CN106251583B - Fatigue driving discrimination method based on driving behavior and eye movement characteristics - Google Patents
Fatigue driving discrimination method based on driving behavior and eye movement characteristics Download PDFInfo
- Publication number
- CN106251583B CN106251583B CN201610869724.9A CN201610869724A CN106251583B CN 106251583 B CN106251583 B CN 106251583B CN 201610869724 A CN201610869724 A CN 201610869724A CN 106251583 B CN106251583 B CN 106251583B
- Authority
- CN
- China
- Prior art keywords
- fatigue
- standard deviation
- driving
- eye movement
- parameter
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
- G08B21/02—Alarms for ensuring the safety of persons
- G08B21/06—Alarms for ensuring the safety of persons indicating a condition of sleep, e.g. anti-dozing alarms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Emergency Management (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention discloses a kind of fatigue driving discrimination method based on driving behavior and eye movement characteristics, includes the following steps：1）Fatigue driving simulated experiment is designed, the eye movement data of driving behavior data and driver of the driver under different fatigue state is acquired by experiment；2）Cutting screening is carried out to experimental data, establishes fatigue driving sample database；3）It is for statistical analysis to the driving behavior parameter of driver under different fatigue state and the conspicuousness of eye movement characteristics parameter differences using the method for one-way analysis of variance, and have found the optimal time window of various characteristic parameters；4）Bivariate Spearman correlation analyses are carried out to the characteristic parameter that preliminary screening goes out；5）Filter out optimal characteristics parameter；6）Establish fatigue driving BP neural network identification model；7）Using Matlab software programming model programs, randomly selects training sample and test sample set pair model is trained and is recognized with fatigue state.
Description
Technical field
The present invention relates to a kind of fatigue driving discrimination method based on driving behavior and eye movement characteristics.
Background technology
With the fast development of communication, the continuous of vehicle guaranteeding organic quantity is risen, and road traffic accident quantity is increasingly
More, traffic safety problem has become a serious social concern.Fatigue driving is to lead to the important original of traffic accident
One of because, in recent years, the correlative study project about fatigue driving is more and more, and fatigue-driving detection technology has also obtained rapidly
Development, existing research have been achieved with fatigue detecting pretty good effect, but still in place of Shortcomings：
1) driver is a progressive formation from regaining consciousness fatigue, this process is often ignored in present research, to driving
When member can not judge that the actual conditions that cannot be also directed to driver give rational prompt message into fatigue state, therefore,
Reasonably the fatigue state of driver is divided and accurate recognition will be the emphasis studied；
2) existing to be based on single finger object detection method, the environment such as space, illumination, weather can not be overcome to accuracy of detection
It influences, will be the important channel for improving detection accuracy and reliability using the detection method merged based on information therefore.
Invention content
It is recognized based on driving behavior and the fatigue driving of eye movement characteristics the technical problem to be solved in the present invention is to provide a kind of
Method.
Fatigue driving discrimination method based on driving behavior and eye movement characteristics, includes the following steps：
1) fatigue driving simulated experiment is designed, driving behavior number of the driver under different fatigue state is acquired by experiment
According to, including steering wheel angle, tarnsition velocity, speed, vehicle acceleration and the eye movement data of driver, including blink, watch attentively, sweep
Depending on data；
2) cutting screening is carried out to experimental data, establishes fatigue driving sample database；
3) use the method for one-way analysis of variance to the driving behavior parameter and eye movement of driver under different fatigue state
The conspicuousness of characteristic parameter difference is for statistical analysis, and has found the optimal time window of various characteristic parameters；Preliminary screening goes out
Driving behavior parameter includes steering wheel absolute mean SAM, corner standard deviation SASTD, steering wheel angular velocity absolute mean
SWM, angular speed standard deviation SWSTD, velocity standard difference Vstd, acceleration mean value Am and standard deviation Astd；Eye movement characteristics parameter packet
