CN105726046B - A kind of driver's alertness condition detection method - Google Patents

Abstract

The invention discloses a method for detecting a driver's vigilance state, and relates to the technical field of active safety for automobile drivers. It can effectively solve the problem of prompting and alarming when the driver is in fatigue driving. It mainly includes the following steps: using the chip, wireless transmission technology and peripheral circuit EEG signal processing module to collect the EEG signal data of the driver in a state of wakefulness or fatigue, edit the collected data into a redundant dictionary, The EEG signals at different time periods on the way are collected in real time, and transmitted to the EEG signal processing module through wireless transmission technology; the EEG signals are denoised using the discrete wavelet transform (DWT) algorithm, down-sampled the EEG signals, and feature extraction of the EEG signals 1. Mode classification of the driver's alertness state. If it is judged that the driver is in a fatigue driving situation, a control signal is output to control the warning device located on the driver's seat to vibrate or buzz the driver. Mainly used for long-distance driving fatigue alarm.

Description

A driver alertness state detection method

technical field

The invention relates to the technical field of active safety for automobile drivers.

technical background

Traffic safety has become a worldwide problem. According to literature review, the number of deaths due to traffic accidents in the world is about 500,000 every year. In recent years, due to the continuous increase of cars in cities, traffic accidents caused by driver fatigue have also increased significantly.

When the driver is in fatigue driving, the judgment ability decreases, the response is slow and the operation error increases. When the driver is slightly fatigued, the gear shift will not be timely or accurate; when the driver is moderately fatigued, the operation action will be sluggish, and sometimes even forget to operate; Sleep phenomenon, in severe cases, will lose the ability to control the vehicle. When the driver is tired, there will be phenomena such as blurred vision, sore back, dull movements, swollen hands and feet, lack of concentration, slow response, incomplete thinking, laxity, anxiety, and impatience. If you still drive the vehicle reluctantly, it may lead to traffic accidents. At present, the alertness detection for drivers mainly focuses on the following three aspects: 1. By detecting the biological signals of the human body, such as EEG signals, EMG signals, heart rate, etc., to judge the state of alertness of the human body; 2. The state of alertness of the human body is judged by detecting the movement of the human body, such as the movement of the head; third, the alertness state of the human body is detected by detecting the blinking frequency of the human body.

Human alertness detection based on EEG signals is the "gold standard" in alertness detection. Compared with other physiological signals, it can more directly reflect the brain's own activities and has a higher temporal resolution. The Chinese Patent Publication No. CN102622600A discloses "A High-Speed Train Driver Alertness Detection Method Based on Facial Image and Eye Movement Analysis". It mainly uses the driver's facial image and eye movement to analyze the driver's alertness state, while the present invention is based on the alertness analysis of the driver's EEG signal, which has better accuracy. The Chinese patent with publication number CN102058413B is based on wavelet transform-based EEG signal alertness detection method. It uses the wavelet function to obtain the eigenvalues of the wavelet coefficients of the EEG signal sequence as the feature set, and then uses the random forest method to sort and simplify the feature set, then uses the sample training support vector machine, and uses the trained support vector machine to analyze the EEG signal. Perform an alertness check. The Chinese patent application with application number 201520878190.7 discloses a wearable driver's EEG signal acquisition headband. It mainly describes the structure of the EEG signal acquisition equipment, but does not describe the EEG signal processing and algorithm in detail. The purpose of this application needs to be realized by using the "headband".

Contents of the invention

The purpose of the present invention is to provide a driver alertness state detection method, which can effectively solve the problem of prompting and alarming when the driver is in fatigue driving.

