A method for detecting driver's emergency braking intention based on EEG signals

Abstract

The present invention discloses a method for detecting the driver's emergency braking intention based on EEG signals. First, an experimental paradigm for detecting the driver's emergency braking intention based on EEG signals is designed. Then, the driver's EEG signals in emergency braking, normal braking and normal driving scenarios are collected and preprocessed through the experimental paradigm to obtain processed EEG signals. Finally, a classification model is preset, and the classification model is trained and tested. The processed EEG signals are classified using the tested classification model to obtain corresponding driving scenarios, wherein the EEG signals in the emergency braking scenario correspond to the driver's emergency braking intention. The method can predict the driver's emergency braking intention in advance by detecting EEG signals before the driver performs emergency braking, and the emergency braking system of the vehicle automatically implements emergency braking, thereby saving braking time, improving driving safety, and achieving more effective human-vehicle interaction.

Description

Technical Field

[0001] The invention belongs to the field of cognitive neuroscience and pattern recognition, and in particular relates to a method for detecting a driver's emergency braking intention based on electroencephalogram signals. Background Art

[0002] One of the key points of human-vehicle interaction technology is to enable the vehicle to know the human's intentions in advance. How to make the vehicle understand the human's intentions in advance has huge application potential by detecting the driver's EEG signals.

[0003] EEG is a signal collected by wearing an electrode cap to collect the brain's electrophysiological activity. It provides a general reflection of the activity of the cerebral cortex nerve cells. EEG contains a lot of information related to cognition. Before a person's behavior, there will be a corresponding EEG activity. By detecting EEG signals, some of a person's upcoming actions can be predicted in advance. In the field of driving safety, accurately detecting the driver's emergency braking intention can make the car's emergency braking system automatically work in advance, thereby avoiding some serious traffic accidents.

[0004] Based on this, a method for detecting driver's emergency braking intention based on EEG signals in a simulated scenario is proposed. Summary of the invention

[0005] In order to solve the deficiencies of the existing technology, the present invention provides a method for detecting a driver's emergency braking intention based on electroencephalogram signals.

[0006] The technical solution adopted by the present invention to solve the technical problem is to provide a method for detecting the driver's emergency braking intention based on EEG signals, the method comprising the following steps:

[0007] S100: Design an experimental paradigm for detecting driver emergency braking intention based on EEG signals;

[0008] S200: collecting the driver's EEG signals in emergency braking, normal braking and normal driving scenarios through an experimental paradigm, and performing preprocessing to obtain processed EEG signals;

[0009] S300: Preset a classification model, obtain a preset processed EEG signal data set, train and test the preset classification model to obtain a tested classification model, classify the processed EEG signals according to the tested classification model to obtain corresponding driving scenarios, wherein the EEG signals in the emergency braking scenario correspond to the driver's emergency braking intention.

[0010] Preferably, in step S100, a driver emergency braking intention detection experimental paradigm based on EEG signals is designed, specifically:

[0011] S110: Build a driving simulation platform;

[0012] S120: Design emergency braking and normal braking scenarios based on the driving simulation platform;

[0013] S130: According to the preset electrode layout standard and sampling rate of the driving simulation platform, an experimental paradigm for detecting the driver's emergency braking intention based on the EEG signal is obtained.

[0014] Preferably, step S200 is specifically:

[0015] S210: Collect the driver's EEG signals in emergency braking, normal braking and normal driving scenarios through an experimental paradigm;

[0016] S220: performing bandpass filtering processing on the acquired EEG signals in each scenario according to a preset frequency band, and then performing notch filter filtering processing according to a preset frequency, and selecting EEG signals in emergency braking, normal braking and normal driving scenarios;

[0017] S230: performing baseline correction on the acquired EEG signal using a preset time to obtain a corrected EEG signal, and processing the corrected EEG signal using a preset amplitude to obtain a processed EEG signal that does not exceed the preset amplitude.

