CN107095670A - Time of driver's reaction Forecasting Methodology - Google Patents

Abstract

The invention discloses a method for predicting a driver's reaction time. The method for predicting the driver's reaction time comprises the steps of: acquiring the driver's brain electrical signal; extracting brain electrical characteristic parameters in the filtered brain electrical signal through wavelet transform; Taking the driver's EEG characteristic parameters as the input layer and the reaction time as the output layer, a BP neural network prediction model is constructed. In the present invention, the EEG characteristic parameters are extracted by wavelet transform as an objective predictive index, combined with the BP neural network method, a predictive algorithm of driver reaction time is constructed, so as to provide a real-time driver's dangerous driving state early warning system for the development of the vehicle Provide theoretical basis with design.

Description

Driver Reaction Time Prediction Method

technical field

The invention relates to the technical field of measuring the reaction time of a specific person, in particular to a method for predicting the reaction time of a driver.

Background technique

The continuous improvement of the speed of high-speed rail puts forward higher requirements on the operating ability of EMU drivers, and the ability of EMU drivers to respond to emergencies is a key factor affecting the safety and reliability of driving operations.

At present, there are few related studies on the prediction of reaction time of EMU drivers (and motor vehicle drivers) at home and abroad. The existing research mainly discusses the correlation between reaction time and neurophysiological signals. Jap et al. studied the correlation between EEG signal and reaction time during the long-term monotonous driving of locomotive drivers in a simulated environment. Time is negatively correlated. Darwent et al. monitored the vigilance of locomotive drivers in a real environment, and the results showed that with the decline of driving vigilance, the reaction time to emergencies continued to prolong, and demonstrated that there was a close relationship between EEG signals and reaction time. connect. Haga et al. designed an experiment on the ability of locomotive drivers to respond to signal lights, and verified that changes in EEG signals can directly reflect the ability of locomotive drivers to respond. Lin et al. used the correlation between EEG signals and driving behavior performance (reaction time, vehicle speed, etc.) to monitor the level of continuous attention in driving in real time.

The above studies have verified the correlation between reaction time and neurophysiological signals from different perspectives, but the correlation between the two has not been used to effectively predict the reaction time. Therefore, how to realize the effective prediction of the reaction time of EMU drivers to emergencies through the neurophysiological signals of EMU drivers is a key technology for building a real-time warning system for dangerous driving status of EMU drivers.

Contents of the invention

In view of this, the present invention aims to provide a method for predicting the driver's reaction time, that is, extracting the EEG characteristic parameters by wavelet transform as an objective predictive index, and in conjunction with the BP neural network method, a prediction algorithm for the driver's reaction time is constructed, In order to provide a theoretical basis for the development and design of the vehicle-mounted real-time driver's dangerous driving state warning system.

Specifically, the driver's reaction time prediction method of the present invention includes the steps of: obtaining the driver's EEG signal; extracting the EEG characteristic parameters in the filtered EEG signal through wavelet transform; The parameters are used as the input layer, and the response time is used as the output layer to build a BP neural network prediction model.

Further, the filtering process specifically includes: performing an overall filtering process on the EEG signal with a bandwidth of 0-35 Hz.

Further, extracting EEG feature parameters through wavelet transform specifically includes:

a. For the filtered EEG signal, denoted as u(n), its wavelet transform is defined as:

In the formula, is the wavelet function; i is the frequency factor; m is the time translation factor; n is the signal duration;

b. The wavelet-transformed signal u(n) is subjected to finite layer decomposition, using the Mallat algorithm:

In the formula, AH is the approximate component; Ci is the detail component at different scales; H is the number of decomposition layers; through the above finite layer decomposition, the wavelet coefficients of θ, α, and β three different frequency bands are obtained;

c. Extract the energy value of the wavelet coefficient for the above-mentioned frequency band as the EEG feature parameter:

In the formula, P _X is the energy value of the corresponding frequency band; S _X (t) is the wavelet coefficient of the corresponding frequency band; t is the time; h _i is the amplitude of the corresponding frequency band;

d. Process the EEG signals of q electrodes, and obtain 3×q EEG characteristic parameters accordingly.

Further, the method also includes: normalizing the obtained EEG characteristic parameters according to the formula, so that the value of the EEG characteristic parameters is between [0,1], so as to eliminate the noise existing in the data:

In the formula, _ximax and _ximin are the maximum and minimum values of the EEG characteristic parameter _xi .