Include frequency of wink BF, duration of fixation mean value FIXT_mean and standard deviation FIXT_std, pan average speed standard deviation
SACV_std and pupil diameter coefficient of variation CVPLD；
4) bivariate Spearman correlation analyses are carried out to the characteristic parameter that preliminary screening goes out, be as follows：
4-1) source and the characteristic parameter of same index are analyzed, the significant feature of correlation can only select one；
4-2) to the characteristic parameter filtered out, correlation analysis is carried out by driving behavior parameter and eye movement parametric classification, together
Sample can only select one to the significant index of correlation；
4-3) to the remaining index progress correlation analysis after above two screenings, the significant index of correlation is adopted
Take step 4-1), 4-2) same processing mode；
5) filtering out optimal characteristics parameter includes：Steering wheel angle standard deviation SASTD, vehicle acceleration standard deviation Astd,
Sweep average speed standard deviation SACV_std and pupil diameter coefficient of variation CVPLD；
6) fatigue driving BP neural network identification model is established, it is first determined the input neuron number of the network structure, defeated
Go out neuron number, the implicit number of plies, hidden layer neuron number；The transmission function of next network layer, the power of each layer initial value of setting
Value finally determines best learning efficiency, chooses anticipation error appropriate, specific as follows：
6-1) input neuron, select optimal characteristics parameter and the input as model in step 5), and by input data into
Row normalized；
6-2) output neuron, setting output are that 1- is awake, 2- is tired, 3- is very tired, and is indicated by a neuron,
It is boundary with 1.5 and 2.5 in the prediction result of network, is awake if predicted value 1≤k≤1.5；If predicted value 1.5
≤ k≤2.5 are very tired if being then fatigue predicted value 2.5≤k≤3；
6-3) imply the number of plies, single hidden layer, node in hidden layer 9；
6-4) transmission function, using tansig as transmission function, output layer uses purelin for the neural network inside of BP
As transmission function；
6-5) initial weight, initial weight are set to the non-zero random value between (- 1,1)；
6-6) learning efficiency and anticipation error, using LM learning algorithms, learning efficiency is automatically adjusted according to anticipation error, nothing
It needs to set；Anticipation error is set as 0.001；
7) Matlab software programming model programs are utilized, training sample is randomly selected and test sample set pair model is instructed
Practice and fatigue state recognizes.
Further, the fatigue driving simulated experiment is specific as follows：
1) experimental facilities：
A, driving analog system, by emulation vehicle, control taiwan area, visual system, computer system, network system and phase
The software systems composition answered；Primary operational equipment is all real vehicles equipment, and the input action of user is captured by sensor, will
The input of user and other correlations carry out analog simulation calculating, at the same simulation result shown by visual system, sound is imitated
Corresponding informance is fed back to user by true system and signal output system；System further includes this vehicle phase in emulation vehicle operation
The export interface of data is closed, can different sample frequencys be set according to Research Requirements, export data carry out secondary analysis；
B, II type eye trackers of Eye Link, are tracked using pupil pattern；
2) experiment scene：Using dull beltway as experiment scene, the dullness beltway is that an annular is double
To the suburb dullness high-grade highway in 6 tracks, lane width 3.75m, road overall length 20km, evenness of road surface, weather is set as fine
It；
3) experimenter：Health, rule of sleeping is interior for 24 hours before test not allow to drink, and does not allow in 12h before testing
Coffee for drinking, tea；
4) experimental program：The speed of 100km/s is kept to be travelled in middle lane, staff's primary subject of record per 5min
The Subjective fatigue grade (being demarcated by sleep scale table Karolinska sleepiness scale, KSS) of personnel, until traveling
60min, experiment terminate.
The beneficial effects of the invention are as follows：
The present invention extracts new fatigue driving Testing index, and verify by variance analysis in existing achievement in research
The validity of index, establishes fatigue driving identification model based on information fusion technology, overcomes based on the tired of single piece of information source
The limitation of labor discrimination method.
Specific implementation mode
The present invention is further elaborated for following specific examples, but not as a limitation of the invention.