[0018] Preferably, the classification model in step S300 includes an EEG signal characterization module and a convolutional neural network. The EEG signal characterization module includes a mapping module and a convolution module. The mapping module is used to map the input two-dimensional EEG signal into a three-dimensional EEG signal topology map according to the topological structure of the electrode. The convolution module is used to perform convolution and normalization on the three-dimensional EEG signal topology map to obtain a new two-dimensional EEG signal with the same size as the original two-dimensional EEG signal. The convolutional neural network is used to classify the new two-dimensional EEG signal to obtain the corresponding driving scene.

[0019] Preferably, the mapping module is used to map the input two-dimensional EEG signal into a three-dimensional EEG signal topology map according to the topological structure of the electrode, specifically including:

[0020] S310: representing the two-dimensional EEG signal in the form of number of electrodes × data points;

[0021] S320: constructing a topological structure diagram of the electrodes according to the positions of the electrodes on the scalp;

[0022] S330: mapping the two-dimensional electroencephalogram signal into a two-dimensional electroencephalogram signal topology map according to the electrode topology map;

[0023] S340: Arrange the two-dimensional EEG signal topology map in chronological order to generate a three-dimensional EEG signal topology map.

[0024] Preferably, the convolution module is used to perform convolution operation and normalization processing on the three-dimensional EEG signal topology map to obtain a new two-dimensional EEG signal with the same size as the original two-dimensional EEG signal, specifically including:

[0025] S350: performing a convolution operation on the three-dimensional EEG signal topology map to generate a convolved EEG signal feature map;

[0026] S360: performing batch normalization processing on the convolved EEG signal feature map to obtain a new two-dimensional EEG signal having the same size as the original two-dimensional EEG signal.

[0027] Preferably, in step S350, a convolution operation is performed on the three-dimensional EEG signal topology map, and the specific formula is:

[0028]

[0029] In the formula, f l+1 (i, j) is the value of the feature map at coordinate (i, j) after the l+1th convolution layer, f l (i+m-1,j+n-1) is the value of the feature map obtained after the lth layer of convolution at the coordinate (i+m-1,j+n-1), W(m,n) is the convolution kernel of size (m,n), and ∑ is the summation operation.

[0030] Preferably, in step S360, batch normalization is performed on the convolved EEG signal feature map, and the specific formula is:

[0031]

[0032] Where x(k) and y(k) are the kth batch of input data and the batch-normalized output data, μ and σ are the mean and standard deviation of the input data, γ and β are the learnable translation and scaling parameters, and ε is a small constant set to prevent the denominator from being zero.

[0033] Preferably, in step S300, a classification model is preset, a preset processed EEG signal data set is obtained, and the preset classification model is trained and tested to obtain a tested classification model, specifically:

[0034] S370: Dividing the preset processed EEG signal data set into a training set and a test set according to a preset percentage according to the time sequence of acquisition;

[0035] S380: intercepting the EEG signals corresponding to the acquired training set according to the preset input data length to obtain the EEG signals corresponding to emergency braking and normal braking in the intercepted training set, using the sliding window with the preset input data length and step size to intercept the EEG signals corresponding to normal driving in the acquired training set to obtain the EEG signals corresponding to normal driving in the intercepted training set, and training the classification model using the EEG signals in the intercepted training set to obtain the trained classification model;

[0036] S390: Use a sliding window with preset input data length and step size to intercept the EEG signal corresponding to the acquired test set to obtain the EEG signal corresponding to the intercepted test set, and use the EEG signal corresponding to the intercepted test set to test the trained classification model to obtain the tested classification model.

[0037] Preferably, the EEG signal corresponding to emergency braking in step S380 is the EEG signal 1000 milliseconds before the driver steps on the brake pedal in the emergency braking scenario, the EEG signal corresponding to normal braking is the EEG signal 1000 milliseconds before the driver steps on the brake pedal in the normal braking scenario, and the EEG signal corresponding to normal driving is the EEG signal captured by sliding a sliding window with a length of 1000 milliseconds and a step size of 500 milliseconds on the corresponding EEG signal in the normal driving scenario.