Further, constructing the BP neural network prediction model specifically includes: taking the driver's EEG characteristic parameters as the input layer, and taking the predicted value of reaction time as the output layer, constructing a BP neural network prediction model with one hidden layer; wherein, the input The number of layer nodes is determined by the number of EEG characteristic parameters of the driver; the number s of hidden layer nodes is selected according to the model training results; the number of output layer nodes is 1.

Further, the connection weight coefficients and offsets between the input layer to the hidden layer and the hidden layer to the output layer in the model are respectively wi _ik , w _k1 , b _ik , b _k1 (i=1,2, ...,3×q,k=1,2,...,s), for the output from any node o in the input layer to any node p in the hidden layer:

y _op ＝f(x _i w _op +b _op )

In the formula, f(·) is the Sigmoid function, namely

The output layer output result is:

In the formula, y _i is the prediction result output of the neural network model; W ₁ is the connection weight coefficient matrix from the input layer to the hidden layer; W ₂ is the connection weight coefficient matrix from the hidden layer to the output layer; _xi is the driver EEG feature parameters; b ₁ is the bias matrix from the input layer to the hidden layer; b ₂ is the bias matrix from the hidden layer to the output layer.

Further, the acquiring the EEG signals of the drivers specifically includes: acquiring the EEG signals of a predetermined number of drivers.

Further, the EEG signal acquisition frequency is 10 Hz; the EEG acquisition device continuously collects the driver's EEG data.

Further, the comparison between the predicted value of the reaction time obtained by the BP neural network model and the actual value of the reaction time collected in the experiment compares the fitting degree and selects the maximum absolute error M1 and the relative mean square error M2 as the evaluation indicators of the prediction effect:

The present invention uses wavelet transform to extract EEG feature parameters as an objective predictive index, and combines the BP neural network method to construct a predictive algorithm of driver's reaction time, which realizes the accurate prediction of the driver's reaction time to emergencies. The development and design of the real-time driver's dangerous driving state warning system provides a theoretical basis.

Description of drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate the embodiments of the invention and together with the description serve to explain the principles of the invention. In the drawings, like reference numerals are used to denote like elements. The drawings in the following description are some, but not all, embodiments of the present invention. Those skilled in the art can obtain other drawings based on these drawings without creative efforts.

FIG. 1 is a schematic flow chart of a driver reaction time prediction method provided by an embodiment of the present invention;

Fig. 2 is a schematic diagram of the location of the stimulus point in the experimental task in the embodiment of the present invention;

Fig. 3 is a schematic diagram of the detection signal flashing in the experimental task in the embodiment of the present invention;

Fig. 4 is the structural representation of the BP neural network model in the embodiment of the present invention;

Fig. 5 is the comparison schematic diagram of BP neural network model prediction reaction time and actual reaction time in the embodiment of the present invention;

FIG. 6 is a schematic diagram of a reaction time error predicted by a BP neural network model in an embodiment of the present invention.

detailed description

In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments It is a part of embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention. It should be noted that, in the case of no conflict, the embodiments in the present application and the features in the embodiments can be combined arbitrarily with each other.

The method for predicting the driver's reaction time according to the embodiment of the present invention will be described in detail below in conjunction with the drawings and scenes.

Referring to Fig. 1, the method of the driver's reaction time prediction method includes the steps of: obtaining the driver's EEG signal; extracting the EEG characteristic parameters in the filtered EEG signal by wavelet transform; The electrical characteristic parameters are used as the input layer, and the response time is used as the output layer to build a BP neural network prediction model.

In order to further illustrate the advantages, principles and effects of the method of the present invention, it will be described in detail below in conjunction with specific experimental scenarios and examples.

1 Experimental method

1.1 Selection of subjects

Select 20 male trainees of a certain EMU driver class as subjects. Those aged between 34 and 38 had a mean of 36.1 years and a standard deviation of 1.8 years; those with driving experience of 6 to 11 years had a mean of 7.2 years and a standard deviation of 1.1 years. The selected subjects had good sleep quality, good physical condition, no bad habits (smoking, alcoholism, etc.), no color weakness or color blindness, and visual acuity or corrected visual acuity was 1.0. 4 hours before the start of the experiment, the subjects were prohibited from drinking coffee or tea and other irritating beverages, and they voluntarily signed the "Informed Consent Form" after fully understanding it.