1, fatigue driving simulated experiment is designed, driving behavior number of the driver under different fatigue state is acquired by experiment
According to specific as follows：
1) experimental facilities：
A, driving analog system, by emulation vehicle, control taiwan area, visual system, computer system, network system and phase
The software systems composition answered；Primary operational equipment is all real vehicles equipment, and the input action of user is captured by sensor, will
The input of user and other correlations carry out analog simulation calculating, at the same simulation result shown by visual system, sound is imitated
Corresponding informance is fed back to user by true system and signal output system；System further includes this vehicle phase in emulation vehicle operation
The export interface of data is closed, can different sample frequencys be set according to Research Requirements, export data carry out secondary analysis；
B, II type eye trackers of Eye Link, are tracked using pupil pattern；
2) experiment scene：Using dull beltway as experiment scene, the dullness beltway is that an annular is double
To the suburb dullness high-grade highway in 6 tracks, lane width 3.75m, road overall length 20km, evenness of road surface, weather is set as fine
It；
3) experimenter：9 subject personnel (men：7, female：2), age 23-27 Sui, driving age 2-5, health,
Sleep rule, it is interior for 24 hours before test not allow to drink, do not allow coffee for drinking, tea in 12h before testing；
4) experimental program：The speed of 100km/s is kept to be travelled in middle lane, staff's primary subject of record per 5min
The Subjective fatigue grade (being demarcated by sleep scale table Karolinska sleepiness scale, KSS) of personnel, until traveling
60min, experiment terminate.
2, cutting screening is carried out to experimental data, establishes fatigue driving sample database；
It is that an interval carries out cutting by 5min, i.e. 5min data are participated in real as a sample, in total 9 subjects
It tests, totally 108 samples；To driving behavior data, the sample data operated containing bend and thread-changing is rejected, for eye movement data,
It is limited by II type eye trackers of Eye Link itself, intelligence is to eyeball opposing headers in horizontal vertical direction [- 30 °, 30 °] range
Eye movement data recorded, therefore the data other than [- 30 °, 30 °] range will be rejected.
After screening, 102, remaining valid data sample, according to the evaluation method of driving fatigue, by 102 samples
Fatigue state division is carried out, establishes fatigue driving sample database, as shown in table 1 below.
1 fatigue driving sample data information of table
3, using the method for one-way analysis of variance to the driving behavior parameter and eye movement of driver under different fatigue state
The conspicuousness of characteristic parameter difference is for statistical analysis, and has found the optimal time window of various characteristic parameters；Preliminary screening goes out
Driving behavior parameter includes steering wheel absolute mean SAM, corner standard deviation SASTD, steering wheel angular velocity absolute mean
SWM, angular speed standard deviation SWSTD, velocity standard difference Vstd, acceleration mean value Am and standard deviation Astd；Eye movement characteristics parameter packet
Include frequency of wink BF, duration of fixation mean value FIXT_mean and standard deviation FIXT_std, pan average speed standard deviation
SACV_std and pupil diameter coefficient of variation CVPLD；
1) to the intuitive analysis of steering wheel angle, can show that the fatigue characteristic of driver typically lasts for 5-15s, 5s,
The related each fatigue characteristic parameter of steering wheel angle is separately designed in tri- time windows of 10s, 15s.Utilize one-way analysis of variance method
The otherness of steering wheel angle absolute value and standard deviation of the quantitative study driver under different fatigue degree, in significance
In the case of α=0.05, the results of analysis of variance is as shown in table 2.
The results of analysis of variance of the otherness of 2 steering wheel angle absolute value of table and standard deviation
Steering wheel angle absolute value SAM, when time window is 10s, F=3.893 is maximum, steering wheel angle absolute value SAM
Optimal time window be 10s；Steering wheel angle standard deviation SASTD, when time window is 10s, F=4.801 is maximum, and steering wheel turns
The optimal time window of mean angular deviation SASTD is 10s.
2) waking state, fatigue state and each 10 sample datas very under fatigue state are randomly selected, respectively press 5s,
Tri- time windows of 10s, 15s calculate steering wheel angular velocity absolute mean SWM, tarnsition velocity standard deviation SWSTD, utilize single factor test side
Difference analyses otherness of the legal quantity research driver under different fatigue degree, in the case of level of significance α=0.05, side
The results are shown in Table 3 for difference analysis.
The results of analysis of variance of table 3 steering wheel angular velocity absolute mean and standard deviation
Steering wheel angular velocity absolute mean SWM, when time window is 5s, F=4.923 is maximum, and steering wheel angular velocity is absolute
The optimal time window of mean value SWM is 5s；Tarnsition velocity standard deviation SWSTD, when time window is 5s, F=4.907 is maximum, corner
The optimal time window of velocity standard difference SWSTD is 5s.
3) speed mean value Vm and vehicle speed standard deviation Vstd is calculated in tri- time windows of 40s, 80s, 120s, utilizes single factor test
Otherness of the method for analysis of variance quantitative study driver under different fatigue degree, in the case of level of significance α=0.05,
The results of analysis of variance is as shown in table 4.