[0038] The above-mentioned method for detecting the driver's emergency braking intention based on EEG signals first designs an experimental paradigm for detecting the driver's emergency braking intention based on EEG signals, then collects the driver's EEG signals in emergency braking, normal braking and normal driving scenarios through the experimental paradigm and preprocesses them to obtain processed EEG signals, and finally presets a classification model, trains and tests the classification model, and uses the tested classification model to classify the processed EEG signals to obtain the corresponding driving scenarios, wherein the EEG signals in the emergency braking scenario correspond to the driver's emergency braking intention. This method can predict the driver's emergency braking intention in advance by detecting EEG signals before the driver performs emergency braking, and the vehicle's emergency braking system automatically implements emergency braking, thereby saving braking time, improving driving safety, and achieving more effective human-vehicle interaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 is a flow chart of a method for detecting a driver's emergency braking intention based on an electroencephalogram signal in one embodiment of the present invention;

[0040] Figure 2 A schematic diagram of a design of an experimental paradigm for detecting a driver's emergency braking intention based on EEG signals in one embodiment of the present invention;

[0041] Figure 3A schematic diagram of a standard electrode layout in an embodiment of the present invention;

[0042] Figure 4 It is a structural diagram of a classification model in one embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings.

[0044] In one embodiment, a method for detecting a driver's emergency braking intention based on an electroencephalogram signal comprises the following steps:

[0045] S100: Design an experimental paradigm for detecting driver emergency braking intention based on EEG signals;

[0046] S200: collecting the driver's EEG signals in emergency braking, normal braking and normal driving scenarios through an experimental paradigm, and performing preprocessing to obtain processed EEG signals;

[0047] S300: Preset a classification model, obtain a preset processed EEG signal data set, train and test the preset classification model to obtain a tested classification model, classify the processed EEG signals according to the tested classification model to obtain corresponding driving scenarios, wherein the EEG signals in the emergency braking scenario correspond to the driver's emergency braking intention.

[0048] Specifically, see Figure 1 , Figure 1 The figure is a flow chart of a method for detecting a driver's emergency braking intention based on EEG signals in one embodiment of the present invention.

[0049] A method for detecting a driver's emergency braking intention based on electroencephalogram signals comprises the following steps: firstly, designing an experimental paradigm for detecting a driver's emergency braking intention based on electroencephalogram signals; then, collecting the driver's electroencephalogram signals in different scenarios such as emergency braking, normal braking and normal driving through the experimental paradigm, and performing preprocessing to obtain processed electroencephalogram signals; finally, presetting a classification model, and training and testing the classification model through a preset processed electroencephalogram signal data set; and using the tested classification model to classify the processed electroencephalogram signals to obtain corresponding driving scenarios, wherein the electroencephalogram signals in the emergency braking scenario correspond to the driver's emergency braking intention.

[0050] In one embodiment, step S100 involves an experimental paradigm for detecting driver emergency braking intention based on EEG signals, specifically:

[0051] S110: Build a driving simulation platform;

[0052] S120: Design emergency braking and normal braking scenarios based on the driving simulation platform;

[0053] S130: According to the preset electrode layout standard and sampling rate of the driving simulation platform, an experimental paradigm for detecting the driver's emergency braking intention based on the EEG signal is obtained.

[0054] Specifically, see Figure 2 and Figure 3 , Figure 2 FIG. 1 is a schematic diagram of a design of an experimental paradigm for detecting a driver's emergency braking intention based on EEG signals in an embodiment of the present invention. Figure 3 FIG. 4 is a schematic diagram of a standard electrode layout in an embodiment of the present invention.