1.2 Experimental equipment

1.2.1 EMU Simulator

The experiment uses a CRH380 train simulator with a 6-degree-of-freedom motion system. The simulator uses a single-channel large-screen forward-looking vision system with a screen resolution of 1920×1200pix and a horizontal viewing angle of up to 160 ^° . The locomotive main operating console is composed of automatic train protection system (Automatic Train Protection, ATP), integrated dedicated digital mobile communication system, DMI display screen, switch, indicator light, speed setting controller, etc. The sound environment of the simulator is generated by a 7.1 digital audio sound system, which can simulate the background sound environment of the running train with high fidelity. The validity of the simulator has passed the systematic test, and its degree of simulation can meet the experimental requirements.

1.2.2 EEG Acquisition Instrument

During the experiment, the 64-channel Neuroscan EEG instrument produced by Compumedics in Australia was used to continuously collect the EEG signal data of the subjects in real time. The EEG instrument adopts the electrode cap of the 10-20 system commonly used by the International Electroencephalography Society, the electrode position has been set, and the FCz electrode is selected as the reference electrode.

1.3 Experimental tasks

The experiment uses the line from Fuzhou to Hefei South Railway Station, with a total length of 808km. The train passes through 22 stations, and when it arrives at the station, it will perform normal station operations. When the train is running in each section, the random signal detection method is used to detect the reaction time of the subjects in the driving operation in real time. During the test driving operation, five possible positions in the front screen will present random signals (red dots) (Figure 2), and the signal presentation time is 120±10s. When the signal appeared at a certain position (Figure 3), the subjects were required to respond as quickly as possible by pressing the keys. If the subject did not press the button 1000ms after the signal appeared, the response was considered invalid.

1.4 Experimental process and data collection

24 hours before the start of the experiment, let the subjects understand the experimental tasks and operating rules, and then operate the simulator according to their daily driving habits until the subjects can operate the simulator proficiently. In order to ensure that the subjects had a high level of vigilance, the experiment was uniformly arranged to be conducted at 8:00 am. Before the start of the formal experiment, in order to make the subjects adapt to the driving simulation environment and enter the experimental state, 15 minutes of driving simulation practice was given. During the formal experiment, the indoor light illumination was 300lx, and the temperature was 24±1°C. The subjects were required to keep the train running at a speed of no less than 220km/h, and the driving time was 2 hours.

Simultaneously record the reaction time of the subjects to random signal stimuli during driving, and the data acquisition frequency is 10Hz. At the same time, the EEG acquisition instrument continuously collects the EEG data of the subjects. In order to remove the interference of other information, the horizontal and vertical oculoelectricity and myoelectricity were recorded. The sampling rate of the EEG signal is set, and the sampling frequency bandwidth is 128Hz, 0.5-100Hz. It is required that the impedance of all electrodes shall not exceed 5kΩ.

2. Reaction time prediction model based on EEG signal

For the EEG data of 20 EMU drivers collected in the above experiment, the wavelet transform is used to extract the EEG characteristic parameters in the filtered EEG data, combined with the EEG characteristic parameters as the input index, and the reaction time as the output The index BP neural network is used to construct the prediction model of the EMU driver's reaction time to emergencies. The specific model construction steps are as follows:

2.1 Extraction of EEG feature parameters

The EEG signal can reflect the activity state of the cerebral cortex. When the EMU driver is at a low level of arousal, the EEG spectrum distribution tends to be in the low frequency band. On the contrary, when the level of arousal is high, the EEG spectrum distribution tends to be in the high frequency band. Existing research results have shown that the three frequency bands of θ (4-8Hz), α (8-13Hz), and β (13-30Hz) in EEG signals are highly correlated with reaction time and can be used as objective predictors of reaction time. Therefore, in the embodiment of the present invention, the wavelet coefficient energy values of the above three frequency bands are extracted by wavelet transform as the EEG feature parameters. Its calculation process is as follows:

(1) The EEG signal collected in the experiment is processed by overall filtering with a bandwidth of 0-35Hz to remove the interference of artifact components such as power frequency electricity and part of myoelectricity.

(2) For the filtered EEG signal, denoted as u(n), its wavelet transform is defined as:

In the formula, is the wavelet function; i is the frequency factor; m is the time translation factor; n is the signal duration.

