning interaction method considering the influence of driving style, comprising the following steps:

[0005] Step 1: Build a driver-in-the-loop system and have the participants complete driving simulation experiments in turn, collecting their electromyography data and driving data.

[0006] Step 2: First, preprocess the electromyography data and driving data collected in Step 1; then, slice the preprocessed electromyography data and driving data into driving segments according to the set time period to obtain the required segments, and extract data from the segments to construct feature parameters for subsequent driving style recognition.

[0007] Step 3: Principal component analysis is used to reduce the dimensionality of the selected feature parameters, retaining the principal components with eigenvalues ​​greater than 1. K-means clustering is used to classify the retained feature values, and the multiple driving feature parameters are clustered into three driving style types: conservative, normal, and aggressive.

[0008] Step 4: Combine the entropy weight method and the TOPSIS method to establish a driving behavior evaluation system and give a comprehensive safety score to the driving behavior of drivers with different styles.

[0009] Step 5: Propose corresponding audible and visual warning strategies for drivers with different driving styles.

[0010] In step 1, the driver-in-the-loop system includes a real-time simulator and a host computer PC. Each of the real-time simulator and the host computer PC runs a Simulink model. Speedgoat runs a vehicle model, and the host computer PC runs a driving scenario-related model. Finally, the real-time simulator realizes the synchronous acquisition of electromyographic signals and vehicle status information.

[0011] The Simulink model in the host PC can receive steering, throttle and brake control signals from the driver. The host PC then inputs the control signals and rolling resistance and aerodynamic resistance information to the real-time simulator via UDP communication. The real-time simulator calculates the vehicle's state based on this information and sends it to the host PC, synchronously collecting the driver's physiological signals and vehicle signals.

[0012] The driving data includes steering wheel angle, lateral and longitudinal speed, lateral and longitudinal acceleration, impact, accelerator pedal signal, and brake pedal signal;

[0013] The electromyographic data included electromyographic signals on the surface of the gastrocnemius muscle and the right sternocleidomastoid muscle.

[0014] Step 2 involves extracting data from the slices, including: selecting the average and standard deviation of the driver's speed, absolute acceleration, absolute impact, accelerator pedal opening, and rate of change of accelerator pedal opening, as well as the mean and median of muscle activation and electromyographic amplitude of the gastrocnemius and right sternocleidomastoid muscles to construct feature parameters for subsequent driving style identification.

[0015] In step 3, safety evaluation indicators are selected to determine driver operation behavior indicators and vehicle status information indicators from both vertical and horizontal perspectives.

[0016] Longitudinal velocity standard deviation:

[0017] ;

[0018] In the formula, V std The standard deviation of longitudinal velocity; V i This represents the actual value of the longitudinal velocity. This represents the average longitudinal velocity.

[0019] Longitudinal acceleration standard deviation:

[0020] ;

[0021] In the formula, It is the standard deviation of longitudinal acceleration; This is the actual value of longitudinal acceleration; It is the average longitudinal acceleration;

[0022] Longitudinal acceleration rate of change:

[0023] ;

[0024] In the formula, It is the rate of change of longitudinal acceleration. This is the actual value of longitudinal acceleration. It is the length of the sampling time period;

[0025] Steering wheel angle entropy value:

[0026] ;

[0027] In the formula, SE It is the entropy value of the steering wheel angle. The probability distribution value of each interval is determined by the frequency of the error value falling within each interval;

[0028] Steering wheel angle standard deviation:

[0029] ;

[0030] In the formula, It is the standard deviation of the steering wheel angle; It is the actual value of the steering wheel angle; It is the average value of the steering wheel angle over one cycle;

[0031] lateral acceleration standard deviation:

[0032] ;

[0033] In the formula, It is the standard deviation of lateral acceleration; This is the actual value of lateral acceleration; It is the average lateral acceleration;

[0034] Root mean square value of electromyographic signal:

[0035] ;

[0036] In the formula, It is the root mean square value of the electromyography signal. It is the length of the time period. It refers to the EMG data collected within this time period;

[0037] Average frequency of electromyographic signals:

[0038] ;

[0039] In the formula, It is the average power of the electromyographic signal. At frequency Power spectral density at that point It's frequency. It is the maximum frequency of the signal.