[0055] exist Figure 2In the experiment, the driving simulation platform was first built: the open source CARLA open city driving simulator was used as the experimental driving simulation platform. The experimental equipment included a driving simulator (including a steering wheel, a driving seat, a brake pedal, an accelerator pedal and a gear shifter), an EEG acquisition device (an EEG amplifier, electrodes and an EEG cap) and two computers and their monitors. The driving simulator is used to drive the simulated vehicle. The EEG acquisition device is responsible for collecting the driver's EEG signals in real time. The computer is used to present the driving simulation environment and record the EEG signals in real time. The driver's braking signal and the received stimulation signal are recorded synchronously in the computer as the label information of the EEG signal through the application program interface of the simulated driver; then the emergency braking and normal braking scenarios were designed: two vehicles were set up in the experiment, one vehicle controlled by the experimental instructor was driving in front as the guide vehicle, and the other was controlled by the driver as the following vehicle. They drove on the CARLA simulation platform at a speed of 90 km / h. The driver's task was to make the driving vehicle follow the leading vehicle in front closely, keeping the distance between 6 and 15 meters. In the emergency braking scenario, the experimental leading vehicle will randomly brake at intervals of 15-60 seconds, and use the brake lights to alert the following vehicles behind. In order to avoid a collision with the leading vehicle in front, the driver needs to immediately emergency brake the following vehicle he is driving. In this process, two signals are synchronously recorded into the computer through the application interface of the simulated driving platform as label information of the real-time collected EEG signals: one is the moment when the brake lights of the leading vehicle light up, and the other is the moment when the driver starts to step on the brake pedal. After 3 seconds, the leading vehicle starts to accelerate and maintain a speed of 90 kilometers per hour. The driver continues to control the following vehicle to closely follow the leading vehicle in front and keep a distance of 6-15 meters from it. The experimental settings of the normal braking scenario are the same as the emergency brake drum scenario, except that the leading vehicle in front always maintains a speed of 90 kilometers per hour and no longer decelerates urgently. In the normal braking scenario, the driver spontaneously brakes the vehicle he is driving normally at intervals of 15-60 seconds, and starts to accelerate and follow the leading vehicle in front 3 seconds after braking, and keeps a distance of 6-15 meters from it. In addition, it should be noted that Figure 2 The braking type shown includes two braking types: emergency braking and normal braking; finally, the electrode layout standard and sampling rate setting are performed: the electrode layout standard used in the present invention is as follows Figure 3 As shown, the electrode distribution conforms to the international 10-20 system. Figure 3In the experiment, 28 electrodes (F5 F3 Fz F4 F6FT7 FC5 FC1 FC2 FC6 FT8 T7 C3 Cz C4 T8 CP5 CP1 CP2 CP6 P5 P3 Pz P4 P6 O1 OzO2) were used to collect EEG signals, 2 electrodes were used as reference (TP9 TP10), 1 electrode was used as ground electrode (FPz), and the impedance of all electrodes was less than 10 kilo-ohms. The sampling rate of EEG signals was set to 200 Hz.

[0056] In one embodiment, step S200 is specifically as follows:

[0057] S210: Collect the driver's EEG signals in emergency braking, normal braking and normal driving scenarios through an experimental paradigm;

[0058] S220: performing bandpass filtering processing on the acquired EEG signals in each scenario according to a preset frequency band, and then performing notch filter filtering processing according to a preset frequency, and selecting EEG signals in emergency braking, normal braking and normal driving scenarios;

[0059] S230: performing baseline correction on the acquired EEG signal using a preset time to obtain a corrected EEG signal, and processing the corrected EEG signal using a preset amplitude to obtain a processed EEG signal that does not exceed the preset amplitude.

[0060] Specifically, step S200 includes: first, collecting the EEG signals of the driver in three different scenarios: normal driving, normal braking and emergency braking; then performing 1-45Hz bandpass filtering and 50Hz notch filter filtering on the collected EEG signals, and selecting the corresponding EEG signals of emergency braking, normal braking and normal driving. The specific selection method is: the EEG signal in the emergency braking scenario is selected as the EEG signal from 2000 milliseconds before the driver steps on the brake pedal to the moment when the brake pedal is stepped on; the EEG signal in the normal braking scenario is selected as the EEG signal from 2000 milliseconds before the driver steps on the brake pedal to the moment when the brake pedal is stepped on; the EEG signal in the normal driving scenario is selected by using a sliding window with a length of 2000 milliseconds and a step size of 500 milliseconds, and sliding on the EEG signal collected during the driver's normal driving process (at least 3000 milliseconds away from any vehicle braking situation) to intercept. Finally, the first 100 milliseconds of data for each type of EEG signal were used for baseline correction. On this basis, the EEG signals with amplitudes greater than 300 mV were removed to obtain the processed EEG signals corresponding to different scenarios of emergency braking, normal braking, and normal driving.