(3) In order to decompose the wavelet-transformed signal u(n) into finite layers, the embodiment of the present invention introduces the Mallat algorithm [9], namely

In the formula, AH is the approximate component; Ci is the detail component at different scales; H is the number of decomposition layers, and the number of layers is taken as 3 in this paper. Therefore, the wavelet coefficients of θ, α, and β three different frequency bands can be obtained through the above-mentioned finite layer decomposition.

(4) The energy value of the wavelet coefficient can reflect the frequency-domain characteristics of the EEG signal, so the corresponding energy value of the above band is extracted as the EEG characteristic parameter.

In the formula, P _X is the energy value of the corresponding frequency band; S _X (t) is the wavelet coefficient of the corresponding frequency band; t is the time; h _i is the amplitude of the corresponding frequency band.

(5) Process the EEG signals of q electrodes according to the steps (1) - (4), and obtain 3×q EEG characteristic parameters correspondingly, denoted as x _i (i=1,2,...,3×q). Due to the different dimensions of the EEG characteristic parameters, the EEG characteristic parameters are normalized according to formula (4), so that the value of the EEG characteristic parameters is between [0,1] to eliminate the noise in the data .

In the formula, _ximax and _ximin are the maximum and minimum values of the EEG characteristic parameter _xi .

2.2 Construction of prediction model based on BP neural network

Taking EMU driver's EEG feature parameters as input layer and reaction time prediction value as output layer, a BP neural network prediction model with one hidden layer was constructed. Among them, the number of input layer nodes is determined by the EMU driver's EEG characteristic parameter number 3×q; the number of hidden layer nodes s is selected according to the model training results; the number of output layer nodes is 1, and its structure diagram is as follows Figure 4 shows.

In this model, the connection weight coefficients and offsets between the input layer and the hidden layer and between the hidden layer and the output layer are respectively set as w _ik , w _k1 , b _ik , b _k1 (i=1,2,... ,3×q,k=1,2,...,s), for the output from any node o in the input layer to any node p in the hidden layer:

y _op ＝f(x _i w _op +b _op ) ⑸

In the formula, f(·) is the Sigmoid function, namely

The output layer output result is:

In the formula, y _i is the output of the neural network, that is, the model prediction result; W ₁ is the connection weight coefficient matrix from the input layer to the hidden layer; W ₂ is the connection weight coefficient matrix from the hidden layer to the output layer; x _i is EMU driver’s EEG characteristic parameter; b ₁ is the bias matrix from input layer to hidden layer; b ₂ is the bias matrix from hidden layer to output layer.

For an EMU driver experimental sample X _i =(x ₁ ,x ₂ ,... _xi ,...,x _3×q ; y _i ), where y _i represents the value obtained by the EMU driver for the ith key press The reaction time is input into the neural network for training, and the network output error of the experimental sample is defined as:

In the formula, is the actual reaction time of the EMU driver. The total error for all experimental samples of EMU drivers is defined as:

In the formula, N is the number of experimental samples. The neural network is trained by adjusting the connection weights through error backpropagation until the total error reaches the minimum, thus completing the training of the neural network.

2.3 Forecasting effect evaluation index

In order to evaluate the degree of fit between the predicted value of the reaction time obtained by the BP neural network model and the actual value of the reaction time collected in the experiment, the embodiment of the present invention selects the maximum absolute error M1 and the relative mean square error M2 as the evaluation indicators of the prediction effect.

3 Beneficial effects and analysis

For the EEG data collected in each reaction time in the experiment, the method in Section 2.1 was used to obtain 96 EEG characteristic parameters after normalization processing, which were used as the input indicators of the BP neural network prediction model. In order to verify that there is a correlation between the EEG characteristic parameters and the reaction time, the embodiment of the present invention uses the Pearson correlation test to test the correlation between the two, so as to provide a theoretical premise for the prediction of the reaction time.

(1) Correlation analysis between EEG characteristic parameters and reaction time. The Pearson correlation test was performed on the three EEG characteristic parameters of θ, α, β and reaction time, and the results are shown in the table.

Table 1 Correlation between EEG characteristic parameters and reaction time

Note: |r| is the absolute value of the Pearson correlation coefficient, and the value of |r| reflects the strength of the correlation between the two. ** indicates that the correlation is extremely significant at the 0.01 significance level; * indicates that the correlation is significant at the 0.05 significance level.