[0040] The steps for determining the objective weights of the indicators using the entropy method in step 4 are as follows:

[0041] Step 4.1.1, calculate the first... j Under this indicator, the first i The feature weight of each object:

[0042] ;

[0043] Step 4.1.2, calculate the first... j Entropy value of the indicator:

[0044] ;

[0045] Step 4.1.3, calculate the coefficient of variation of the indicators:

[0046] ;

[0047] Step 4.1.4, Determine the objective weights:

[0048] .

[0049] It has m One driver, n Evaluation indicators C n , X ij It is the first i The first driver's j Item indicator value;

[0050] The steps for multi-attribute decision-making using the TOPSIS method in step 4 are as follows:

[0051] Step 4.2.1: Establish a decision matrix based on the original data:

[0052] (13);

[0053] Step 4.2.2: Normalize the data for different indicators to obtain the normalized matrix σ.

[0054] (14);

[0055] Step 4.2.3: Assign weights to each indicator, and finally obtain the evaluation matrix V:

[0056] (15);

[0057] in, , ;

[0058] Step 4.2.4 yields the ideal solution. and negative ideal solution :

[0059] , (16);

[0060] If the indicator is a very large indicator, then:

[0061] (17);

[0062] If the indicator is a very small indicator, then:

[0063] (18);

[0064] Step 4.2.5: Calculate the distance between each experimenter and the positive and negative ideal solutions:

[0065] ;

[0066] ;

[0067] Step 4.2.6, calculate the score for each experimenter:

[0068] ;

[0069] Score The closer a value is to 1, the safer the driving behavior.

[0070] In step 5, a head-up display device is used to project different colored signal lights to achieve light warning. As the driver's safety score changes from 1 to 0, the color of the signal light will also change from green to yellow, orange, and red. According to the color vision principle of ergonomics, red light is used to represent a warning signal; yellow and orange lights are used to represent warning signals; and green light is used to represent normal operation.

[0071] Step 5 uses two-way MANCOVA analysis to obtain the impact of different driver styles on the trust and acceptance of the warning strategy, and introduces adjustment parameters: , , , And the optimal values ​​of these parameters are determined using the particle swarm optimization algorithm:

[0072] Initial slope of the model and intercept Given known fixed values, the two external influencing factors are confidence levels. and acceptance For model parameters and To generate an effect, the following parameter adjustment model is defined:

[0073] ;

[0074] ;

[0075] The particle swarm optimization algorithm is used to minimize the mean squared error between the model's predicted values ​​and the actual observed values. The objective function for optimization is defined as follows:

[0076] ;

[0077] in, It is the output of the model prediction. These are actual observational data.

[0078] The specific functional relationship of the adjusted parameters:

[0079] Conservative type:

[0080] ;

[0081] ;

[0082] Normal type:

[0083] ;

[0084] ;

[0085] Radical type:

[0086] ;

[0087] ;

[0088] In this strategy, after parameter adjustment, if If the value is less than 500, the light warning system will still show a green light. If the value is greater than 500, the light warning system will still show a red light; if If the value is less than 500, the buzzer frequency of the sound warning system will still be 500Hz. If the value is greater than 2000, the buzzer frequency of the audible warning system will remain at 2000Hz; after adjustment, the threshold for a conservative driver to trigger the audible warning is... The threshold for a normal driver to trigger an audible warning is: The threshold for an aggressive driver to trigger an audible warning is: .

[0089] The present invention has the following beneficial effects:

[0090] 1. Mainstream driving style recognition methods only use single vehicle operation and running status signals. The method of this invention integrates the driver's physiological signals with the vehicle's operation and running status, increasing the objective evaluation of the driver's own feelings. Physiological signals can not only serve as a basis for classifying different driving styles, but also reflect the differences in driving behavior between drivers of different styles.

[0091] 2. Driving style is used to characterize a driver's behavioral characteristics in operating a vehicle. In actual driving, drivers may exhibit different driving styles due to subjective and objective factors such as road conditions, emotions, and time costs. This method first identifies the driving style at different time intervals throughout the entire driving process, thereby making a precise judgment on driving behavior and accurately identifying the driving style with the highest safety.

[0092] 3. This invention proposes corresponding sound and light warning strategies for drivers with different driving styles, which greatly improves the trust and acceptance of the warning strategies by drivers with different styles. Attached Figure Description

[0093] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0094] Figure 1 The present invention provides a logical block diagram of an audio-visual early warning strategy.

[0095] Figure 2 This is a signal flow diagram of the driver-in-the-loop system constructed according to the present invention.

[0096] Figure 3 (a) and (b) show the locations of the gastrocnemius and right sternocleidomastoid muscles.