[0061] In one embodiment, the classification model in step S300 includes an EEG signal characterization module and a convolutional neural network. The EEG signal characterization module includes a mapping module and a convolution module. The mapping module is used to map the input two-dimensional EEG signal into a three-dimensional EEG signal topology map according to the topological structure of the electrode. The convolution module is used to perform convolution and normalization on the three-dimensional EEG signal topology map to obtain a new two-dimensional EEG signal of the same size as the original two-dimensional EEG signal. The convolutional neural network is used to classify the new two-dimensional EEG signal to obtain the corresponding driving scene.

[0062] Specifically, see Figure 4 , Figure 4 It is a structural diagram of a classification model in one embodiment of the present invention.

[0063] exist Figure 4 In the classification model, an EEG signal representation module based on the electrode topology and a convolutional neural network are included. The EEG signal representation module includes a mapping module and a convolutional module. The mapping module is used to map the input two-dimensional EEG signal into a three-dimensional EEG signal topology map according to the electrode topology. The convolutional module is used to perform convolution operations and normalization processing on the obtained three-dimensional EEG signal topology map to obtain a new EEG signal with the same size as the original two-dimensional EEG signal. The convolutional neural network is used to classify the obtained new two-dimensional EEG signal to obtain the corresponding driving scene. When the driving scene is an emergency braking scene, the EEG signal in the emergency braking scene is the driver's emergency braking intention.

[0064] In one embodiment, the mapping module is used to map the input two-dimensional EEG signal into a three-dimensional EEG signal topology map according to the topological structure of the electrode, specifically including:

[0065] S310: representing the two-dimensional EEG signal in the form of number of electrodes × data points;

[0066] S320: constructing a topological structure diagram of the electrodes according to the positions of the electrodes on the scalp;

[0067] S330: mapping the two-dimensional electroencephalogram signal into a two-dimensional electroencephalogram signal topology map according to the electrode topology map;

[0068] S340: Arrange the two-dimensional EEG signal topology map in chronological order to generate a three-dimensional EEG signal topology map.

[0069] Specifically, the two-dimensional EEG signal is first represented in the form of the number of electrodes × data points: the number of electrodes of the two-dimensional EEG signal is set to the number of electrodes C used to collect EEG, and the data point is set to N, N = T × L, which represents the number of data points that constitute the discrete EEG signal collected by a single electrode, where T is a fixed duration and L is the EEG signal sampling rate; then, the electrode topology diagram is constructed according to the position of the electrode on the scalp: the electrode layout adopts the international 10-20 system, and the corresponding electrode topology diagram is a 9×9 matrix. The size of the electrode topology diagram can be adjusted according to the position of the electrodes used to collect EEG signals. ; Then, the two-dimensional EEG signal is mapped into a two-dimensional EEG topological map according to the electrode topological structure diagram. Specifically, the voltage amplitude of the two-dimensional EEG signal is mapped onto the electrode topological structure diagram according to the corresponding relationship between the electrode and the topological structure diagram. After mapping, the amplitude of the EEG signal at the corresponding electrode position on the topological structure diagram remains unchanged. If there is no corresponding electrode corresponding to the position on the electrode topological structure diagram or the electrode at this position is not used in the EEG signal, the value of this position is 0, thereby obtaining a two-dimensional EEG topological structure diagram at a certain moment; finally, the two-dimensional EEG topological structure diagram is arranged in chronological order to generate a three-dimensional EEG signal topological map.

[0070] In one embodiment, the convolution module is used to perform convolution operation and normalization processing on the three-dimensional EEG signal topology map to obtain a new two-dimensional EEG signal with the same size as the original two-dimensional EEG signal, specifically including:

[0071] S350: performing a convolution operation on the three-dimensional EEG signal topology map to generate a convolved EEG signal feature map;

[0072] S360: performing batch normalization processing on the convolved EEG signal feature map to obtain a new two-dimensional EEG signal having the same size as the original two-dimensional EEG signal.

[0073] In one embodiment, in step S350, a convolution operation is performed on the three-dimensional EEG signal topology map, and the specific formula is:

[0074]

[0075] In the formula, f l+1 (i, j) is the value of the feature map at coordinate (i, j) after the l+1th convolution layer, f l (i+m-1,j+n-1) is the value of the feature map obtained after the lth layer of convolution at the coordinate (i+m-1,j+n-1), W(m,n) is the convolution kernel of size (m,n), and ∑ is the summation operation.