The results of analysis of variance of table 4 speed mean value and standard deviation
It is anisotropic that there is no significance differences between speed mean value Vm different fatigue levels, therefore speed mean value is unsuitable for fatigue driving
State-detection；Vehicle speed standard deviation Vstd, when time window is 80s, F=46.424 values are maximum, vehicle speed standard deviation Vstd's
Optimal time window is 80s.
4) acceleration absolute value and standard deviation are calculated in tri- time windows of 25s, 50s, 75s, utilizes one-way analysis of variance
Otherness of the legal quantity research driver under different fatigue degree, in the case of level of significance α=0.05, variance analysis
The results are shown in Table 5.
The results of analysis of variance of table 5 acceleration absolute value and standard deviation
For acceleration absolute value Am and acceleration standard deviation Astd, when time window is 25s, the two is in different fatigue shape
Difference between state is most notable, and therefore, acceleration absolute value Am and acceleration standard deviation Astd optimal time windows are 25s.
5) 10 sample datas of fatigue state carry out one-way analysis of variance from randomly selecting 3 in sample database,
Time window is 30s and 60s, and for frequency of wink in (α=0.05), the results of analysis of variance is as shown in table 6 below.
The results of analysis of variance of 6 frequency of wink of table
No matter time window is 30s or 60s, there is significant difference in frequency of wink BF, and access time window is 30s.
6) time window is 30s and 60s, mean value and standard deviation to driver's duration of fixation under different fatigue state into
Row one-way analysis of variance, as a result as shown in table 7 below.
The results of analysis of variance of table 7 duration of fixation mean value and standard deviation
There is significant difference in duration of fixation mean value and standard deviation, upon extracting between window when being 30s, F values are maximum, because
This, optimal time window is 30s.
7) time window is 30s and 60s, and Dan Yin is carried out to driver's sweep amplitude mean value under different fatigue state and standard deviation
Plain variance analysis, as a result as shown in table 8 below.
The results of analysis of variance of table 8 sweep amplitude mean value and standard deviation
No matter time window is 30s or 60s, sweep amplitude mean value and standard deviation in the given feelings of level of signifiance α=0.05
Under condition, significant difference is not present, therefore sweep amplitude cannot function as the characteristic parameter of fatigue driving differentiation.
8) time window be 30s and 60s, under different fatigue state driver sweep average speed mean value and standard deviation into
Row one-way analysis of variance, as a result as shown in table 9 below.
Table 9 sweeps the results of analysis of variance of average speed mean value and standard deviation
No matter time window is 30s or 60s, pan average speed mean value in given level of signifiance α=0.05,
There is no significant differences, therefore sweep the characteristic parameter that average speed mean value cannot function as fatigue driving differentiation；The average speed of pan
Standard deviation is spent in 30s, and F is maximum, and therefore, optimal time window is 30s.
9) time window is 30s and 60s, and single factor test side is carried out to driver's pupil diameter coefficient of variation under different fatigue state
Difference is analysed, as a result as shown in the following table 10.
The results of analysis of variance of the 10 pupil diameter coefficient of variation of table
When time window is 30s, F is maximum, and therefore, pupil diameter coefficient of variation optimal time window is 30s.
10) it selects Bonferroni to examine and is used as Multiple range test method, obtain the Multiple range test of fatigue driving characteristic parameter
As a result as shown in table 11.
The result of multiple comparisons (α=0.05) of 11 driving fatigue characteristic parameter of table
It can be seen that in the case of given level of signifiance α=0.05, there is system between different fatigue level from upper table
Meter meaning on significant difference characteristic parameter have steering wheel angle standard deviation SASTD, steering wheel angular velocity absolute mean SWM,
Acceleration standard deviation Astd, pan average speed standard deviation SACV_std, pupil diameter coefficient of variation CVPLD；
There are conspicuousnesses between awake and very tired and between tired and very tired by steering wheel absolute mean SAM
The identification of difference, seven pairs of fatigue drivings still has certain contribution, therefore, the characteristic parameter that can be recognized as fatigue driving.