[0097] Figure 4 (a) and (b) are schematic diagrams comparing the original electromyographic signals acquired in this invention with the electromyographic signals obtained after preprocessing with a Butterworth filter.

[0098] Figure 5 This is a graph showing the results of K-means clustering.

[0099] Figure 6 The driving behavior safety score is obtained by combining the entropy weight method and the TOPSIS method.

[0100] Figure 7 This shows the change in safety score before and after the warning. Detailed Implementation

[0101] The embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0102] Example 1:

[0103] A sound-and-light warning interaction method that takes into account the influence of driving style includes the following steps:

[0104] Step 1, build as follows Figure 2 The driver-in-the-loop system shown allows participants to sequentially complete driving simulation experiments. Electromyography (EMG) and driving data are collected from these participants. Driving data includes steering wheel angle, lateral and longitudinal velocity, lateral and longitudinal acceleration, impact force, accelerator pedal signal, and brake pedal signal. EMG data includes surface EMG signals from the gastrocnemius muscle and the right sternocleidomastoid muscle. The gastrocnemius and right sternocleidomastoid muscles are shown below. Figure 3 As shown.

[0105] Step 2: First, preprocess the electromyographic signals acquired in Step 1 to remove noise. Then, slice the raw data acquired in Step 1 into 8-second segments to obtain the required driving segments. Select the average and standard deviation of the driver's speed, absolute acceleration, absolute impact, accelerator pedal opening, and rate of change of accelerator pedal opening, as well as the mean and median electromyographic amplitude of the gastrocnemius and right sternocleidomastoid muscles, as parameters for driving style classification and recognition.

[0106] Step 3: Principal component analysis (PCA) is used to reduce the dimensionality of the selected feature parameters, retaining principal components with eigenvalues ​​greater than 1. K-means clustering is then used to classify the experimental data, categorizing experimenters into conservative, normal, and aggressive types.

[0107] Step 4: Combine the entropy weight method and the TOPSIS method to establish a driving behavior evaluation system and give a comprehensive safety score to the driving behavior of drivers with different styles.

[0108] Step 5: Propose corresponding audible and visual warning strategies for drivers with different driving styles.

[0109] Example 2:

[0110] To address the shortcomings of existing driving warning methods, this invention proposes an audio-visual warning interaction method that considers driving style. The specific implementation method is as follows:

[0111] 1. Set up as follows Figure 2The driver-in-the-loop system consists of a real-time simulator and a host PC, each running a Simulink model. Speedgoat runs the vehicle model, and the host PC runs driving scenario-related models. Finally, synchronous acquisition of electromyography (EMG) signals and vehicle state information is achieved on the real-time simulation platform. The driver-in-the-loop system and its signal flow are as follows: Figure 2 As shown, the Simulink model in the PC can receive steering, accelerator, and brake signals from the driver. The PC then uses UDP communication to input the control signals and information such as rolling resistance and aerodynamic drag into the real-time simulation platform. The real-time simulator calculates the vehicle's state based on this information and sends it to the host PC. The driver's physiological signals and vehicle signals are collected synchronously.

[0112] 2. First, the collected electromyography (EMG) and driving data are preprocessed. The driving data is cleaned to remove unreasonable data and ensure all parameters are within reasonable ranges. For the EMG signals, a Butterworth 6th-order filter is used for bandpass filtering from 10 to 450 Hz, and a 50 Hz notch filter is used to eliminate power frequency interference. Median smoothing is also performed. The preprocessed data is as follows: Figure 4 As shown, the preprocessed data was sliced ​​into 8-second segments representing driving conditions. The average and standard deviation of the driver's speed, absolute acceleration, absolute impact, accelerator pedal opening, and rate of change of accelerator pedal opening, as well as the mean and median electromyographic amplitude of the gastrocnemius and right sternocleidomastoid muscles, were selected to construct feature parameters for subsequent driving style identification.

[0113] 3. Principal Component Analysis (PCA) was used to reduce the dimensionality of the constructed feature parameters, thereby decreasing data complexity. Principal components with eigenvalues ​​greater than 1 were selected, while those with eigenvalues ​​less than 1 were discarded. K-means clustering analysis was then used to cluster the multiple driving feature parameters into three driving style types: conservative, normal, and aggressive. The clustering results are shown below. Figure 5 As shown, a driver's driving style is determined by the composition of their driving segments. For example, if a driver's driving segments comprise the largest proportion of their total driving segments, then that driver is a normal driving type.