The technical scheme that the present invention realizes that its object of the invention adopts is: a kind of driver alertness state detection method comprises the following steps:

1. A driver alertness state detection method, comprising the following steps:

The first step is to build functional modules

Use the development board including the DSP chip TMS320F28335 with wireless transmission function to build the EEG signal processing module;

The second step, the collection of the driver's EEG signal

First, the driver wears an EEG signal collection headband to collect the EEG signal data of the driver in a state of wakefulness or fatigue, edit the collected data into a redundant dictionary, and store it in the EEG signal processing module Middle; secondly, real-time collection of the EEG signals of the driver at different periods of time while driving, and the collected EEG signals of the driver are transmitted to the EEG signal processing module through wireless transmission technology;

The third step, denoising of EEG signal

The EEG signal processing module uses the discrete wavelet transform (DWT) algorithm to carry out a 5dB wavelet 6-layer decomposition of the original EEG signal to the collected driver's EEG signal, and obtains the sub-band wavelet detail coefficient (D _i , i=1 ,2,3,4,5,6) and approximation coefficients (A _i , i=1,2,3,4,5,6); by reconstructing the decomposition coefficients D ₃ ,, D ₄ , D ₅ , D ₆ to remove Low-frequency and high-frequency interference in the original signal to obtain useful EEG signals;

The fourth step, downsampling of EEG signals

The EEG signal processing module performs down-sampling processing on the driver's EEG signal after denoising, and the sampling frequency is 128Hz;

The fifth step, feature extraction of EEG signal

The EEG signal processing module performs feature extraction based on the Fast Fourier Transform (FFT) algorithm for the EEG signal after down-sampling processing, and the FFT algorithm is 128 points;

Step 6. Pattern Classification of Driver Alertness State

The EEG signal processing module carries out pattern classification based on the K-SVD sparse classification representation algorithm to the driver's EEG signal through feature extraction, and obtains the driver's alertness state class(y);

class(y)＝argmin _i ||y-Dα′ _i || ₂

Among them, y is the data to be classified. D=[D _a , D _d ] is the redundant dictionary of the driver's alertness state, D _a is the redundant dictionary of the driver in the awake state, and D _d is the redundant dictionary of the driver in the fatigue state. Yu dictionary, α is the data to be classified, and the sparse coefficient of y;

The seventh step, alarm reminder

If the driver's alertness status transmitted from the EEG signal processing module is judged to be the driver's fatigue driving condition, a control signal is output to control the warning device located on the driver's seat to vibrate or buzz to remind the driver.

Compared with prior art, the beneficial effect of the present invention is:

1. The N=128 FFT algorithm is used in the feature extraction stage, which greatly reduces the amount of calculation compared with the traditional discrete Fourier transform (DFT) algorithm, making the feature extraction time only about 1/37 of the DFT algorithm. Thereby reducing the time of the whole process.

Second, the sparse representation classification algorithm based on K-SVD is used in the pattern classification stage. The purpose of signal sparse representation is to use as few atoms as possible to represent the signal in a given over-complete dictionary, so that a more concise representation of the signal can be obtained, so that we can more easily obtain the information contained in the signal, K- SVD is a classic dictionary training algorithm. According to the principle of minimum error, the error item is decomposed by SVD, and the decomposition item that minimizes the error is selected as the updated dictionary atom and the corresponding atomic coefficient. After continuous iteration, an optimized solution is obtained.

Description of drawings

FIG. 1 is a flow chart of the implementation of the alertness detection algorithm based on EEG signals in the present invention.

Fig. 2 is a time-domain comparison diagram of the EEG signal before and after denoising according to the present invention.

Fig. 3 is a comparison diagram of the frequency domain of the EEG signal before and after denoising according to the present invention.

Fig. 4 is a time-domain comparison diagram between the awake state and the fatigue state after denoising the EEG signal according to the present invention.

Fig. 5 is a frequency-domain comparison diagram between the awake state and the fatigue state after denoising the EEG signal according to the present invention.

Specific implementation method

Embodiment 1, the subject of the experiment is a driver who has driven for more than three hours and is recorded as driver 1.