[0076] In one embodiment, in step S360, batch normalization is performed on the convolved EEG signal feature map, and the specific formula is:

[0077]

[0078] Where x(k) and y(k) are the kth batch of input data and the batch-normalized output data, μ and σ are the mean and standard deviation of the input data, γ and β are the learnable translation and scaling parameters, and ε is a small constant set to prevent the denominator from being zero.

[0079] Specifically, firstly, a convolution operation is performed on the three-dimensional EEG signal topology map to generate an EEG signal feature map after convolution. Two convolution strategies can be used to complete the above operation: one is to use C convolution kernels of size 5×5 with a step size of 1 to convolve the three-dimensional EEG signal topology map to obtain a feature map, and then select C convolution kernels of the same size as the actual size of the feature map to perform a convolution operation on the feature map to generate an EEG signal feature map of size (number of electrodes × data points) after convolution; the other is to use C convolution kernels of size 3×3 with a step size of 1 The three-dimensional EEG signal topology map is convolved to obtain a feature map, and then C 3×3 convolution kernels are used to convolve the obtained feature map with a step size of 1 until the final feature map is smaller than the size of the convolution kernel 3×3, and then C convolution kernels of the same size as the actual size of the feature map are selected to perform a convolution operation on the feature map to generate an EEG signal feature map with a convolution size of (number of electrodes × data points); then the convolved EEG signal feature map is batch normalized to obtain a new signal with the same size as the original two-dimensional EEG signal.

[0080] Furthermore, the formula for the convolution operation is expressed as:

[0081]

[0082] In the formula, f l+1 (i, j) is the value of the feature map at coordinate (i, j) after the l+1th convolution layer, f l (i+m-1,j+n-1) is the value of the feature map obtained after the lth layer of convolution at the coordinate (i+m-1,j+n-1), W(m,n) is the convolution kernel of size (m,n), and ∑ is the summation operation.

[0083] Furthermore, the batch normalization formula is expressed as:

[0084]

[0085] Where x(k) and y(k) are the kth batch of input data and the batch-normalized output data, μ and σ are the mean and standard deviation of the input data, γ and β are the learnable translation and scaling parameters, and ε is a small constant set to prevent the denominator from being zero.

[0086] In one embodiment, in step S300, a classification model is preset, a preset processed EEG signal data set is obtained, and the preset classification model is trained and tested to obtain a tested classification model, specifically:

[0087] S370: Dividing the preset processed EEG signal data set into a training set and a test set according to a preset percentage according to the time sequence of acquisition;

[0088] S380: intercepting the EEG signals corresponding to the acquired training set according to the preset input data length to obtain the EEG signals corresponding to emergency braking and normal braking in the intercepted training set, using the sliding window with the preset input data length and step size to intercept the EEG signals corresponding to normal driving in the acquired training set to obtain the EEG signals corresponding to normal driving in the intercepted training set, and training the classification model using the EEG signals in the intercepted training set to obtain the trained classification model;

[0089] S390: Use a sliding window with preset input data length and step size to intercept the EEG signal corresponding to the acquired test set to obtain the EEG signal corresponding to the intercepted test set, and use the EEG signal corresponding to the intercepted test set to test the trained classification model to obtain the tested classification model.

[0090] In one embodiment, the EEG signal corresponding to emergency braking in step S380 is the EEG signal 1000 milliseconds before the driver steps on the brake pedal in the emergency braking scenario, the EEG signal corresponding to normal braking is the EEG signal 1000 milliseconds before the driver steps on the brake pedal in the normal braking scenario, and the EEG signal corresponding to normal driving is an EEG signal captured by sliding a sliding window with a length of 1000 milliseconds and a step size of 500 milliseconds on the EEG signal corresponding to the normal driving scenario.