[0114] 4. Select safety evaluation indicators, and determine driver operation behavior indicators and vehicle status information indicators from both longitudinal and horizontal perspectives;

[0115] Longitudinal velocity standard deviation:

[0116] ;

[0117] In the formula, V std The standard deviation of longitudinal velocity; V iThis represents the actual value of the longitudinal velocity. This represents the average longitudinal velocity.

[0118] Longitudinal acceleration standard deviation:

[0119] ;

[0120] In the formula, It is the standard deviation of longitudinal acceleration; This is the actual value of longitudinal acceleration; It is the average longitudinal acceleration;

[0121] Longitudinal acceleration rate of change:

[0122] ;

[0123] In the formula, It is the rate of change of longitudinal acceleration. This is the actual value of longitudinal acceleration. It is the length of the sampling time period;

[0124] Steering wheel angle entropy value:

[0125] ;

[0126] In the formula, SE It is the entropy value of the steering wheel angle. The probability distribution value of each interval is determined by the frequency of the error value falling within each interval;

[0127] Steering wheel angle standard deviation:

[0128] ;

[0129] In the formula, It is the standard deviation of the steering wheel angle; It is the actual value of the steering wheel angle; It is the average value of the steering wheel angle over one cycle;

[0130] lateral acceleration standard deviation:

[0131] ;

[0132] In the formula, It is the standard deviation of lateral acceleration; This is the actual value of lateral acceleration; It is the average lateral acceleration;

[0133] Root mean square value of electromyographic signal:

[0134] ;

[0135] In the formula, It is the root mean square value of the electromyography signal. It is the length of the time period. It refers to the EMG data collected within this time period;

[0136] Average frequency of electromyographic signals:

[0137] ;

[0138] In the formula, It is the average power of the electromyographic signal. At frequency Power spectral density at that point It's frequency. It is the maximum frequency of the signal.

[0139] 5. Combining the entropy weight method and the TOPSIS method to score driving safety.

[0140] The steps for determining the objective weights of indicators using the entropy method are as follows:

[0141] (1) Calculate the first j Under this indicator, the first i The feature weight of each object:

[0142] ;

[0143] (2) Calculate the first j Entropy value of the indicator:

[0144] ;

[0145] (3) Calculate the coefficient of variation of the indicators:

[0146] ;

[0147] (4) Determine objective weights:

[0148] .

[0149] It has m One driver, n Evaluation indicators C n , X ij It is the first i The first driver's j Item indicator value;

[0150] The steps for multi-attribute decision-making using the TOPSIS method in step 4 are as follows:

[0151] (1) Establish a decision matrix based on the original data:

[0152] (13);

[0153] (2) Normalize the data for different indicators to obtain the normalized matrix σ:

[0154] (14);

[0155] (3) Assign weights to each indicator to obtain the final evaluation matrix V:

[0156] (15);

[0157] in, , ;

[0158] (4) Obtain the positive ideal solution and negative ideal solution :

[0159] , (16);

[0160] If the indicator is a very large indicator, then:

[0161] (17);

[0162] If the indicator is a very small indicator, then:

[0163] (18);

[0164] (5) Calculate the distance between each experimenter and the positive and negative ideal solutions:

[0165] ;

[0166] ;

[0167] (6) Calculate the scores of each experimenter:

[0168] ;

[0169] Score The closer a value is to 1, the safer the driving behavior.

[0170] 6. Previous vehicle visual warning systems mostly displayed warning information on the main control panel and instrument panel. However, head-up displays (HUDs) can optimize driver attention allocation. Therefore, this invention uses HUDs to project different colored signal lights to achieve light-based warnings. As the driver's safety score changes from 1 to 0, the signal light color changes from green to yellow, orange, and red. Green light has a wavelength of 500–570 nm, yellow light 570–590 nm, orange light 590–620 nm, and red light 620–700 nm. According to the principles of color vision in ergonomics, red evokes a strong emotional response, making people highly alert; therefore, a red light represents a warning signal. Yellow and orange can remind people to be more alert, but do not necessarily require immediate stopping or emergency measures; therefore, yellow and orange lights represent warning signals. Green is usually associated with concepts such as safety, permission, and normal operation, indicating that it is permissible to continue or that everything is normal; therefore, a green light represents normal operation.