Its concrete implementation method is, a kind of driver's vigilance state detection method, comprises the following steps:

A. Acquisition of the driver's EEG signal

The driver wears an EEG signal collection cap on his head. When the driver is driving the car, the EEG signal collection cap collects the driver's EEG signal, obtains the EEG signal f ₁ of driver 1, and collects the driver's EEG signal f 1 . The EEG signal _f1 is transmitted to the EEG signal processing module through wireless transmission technology;

B. Denoising of the driver's EEG signal

The EEG signal processing module uses the DWT algorithm to denoise the driver's EEG signal f ₁ collected by the EEG signal acquisition cap, and obtains the driver's EEG signal g ₁ after denoising;

C. Downsampling of the driver's EEG signal

The EEG signal processing module performs down-sampling processing on the denoised driver's EEG signal g ₁ to obtain the down-sampled driver's EEG signal k ₁ to reduce the amount of sample data;

D. Feature extraction of the driver's EEG signal

The EEG signal processing module carries out feature extraction based on the Fast Fourier Transform (FFT) algorithm to the driver's EEG signal _k after the down-sampling process, and obtains the data _Y to be classified;

E. Pattern Classification of Driver Alertness States

The EEG signal processing module carries out pattern classification based on the K-SVD sparse classification representation algorithm to the driver's EEG signal through feature extraction, and obtains the driver's alertness state class(y);

class(y)＝argmin _i ||Y ₁ -Dα′ _i || ₂

Among them, Y ₁ is the data to be classified, D=[D _a , D _d ] is the redundant dictionary of the driver's state, D _a is the redundant dictionary of the driver in the awake state, and D _d is the redundant dictionary of the driver in the fatigue state. Yu dictionary, α is the sparse coefficient of the data to be classified y;

The result obtained is that the driver is in a state of fatigue, and the confidence level of the test is 0.9565.

Embodiment 2/3/4 Repeat the operations of steps A to E above.

The inventive method sees the following table to the driver vigilance detection test result.

driver number 1 2 3 4 Experimental results fatigue fatigue fatigue wide awake Confidence 0.9565 0.9504 0.9647 0.9387

It can be seen that the present invention has a high degree of confidence in the detection of driver alertness.

Claims (1)

1. a kind of driver's alertness condition detection method, includes the following steps：

The first step, constructing function module

Using including having the development board of the dsp chip TMS320F28335 of wireless transmission function, EEG Processing mould is built Block；

The acquisition of second step, driver's EEG signals

First, one eeg signal acquisition headband of the head-mount of driver is in driver in awake or fatigue state brain Electrical signal data is acquired, and by collected data edition redundant dictionary, and is existed in storage EEG Processing module；Its It is secondary, the EEG signals of different periods in driver drives vehicle way are acquired in real time, collected driver's EEG signals pass through Radio Transmission Technology is transferred to EEG Processing module；

The denoising of third step, EEG signals

EEG Processing module is to the EEG signals of collected driver with wavelet transform (DWT) algorithm to original EEG signals carry out 6 layers of decomposition of a 5dB small echo, obtain subband wavelet details coefficient (D_i, i=1,2,3,4,5,6) and it is near Like coefficient (A_i, i=1,2,3,4,5,6)；By rebuilding decomposition coefficient D₃, D₄, D₅, D₆Remove the low frequency and high frequency in original signal Interference is so as to obtain useful EEG signals；

The down-sampling of 4th step, EEG signals

To carrying out down-sampling processing by the EEG signals of the driver after denoising, sample frequency is EEG Processing module 128Hz；

The feature extraction of 5th step, EEG signals

EEG Processing module carries out based on Fast Fourier Transform (FFT) (FFT) EEG signals after down-sampling processing The feature extraction of algorithm, fast fourier transform algorithm are 128 points；

The pattern classification of 6th step, driver's alertness state

EEG Processing module carries out the EEG signals of the driver Jing Guo feature extraction the sparse classification chart based on K-SVD Show the pattern classification of algorithm, obtain the alertness state class (y) of driver；

Class (y)=argmin_i||y-Dα'_i||₂

Wherein y is data D=[D to be sorted_a,D_d] be driver's alertness state redundant dictionary, D_aIt is in clear for driver The redundant dictionary for the state of waking up, D_dThe redundant dictionary of fatigue state is in for driver, α is the sparse coefficient of data y to be sorted；

7th step, warning reminding

Driver's alertness state that EEG Processing module transmits, if judging to be driver there are fatigue driving situation, Output control signal, the alarming device that control is located on pilot set shakes driver or buzzing is reminded.