It can be seen from the table that the three EEG characteristic parameters have a significant correlation with the reaction time to varying degrees, which once again demonstrates the high correlation between the EEG signal and the reaction time in previous studies. Judging from the strength of the correlation, it is obvious that the EEG characteristic parameter α is more related to the reaction time than the other two items, and at the same time it is the most significant.

In addition, existing studies have shown that changes in the activity of the cerebral cortex can directly reflect the driver's mental state, and the EEG signal is highly correlated with the driver's mental state (fatigue, lethargy, etc.). In the classic cognitive psychology experiments such as visual detection, it has also been demonstrated that there is a high correlation between the change of EEG signal and the reaction time. However, this study verified the correlation between the EEG characteristic parameters and the reaction time in practice. During the operation of the driving task, it is also related to the reaction time of the driver of the EMU.

(2) Analysis of model output results. In order to determine the optimal network structure of the BP neural network prediction model, and in order to reflect the integrity and representativeness of the training samples, the embodiment of the present invention randomly selected 7 samples from the experimental samples of 20 EMU drivers, and compared the obtained 140 samples. The sample is used as a training sample, and the maximum absolute error and relative mean square error generated by the final prediction result are used as evaluation criteria. After several repeated trainings, the optimal network structure was determined to be 96-20-1 (input layer-hidden layer-output layer). Combined with the optimal network structure, the above 140 samples (75% of the samples were randomly selected as training samples and the rest as test samples) were substituted into the model for retraining and testing. The prediction results are shown in Figure 5 and Figure 6.

It can be seen from Figures 5 and 6 that the model prediction results are closer to the actual results, indicating that the model prediction effect is better. Therefore to each driver's experimental sample, 75% are randomly selected as training samples, and the rest are used as test samples to train and test the neural network model respectively. Simultaneously, in order to understand the accuracy of the proposed model, the present invention uses existing The Yesian prediction model predicts the reaction time, and the obtained prediction results are shown in Table 2.

Table 2 The prediction results obtained by each driver in the prediction model

On the whole, the maximum absolute error (11.01%) and relative mean square error (8.15%) of the results obtained by artificial neural network are higher than those of Bayesian prediction model (14.6%, 10.7%). At the same time, it can be seen from the prediction results of each driver's reaction time that the maximum absolute error obtained by artificial neural network is less than 15%, while the maximum relative mean square error is 10.89%, and the minimum relative mean square error is 6.93%; In the results obtained by using the Bayesian prediction model, the maximum absolute error is between 11% and 20%, while the maximum relative mean square error is 15.36%, and the minimum relative mean square error is 8.14%, which shows that the prediction accuracy of artificial neural network is high. in Bayesian predictive models. Therefore, the prediction model proposed in the embodiment of the present invention is reliable, and can accurately predict the reaction time of the EMU driver to an emergency, thereby effectively reducing the accident rate.

As can be seen from the above, the embodiment of the present invention is based on the 2h motor car simulation driving experiment, and the prediction of the reaction time of the driver of the motor car unit to the emergency event has been studied, and its beneficial results and conclusions are as follows:

(1) Based on the EEG signals of EMU drivers, wavelet transform is used to extract three EEG indicators, θ, α, and β, which can be used for the evaluation of the EMU driver's reaction level and the prediction of reaction time. Combined with BP neural network, a prediction model for the reaction time of EMU drivers is constructed.

(2) It can be seen from the results that the three EEG characteristic parameters have a significant correlation with the reaction time, indicating that the EEG signal can directly reflect the driver's mental state, thus providing a research basis for the prediction of the reaction time.

(3) The final results show that the maximum absolute error between the driver's reaction time predicted by the model and the driver's actual reaction time is 11.01% (1.75%), and the relative mean square error is 8.51% (1.37%), which is lower than other predictions model, which shows that the method has high accuracy.

The embodiment of the present invention realizes the accurate prediction of the reaction time of the EMU driver to an emergency, and the research results provide a theoretical basis for the development and design of the vehicle-mounted real-time EMU driver's dangerous driving state warning system. In the future, the applicability of this method in the actual running state of the train can be further verified.

Those of ordinary skill in the art can understand that all or part of the steps/units/modules of the above embodiments can be implemented by hardware related to program instructions, and the aforementioned programs can be stored in a computer-readable storage medium. When the program is executed, Execution includes the corresponding steps in the units of the above-mentioned embodiments; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk and other various media that can store program codes.