[0171] 7. Two-way MANCOVA analysis was used to obtain the impact of different driving styles on the trust and acceptance of warning strategies. In order to enhance the adaptability of the model and enable it to dynamically respond to changes in the trust and acceptance of warning methods by drivers of different driving styles as statistically analyzed in the previous section, an adjustment parameter was introduced. , , , The optimal values ​​of these parameters are determined using a particle swarm optimization algorithm.

[0172] Initial slope of the model and intercept Given known fixed values, the two external influencing factors are confidence levels. and acceptance For model parameters and To generate an effect, the following parameter adjustment model is defined:

[0173] ;

[0174] ;

[0175] The particle swarm optimization algorithm is used to minimize the mean squared error between the model's predicted values ​​and the actual observed values. The objective function for optimization is defined as follows:

[0176] ;

[0177] in, It is the output of the model prediction. These are actual observational data.

[0178] In summary, by introducing adjustment parameters and using the particle swarm optimization algorithm, we can adaptively adjust the parameters of the linear model to achieve the best fit under different driving style influences. The specific functional relationships of the adjusted parameters are shown below.

[0179] Conservative type:

[0180] ;

[0181] ;

[0182] Normal type:

[0183] ;

[0184] ;

[0185] Radical type:

[0186] ;

[0187] ;

[0188] In this strategy, after parameter adjustment, if If the value is less than 500, the light warning system will still show a green light. If the value is greater than 500, the light warning system will still display a red light. If the value is less than 500, the buzzer frequency of the sound warning system will still be 500Hz. If the value is greater than 2000, the buzzer frequency of the audible warning system will remain at 2000Hz. After adjustment, the threshold for a conservative driver to trigger the audible warning is... , The threshold for a normal driver to trigger an audible warning is: , The threshold for an aggressive driver to trigger an audible warning is: , The early warning strategy logic diagram is as follows: Figure 1 As shown. The security scores before and after the warning are as follows: Figure 7 As shown.

Claims

1. A sound-light warning interaction method considering the influence of driving style, characterized in that, Includes the following steps: Step 1: Build a driver-in-the-loop system and have the participants complete driving simulation experiments in turn, collecting their electromyography data and driving data. Step 2: First, preprocess the electromyography data and driving data collected in Step 1; The preprocessed electromyography data and driving data are then sliced ​​into driving segments according to a set time period to obtain the required segments, and data is extracted from the segments to construct feature parameters for subsequent driving style recognition. Step 3: Principal component analysis is used to reduce the dimensionality of the selected feature parameters, retaining the principal components with eigenvalues ​​greater than 1. K-means clustering is used to classify the retained feature values, and the multiple driving feature parameters are clustered into three driving style types: conservative, normal, and aggressive. Step 4: Combine the entropy weight method and the TOPSIS method to establish a driving behavior evaluation system and give a comprehensive safety score to the driving behavior of drivers with different styles. Step 5: Propose corresponding audible and visual warning strategies for drivers with different driving styles; In step 1, the driver-in-the-loop system includes a real-time simulator and a host computer PC. Each of the real-time simulator and the host computer PC runs a Simulink model. Speedgoat runs a vehicle model, and the host computer PC runs a driving scenario-related model. Finally, the real-time simulator realizes the synchronous acquisition of electromyographic signals and vehicle status information. The Simulink model in the host PC can receive steering, throttle and brake control signals from the driver. The host PC then inputs the control signals and rolling resistance and wind resistance information to the real-time simulator via UDP communication. The real-time simulator calculates the vehicle's state based on this information and sends it to the host PC, synchronously collecting the driver's physiological signals and vehicle signals. Electromyographic data include surface electromyographic signals of the gastrocnemius muscle and the right sternocleidomastoid muscle; Step 5 uses two-way MANCOVA analysis to obtain the impact of different driver styles on the trust and acceptance of the warning strategy, and introduces adjustment parameters: , , , And the optimal values ​​of these parameters are determined using the particle swarm optimization algorithm: model initial slope and intercept Given known fixed values, the two external influencing factors are confidence levels. and acceptance For model parameters and To generate an effect, the following parameter adjustment model is defined: ； ； The particle swarm optimization algorithm is used to minimize the mean squared error between the model's predicted values ​​and the actual observed values. The objective function for optimization is defined as follows: ； in, It is the output of the model prediction. These are actual observational data; In step 5, a head-up display device is used to project different colored signal lights to achieve light warning. As the driver's safety score changes from 1 to 0, the color of the signal light will also change from green to yellow, orange, and red. According to the color vision principle of ergonomics, red light is used to represent a warning signal; yellow and orange lights are used to represent warning signals; and green light is used to represent normal operation. The specific functional relationship of the adjusted parameters: Conservative type: ； ； Normal type: ； ； Radical type: ； ； In this strategy, after parameter adjustment, if If the value is less than 500, the light warning system will still show a green light. If the value is greater than 500, the light warning system will still show a red light; if If the value is less than 500, the buzzer frequency of the sound warning system will still be 500Hz. If the value is greater than 2000, the buzzer frequency of the audible warning system will remain at 2000Hz; after adjustment, the threshold for a conservative driver to trigger the audible warning is... The threshold for a normal driver to trigger an audible warning is: The threshold for an aggressive driver to trigger an audible warning is: .