4, bivariate Spearman correlation analyses are carried out to the characteristic parameter that preliminary screening goes out, be as follows：
4-1) source and the characteristic parameter of same index are analyzed, the significant feature of correlation can only select one；
4-2) to the characteristic parameter filtered out, correlation analysis is carried out by driving behavior parameter and eye movement parametric classification, together
Sample can only select one to the significant index of correlation；
4-3) to the remaining index progress correlation analysis after above two screenings, the significant index of correlation is adopted
Take step 4-1), 4-2) same processing mode；
5, filtering out optimal characteristics parameter includes：Steering wheel angle standard deviation SASTD, vehicle acceleration standard deviation Astd,
Sweep average speed standard deviation SACV_std and pupil diameter coefficient of variation CVPLD；Optimal characteristics parameter set is as shown in table 12 below.
12 optimal characteristics parameter set of table
| Characteristic parameter | Meaning | Optimal time window |
| CVPLD | The pupil diameter coefficient of variation | 30s |
| SACV_std | Sweep average speed standard deviation | 30s |
| Astd | Vehicle acceleration standard deviation | 25s |
| SASTD | Steering wheel angle standard deviation | 10s |
6, fatigue driving BP neural network identification model is established, it is first determined the input neuron number of the network structure, defeated
Go out neuron number, the implicit number of plies, hidden layer neuron number；The transmission function of next network layer, the power of each layer initial value of setting
Value finally determines best learning efficiency, chooses anticipation error appropriate, specific as follows：
6-1) input neuron, select optimal characteristics parameter and the input as model in step 5), and by input data into
Row normalized；
6-2) output neuron, setting output are that 1- is awake, 2- is tired, 3- is very tired, and is indicated by a neuron,
It is boundary with 1.5 and 2.5 in the prediction result of network, is awake if predicted value 1≤k≤1.5；If predicted value 1.5
≤ k≤2.5 are very tired if being then fatigue predicted value 2.5≤k≤3；
6-3) imply the number of plies, single hidden layer, node in hidden layer 9；
6-4) transmission function, using tansig as transmission function, output layer uses purelin for the neural network inside of BP
As transmission function；
6-5) initial weight, initial weight are set to the non-zero random value between (- 1,1)；
6-6) learning efficiency and anticipation error, using LM learning algorithms, learning efficiency is automatically adjusted according to anticipation error, nothing
It needs to set；Anticipation error is set as 0.001；
7, it using Matlab software programming model programs, randomly selects training sample and test sample set pair model is instructed
Practice and fatigue state recognizes.
150 groups of sample datas are randomly selected from the fatigue driving sample database built, wherein regaining consciousness, being tired, very
Each 50 groups of tired sample is used as training sample, and training sample data importing model is trained.
The correct recognition rata of verification for model validity, model can be used as assessment indicator, and calculation formula is：
Wherein, n indicates fatigue state, and 1- is awake, 2- fatigues, and 3- is very tired；
PnIndicate n-th grade of tired correct recognition rata of model pair；
qnIt indicates in n-th grade of testing fatigue sample, the sample number being correctly validated；
QnIndicate n-th grade of testing fatigue total number of samples.
As another judging quota of model validity, calculation formula is as follows：
N=1,2,3；K=1,2,3；n≠k
Wherein, n, k indicate fatigue state, and 1- is awake, 2- fatigues, and 3- is very tired；
PnkIndicate that n-th grade of fatigue of model pair is identified as the correct recognition rata of k grades of fatigue；
rnkIt indicates in n-th grade of testing fatigue sample, is mistaken for the number of samples of k grades of fatigue states.
8 drivers are recruited, above-mentioned identical experiment is carried out, records their driving behavior data and eye movement data, everyone
Awake, tired, the very tired each 30 groups of sample datas of extraction are used to test trained identification model, test result
Accuracy rate and False Rate are as shown in table 13.
Testing result of the 13 fatigue driving identification model of table to different drivers
The result shows that the model has arrived separately at 83.3% to awake, tired, very tired three kinds of states of driver,
69.6% and 79.6% average accuracy of identification can be used for the fatigue state detection of driver.
The driver that the present invention selects has certain limitation, does not completely eliminate the individual in tired rack detection
Difference, if expanding the sample size of fatigue driving sample database, accuracy of identification returns the raising for having bigger, can also model be made to have
There is extensive adaptability.