The specific embodiments described above have further described the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above descriptions are only specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims (9)

1. a kind of time of driver's reaction Forecasting Methodology, it is characterised in that methods described includes step：

Obtain the EEG signals of driver；

The brain electrical feature parameter in the EEG signals after filtered processing is extracted by wavelet transformation；

Using the brain electrical feature parameter of driver as input layer, using the reaction time as output layer, BP neural network prediction is built Model.

2. time of driver's reaction Forecasting Methodology as claimed in claim 1, it is characterised in that the filtering process is specifically wrapped Include：EEG signals carry out overall filtering process with 0-35Hz bandwidth.

3. time of driver's reaction Forecasting Methodology as claimed in claim 2, it is characterised in that brain electricity is extracted by wavelet transformation Characteristic parameter is specifically included：

A, for the EEG signals after filtered processing, be designated as u (n), then its wavelet transformation is defined as：

In formula,For wavelet function；I is frequency factor；M is the time-shifting factor；N is signal duration；

Signal u (n) after b, wavelet transformation carries out finite layer decomposition, using Mallat algorithms：

<mrow> <mi>u</mi> <mrow> <mo>(</mo> <mi>n</mi> <mo>)</mo> </mrow> <mo>=</mo> <msub> <mi>A</mi> <mi>H</mi> </msub> <mo>+</mo> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>H</mi> </munderover> <msub> <mi>C</mi> <mi>i</mi> </msub> </mrow>

In formula, AH is approximation component；Ci is the details coefficients under different scale；H is Decomposition order；Decomposed by above-mentioned finite layer Obtain the wavelet coefficient of 3 kinds of different frequency ranges of θ, α, β；

C, the energy value to above-mentioned frequency extraction wavelet coefficient are used as brain electrical feature parameter：

<mrow> <msub> <mi>P</mi> <mi>X</mi> </msub> <mo>=</mo> <mfrac> <mn>1</mn> <mi>t</mi> </mfrac> <msup> <mrow> <mo>|</mo> <msub> <mi>S</mi> <mi>X</mi> </msub> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> <mo>|</mo> </mrow> <mn>2</mn> </msup> <mi>d</mi> <mi>t</mi> <mo>=</mo> <mfrac> <mn>1</mn> <mi>n</mi> </mfrac> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msup> <mrow> <mo>|</mo> <msub> <mi>h</mi> <mi>i</mi> </msub> <mo>|</mo> </mrow> <mn>2</mn> </msup> </mrow>

In formula, P_XFor the energy value of corresponding band；S_X(t) it is the wavelet coefficient of corresponding band；T is the time；h_iFor corresponding band Amplitude；

D, the EEG signals to q electrode are handled, and accordingly obtain 3 × q brain electrical feature parameters.

4. time of driver's reaction Forecasting Methodology as claimed in claim 3, it is characterised in that methods described also includes：To institute Obtained brain electrical feature parameter is normalized by formula, makes the numerical value of brain electrical feature parameter between [0,1], to eliminate number The noise present in：

<mrow> <mi>x</mi> <mo>=</mo> <mfrac> <mrow> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>-</mo> <msub> <mi>x</mi> <mrow> <mi>i</mi> <mi>m</mi> <mi>i</mi> <mi>n</mi> </mrow> </msub> </mrow> <mrow> <msub> <mi>x</mi> <mrow> <mi>i</mi> <mi>m</mi> <mi>a</mi> <mi>x</mi> </mrow> </msub> <mo>-</mo> <msub> <mi>x</mi> <mrow> <mi>i</mi> <mi>m</mi> <mi>i</mi> <mi>n</mi> </mrow> </msub> </mrow> </mfrac> </mrow>

In formula, x_{i max}With x_{i min}For brain electrical feature parameter x_iMaxima and minima.

5. the time of driver's reaction Forecasting Methodology as described in any one of Claims 1-4, it is characterised in that build BP nerves Network Prediction Model is specifically included：Using the brain electrical feature parameter of driver as input layer, using reaction time predicted value as defeated Go out layer, build the BP neural network forecast model containing 1 hidden layer；Wherein, input layer number is special by the brain electricity of driver Levy number of parameters decision；Hidden layer node number s is then preferentially chosen by model training result；Output layer node number is 1.