[0091] Specifically, the process of training and testing the preset classification model to obtain the tested classification model is as follows:

[0092] 1) For the EEG signal dataset consisting of EEG signals of emergency braking, normal braking and normal driving, the first 50% of the EEG signals of each category are used as the training set, and the last 50% are used as the test set according to the time sequence of EEG signal collection;

[0093] 2) Set the input data length to 1000 milliseconds, intercept the EEG signals corresponding to emergency braking and normal braking in the training set, and obtain the intercepted EEG signals corresponding to emergency braking and normal braking in the training set; use a sliding window with an input data length of 1000 milliseconds and a step size of 500 milliseconds to intercept the EEG signals corresponding to normal driving in the training set, and obtain the intercepted EEG signals corresponding to normal driving in the training set. Preferably, the EEG signal corresponding to emergency braking is the EEG signal 1000 milliseconds before the driver steps on the brake pedal in the emergency braking scenario, the EEG signal corresponding to normal braking is the EEG signal 1000 milliseconds before the driver steps on the brake pedal in the normal braking scenario, and the EEG signal corresponding to normal driving is the EEG signal intercepted by sliding a sliding window with a length of 1000 milliseconds and a step size of 500 milliseconds on the corresponding EEG signal in the normal driving scenario. Use the intercepted EEG signals in the training set to train the classification model to obtain the trained classification model;

[0094] 3) Use a sliding window with a length of 1000 mm and a step length of 100 milliseconds to intercept the EEG signals corresponding to the test set, and obtain the EEG signals corresponding to emergency braking, normal braking, and normal driving in the intercepted test set. Use the EEG signals in the intercepted test set to test the trained classification model and obtain the tested classification model.

[0095] The above-mentioned method for detecting the driver's emergency braking intention based on EEG signals first designs an experimental paradigm for detecting the driver's emergency braking intention based on EEG signals, then collects the driver's EEG signals under emergency braking, normal braking and normal driving through the experimental paradigm and preprocesses them to obtain processed EEG signals, and finally presets a classification model, trains and tests the classification model, and uses the tested classification model to classify the processed EEG signals to obtain the corresponding driving scene, wherein the EEG signals in the emergency braking scene correspond to the driver's emergency braking intention. This method can predict the driver's emergency braking intention in advance by detecting EEG signals before the driver performs emergency braking, and the vehicle's emergency braking system automatically implements emergency braking, thereby saving braking time, improving driving safety, and achieving more effective human-vehicle interaction.

[0096] The above is a detailed introduction to a method for detecting driver emergency braking intention based on EEG provided by the present invention. This article uses specific examples to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the core idea of ​​the present invention. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of the present invention, the present invention can also be improved and modified, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

Claims

1. A method for detecting driver's emergency braking intention based on EEG signals, characterized in that: The method comprises the following steps: S100: Design an experimental paradigm for detecting driver emergency braking intention based on EEG signals; S200: collecting the driver's EEG signals in emergency braking, normal braking and normal driving scenarios through the experimental paradigm, and performing preprocessing to obtain processed EEG signals; S300: Preset a classification model, obtain a preset processed EEG signal data set, train and test the preset classification model to obtain a tested classification model, classify the processed EEG signals according to the tested classification model to obtain corresponding driving scenarios, wherein the EEG signals in the emergency braking scenario correspond to the driver's emergency braking intention; the classification model in step S300 includes an EEG signal characterization module and a convolutional neural network, the EEG signal characterization module includes a mapping module and a convolution module, the mapping module is used to map the input two-dimensional EEG signal into a three-dimensional EEG signal topology map according to the topological structure of the electrode, the convolution module is used to perform convolution operation and normalization processing on the three-dimensional EEG signal topology map to obtain a new two-dimensional EEG signal with the same size as the original two-dimensional EEG signal, and the convolutional neural network is used to classify the new two-dimensional EEG signal to obtain the corresponding driving scenario; The specific formula for normalization is: ; In the formula, and They are Batch input data and batch normalized output data, and are the mean and standard deviation of the input data, and are the learnable translation and scaling parameters, It is a small constant set to prevent the denominator from being 0.

2. The method for detecting driver's emergency braking intention based on EEG signals as claimed in claim 1, characterized in that: In step S100, an experimental paradigm for detecting the driver's emergency braking intention based on EEG signals is designed, specifically: S110: Build a driving simulation platform; S120: Designing emergency braking and normal braking scenarios according to the driving simulation platform; S130: According to the preset electrode layout standard and sampling rate of the driving simulation platform, an experimental paradigm for detecting the driver's emergency braking intention based on the EEG signal is obtained.