2. The sound-light warning interaction method considering the influence of driving style according to claim 1, characterized in that, The driving data includes steering wheel angle, lateral and longitudinal speed, lateral and longitudinal acceleration, impact, accelerator pedal signal, and brake pedal signal.

3. The sound-light warning interaction method considering the influence of driving style according to claim 1, characterized in that, Step 2 involves extracting data from the slices, including: selecting the average and standard deviation of the driver's speed, absolute acceleration, absolute impact, accelerator pedal opening, and rate of change of accelerator pedal opening, as well as the mean and median of muscle activation and electromyographic amplitude of the gastrocnemius and right sternocleidomastoid muscles to construct feature parameters for subsequent driving style identification.

4. The sound-light warning interaction method considering the influence of driving style according to claim 1, characterized in that, In step 3, safety evaluation indicators are selected to determine driver operation behavior indicators and vehicle status information indicators from both vertical and horizontal perspectives. Longitudinal velocity standard deviation: ； In the formula, V std The standard deviation of longitudinal velocity; V i This represents the actual value of the longitudinal velocity. This represents the average longitudinal velocity. Longitudinal acceleration standard deviation: ； In the formula, It is the standard deviation of longitudinal acceleration; This is the actual value of longitudinal acceleration; It is the average longitudinal acceleration; Longitudinal acceleration rate of change: ； In the formula, It is the rate of change of longitudinal acceleration. This is the actual value of longitudinal acceleration. It is the length of the sampling time period; Steering wheel angle entropy value: ； In the formula, SE It is the entropy value of the steering wheel angle. The probability distribution value of each interval is determined by the frequency of the error value falling within each interval; Steering wheel angle standard deviation: ； In the formula, It is the standard deviation of the steering wheel angle; It is the actual value of the steering wheel angle; It is the average value of the steering wheel angle over one cycle; lateral acceleration standard deviation: ； In the formula, It is the standard deviation of lateral acceleration; This is the actual value of lateral acceleration; It is the average lateral acceleration; Root mean square value of electromyography signal: ； In the formula, It is the root mean square value of the electromyography signal. It is the length of the time period. It refers to the EMG data collected within this time period; Average frequency of electromyographic signals: ； In the formula, It is the average power of the electromyographic signal. At frequency Power spectral density at that point It's frequency. It is the maximum frequency of the signal.

5. The sound-light warning interaction method considering the influence of driving style according to claim 4, characterized in that, The steps for determining the objective weights of the indicators using the entropy method in step 4 are as follows: Step 4.1.1, calculate the first... j Under this indicator, the first i The feature weight of each object: ； Step 4.1.2, calculate the first... j Entropy value of the indicator: ； Step 4.1.3, calculate the coefficient of variation of the indicators: ； Step 4.1.4, Determine the objective weights: 。 6. The sound-light warning interaction method considering the influence of driving style according to claim 5, characterized in that, It has m One driver, n Evaluation indicators C n , X ij It is the first i The first driver's j Item indicator value; The steps for multi-attribute decision-making using the TOPSIS method in step 4 are as follows: Step 4.2.1: Establish a decision matrix based on the original data: （13）； Step 4.2.2: Normalize the data for different indicators to obtain the normalized matrix σ: （14）； Step 4.2.3: Assign weights to each indicator, and finally obtain the evaluation matrix V: （15）； in, , ; Step 4.2.4 yields the ideal solution. and negative ideal solution : ， （16）； If the indicator is a very large indicator, then: (17)； If the indicator is a very small indicator, then: (18)； Step 4.2.5: Calculate the distance between each experimenter and the positive and negative ideal solutions: ； ； Step 4.2.6, calculate the score for each experimenter: ； Score The closer a value is to 1, the safer the driving behavior.

Images