6. time of driver's reaction Forecasting Methodology as claimed in claim 5, it is characterised in that input layer is to hidden in the model It is respectively w containing layer, hidden layer to the connection weight coefficient between output layer and biasing_ik,w_k1,b_ik,b_k1(i=1,2 ..., 3 × Q, k=1,2 ..., s), for input layer arbitrary node o to hidden layer arbitrary node p output：

y_op=f (x_iw_op+b_op)

In formula, f () is Sigmoid functions, i.e.,

<mrow> <mi>f</mi> <mrow> <mo>(</mo> <mi>&amp;sigma;</mi> <mo>)</mo> </mrow> <mo>=</mo> <mfrac> <mn>1</mn> <mrow> <mn>1</mn> <mo>+</mo> <msup> <mi>e</mi> <mrow> <mo>-</mo> <mi>&amp;sigma;</mi> </mrow> </msup> </mrow> </mfrac> </mrow>

Output layer output result is：

<mrow> <msub> <mi>y</mi> <mi>i</mi> </msub> <mo>=</mo> <mi>f</mi> <mrow> <mo>(</mo> <msubsup> <mi>W</mi> <mn>2</mn> <mi>T</mi> </msubsup> <mo>(</mo> <mrow> <mi>f</mi> <mrow> <mo>(</mo> <mrow> <msubsup> <mi>W</mi> <mn>1</mn> <mi>T</mi> </msubsup> <msub> <mi>x</mi> <mi>i</mi> </msub> </mrow> <mo>)</mo> </mrow> <mo>+</mo> <msub> <mi>b</mi> <mn>1</mn> </msub> </mrow> <mo>)</mo> <mo>+</mo> <msub> <mi>b</mi> <mn>2</mn> </msub> <mo>)</mo> </mrow> </mrow>

In formula, y_iExported for Neural Network model predictive result；W₁For the connection weight number system matrix number of input layer to hidden layer；W₂ For the connection weight number system matrix number of hidden layer to output layer；x_iFor driver's brain electrical feature parameter；b₁For input layer to hidden layer Bias matrix；b₂For the bias matrix of hidden layer to output layer.

7. time of driver's reaction Forecasting Methodology as claimed in claim 6, it is characterised in that the brain electricity of the acquisition driver Signal is specifically included：Gather the EEG signals of the driver of predetermined number.

8. time of driver's reaction Forecasting Methodology as claimed in claim 7, it is characterised in that eeg signal acquisition frequency is 10Hz；Electroencephalogramdata data collector continuous acquisition driver's eeg data.

9. time of driver's reaction Forecasting Methodology as claimed in claim 7, it is characterised in that obtained by BP neural network model Comparison of the reaction time predicted value with testing gathered reaction time actual value fitting degree use maximum absolute error M1 Prediction effect assessment indicator is used as with relative mean square error M2：

<mrow> <msub> <mi>M</mi> <mn>1</mn> </msub> <mo>=</mo> <munder> <mrow> <mi>m</mi> <mi>a</mi> <mi>x</mi> </mrow> <mi>i</mi> </munder> <mrow> <mo>|</mo> <mfrac> <mrow> <msub> <mi>y</mi> <mi>i</mi> </msub> <mo>-</mo> <msubsup> <mi>y</mi> <mi>i</mi> <mo>*</mo> </msubsup> </mrow> <msub> <mi>y</mi> <mi>i</mi> </msub> </mfrac> <mo>|</mo> </mrow> <mo>&amp;times;</mo> <mn>100</mn> <mi>%</mi> </mrow>

<mrow> <msub> <mi>M</mi> <mn>2</mn> </msub> <mo>=</mo> <msqrt> <mrow> <mfrac> <mn>1</mn> <mi>N</mi> </mfrac> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>N</mi> </munderover> <msup> <mrow> <mo>(</mo> <mfrac> <mrow> <msub> <mi>y</mi> <mi>i</mi> </msub> <mo>-</mo> <msubsup> <mi>y</mi> <mi>i</mi> <mo>*</mo> </msubsup> </mrow> <msub> <mi>y</mi> <mi>i</mi> </msub> </mfrac> <mo>)</mo> </mrow> <mn>2</mn> </msup> </mrow> </msqrt> <mo>&amp;times;</mo> <mn>100</mn> <mi>%</mi> <mo>.</mo> </mrow> 2