3. The method for detecting driver's emergency braking intention based on EEG signals as claimed in claim 2, characterized in that: The step S200 is specifically as follows: S210: collecting the driver's EEG signals in emergency braking, normal braking and normal driving scenarios through the experimental paradigm; S220: performing bandpass filtering processing on the acquired EEG signals in each scenario according to a preset frequency band, and then performing notch filter filtering processing according to a preset frequency, and selecting EEG signals in emergency braking, normal braking and normal driving scenarios; S230: performing baseline correction on the acquired EEG signal using a preset time to obtain a corrected EEG signal, and processing the corrected EEG signal using a preset amplitude to obtain a processed EEG signal that does not exceed the preset amplitude.

4. The method for detecting driver's emergency braking intention based on EEG signals as claimed in claim 3, characterized in that: The mapping module is used to map the input two-dimensional EEG signal into a three-dimensional EEG signal topology map according to the topological structure of the electrode, and specifically includes: S310: representing the two-dimensional EEG signal in the form of number of electrodes × data points; S320: constructing a topological structure diagram of the electrodes according to the positions of the electrodes on the scalp; S330: Mapping the two-dimensional electroencephalogram signal into a two-dimensional electroencephalogram signal topology map according to the electrode topology map; S340: Arrange the two-dimensional EEG signal topology map in chronological order to generate a three-dimensional EEG signal topology map.

5. The method for detecting driver's emergency braking intention based on EEG signals as claimed in claim 4, characterized in that: The convolution module is used to perform convolution operation and normalization processing on the three-dimensional EEG signal topology map to obtain a new two-dimensional EEG signal with the same size as the original two-dimensional EEG signal, specifically including: S350: performing a convolution operation on the three-dimensional EEG signal topology map to generate a convolved EEG signal feature map; S360: performing batch normalization processing on the convolved EEG signal feature map to obtain a new two-dimensional EEG signal having the same size as the original two-dimensional EEG signal.

6. The method for detecting driver's emergency braking intention based on EEG signals as claimed in claim 5, characterized in that: In step S350, a convolution operation is performed on the three-dimensional EEG signal topology map, and the specific formula is: ； In the formula, It is The feature map obtained after the layer convolution is in coordinates The value at For the The feature map obtained after the layer convolution is in coordinates The value at The size is The convolution kernel, For the sum operation.

7. The method for detecting driver's emergency braking intention based on EEG signals as claimed in claim 1, characterized in that: In step S300, a classification model is preset, a preset processed EEG signal data set is obtained, and the preset classification model is trained and tested to obtain a tested classification model, specifically: S370: Dividing the preset processed EEG signal data set into a training set and a test set according to a preset percentage according to the time sequence of acquisition; S380: intercepting the EEG signals corresponding to the acquired training set according to the preset input data length to obtain the EEG signals corresponding to emergency braking and normal braking in the intercepted training set, using a sliding window with preset input data length and step length to intercept the EEG signals corresponding to normal driving in the acquired training set to obtain the EEG signals corresponding to normal driving in the intercepted training set, and using the EEG signals in the intercepted training set to train the classification model to obtain a trained classification model; S390: Use a sliding window with preset input data length and step size to intercept the EEG signal corresponding to the acquired test set to obtain the EEG signal corresponding to the intercepted test set, and use the EEG signal corresponding to the intercepted test set to test the trained classification model to obtain the tested classification model.

8. The method for detecting driver's emergency braking intention based on EEG signals as claimed in claim 7, characterized in that: The EEG signal corresponding to the emergency braking in step S380 is the EEG signal 1000 milliseconds before the driver steps on the brake pedal in the emergency braking scenario. The EEG signal corresponding to normal braking is the EEG signal 1000 milliseconds before the driver steps on the brake pedal in the normal braking scenario. The EEG signal corresponding to normal driving is an EEG signal captured by sliding a sliding window with a length of 1000 milliseconds and a step size of 500 milliseconds on the corresponding EEG signal in the normal driving scenario.

Images