Claims (2)
1. the fatigue driving discrimination method based on driving behavior and eye movement characteristics, which is characterized in that include the following steps：
1) fatigue driving simulated experiment is designed, driving behavior data of the driver under different fatigue state are acquired by experiment,
Including steering wheel angle, tarnsition velocity, speed, vehicle acceleration and the eye movement data of driver, including blinks, watches attentively, sweeps
Data；
2) cutting screening is carried out to experimental data, establishes fatigue driving sample database；
3) use the method for one-way analysis of variance to the driving behavior parameter and eye movement characteristics of driver under different fatigue state
The conspicuousness of parameter differences is for statistical analysis, and has found the optimal time window of various characteristic parameters；Preliminary screening goes out to drive
Cybernetics control number includes steering wheel absolute mean SAM, corner standard deviation SASTD, steering wheel angular velocity absolute mean SWM, angle
Velocity standard difference SWSTD, velocity standard difference Vstd, acceleration mean value Am and standard deviation Astd；Eye movement characteristics parameter includes blink
Frequency BF, duration of fixation mean value FIXT_mean and standard deviation FIXT_std, pan average speed standard deviation SACV_std and
Pupil diameter coefficient of variation CVPLD；
4) bivariate Spearman correlation analyses are carried out to the characteristic parameter that preliminary screening goes out, be as follows：
4-1) to being analyzed from the characteristic parameter of same index, the significant feature of correlation can only select one；
4-2) to the characteristic parameter filtered out, correlation analysis is carried out by driving behavior parameter and eye movement parametric classification, it is equally, right
The significant index of correlation can only select one；
4-3) to the remaining index progress correlation analysis after the screening of above two step, the significant index of correlation is taken
Step 4-1), 4-2) same processing mode；
5) filtering out optimal characteristics parameter includes：Steering wheel angle standard deviation SASTD, vehicle acceleration standard deviation Astd, pan
Average speed standard deviation SACV_std and pupil diameter coefficient of variation CVPLD；
6) fatigue driving BP neural network identification model is established, it is first determined the input neuron number of the network structure, output god
Through first number, the implicit number of plies, hidden layer neuron number；Secondly the transmission function of each network layer, the power of each layer initial value of setting are determined
Value finally determines best learning efficiency, chooses anticipation error appropriate, specific as follows：
Neuron 6-1) is inputted, the optimal characteristics parameter filtered out in step 5) is selected：Steering wheel angle standard deviation SASTD, vehicle
Acceleration standard deviation Astd, pan average speed standard deviation SACV_std and pupil diameter coefficient of variation CVPLD are as model
Input, and input data is normalized；
6-2) output neuron, setting output is that 1- is awake, 2- is tired, 3- is very tired, and is indicated by a neuron, in net
It is boundary with 1.5 and 2.5 in the prediction result of network, is awake if predicted value 1≤k≤1.5；If 1.5 ＜ k ＜ of predicted value
2.5, it is very tired if being then fatigue predicted value 2.5≤k≤3；
6-3) imply the number of plies, single hidden layer, node in hidden layer 9；
6-4) transmission function, using tansig as transmission function, output layer uses purelin conducts for the neural network inside of BP
Transmission function；
6-5) initial weight, initial weight are set to the non-zero random value between (- 1,1)；
6-6) learning efficiency and anticipation error, using LM learning algorithms, learning efficiency is automatically adjusted according to anticipation error, without setting
It is fixed；Anticipation error is set as 0.001；
7) Matlab software programming model programs are utilized, randomly select training sample and test sample set pair model be trained and
Fatigue state recognizes.
2. the fatigue driving discrimination method according to claim 1 based on driving behavior and eye movement characteristics, which is characterized in that
The fatigue driving simulated experiment is specific as follows：
1) experimental facilities：
A, driving analog system, by emulation vehicle, control taiwan area, visual system, computer system, network system and corresponding
Software systems form；Primary operational equipment is all real vehicles equipment, the input action of user is captured by sensor, by user
Input carry out analog simulation calculating, while simulation result shown by visual system, sound simulation system and signal it is defeated
Go out system and corresponding informance is fed back into user；System further includes that the export of this vehicle related data in emulation vehicle operation connects
Mouthful, can different sample frequencys be set according to Research Requirements, export data carry out secondary analysis；
B, II type eye trackers of Eye Link, are tracked using pupil pattern；
2) experiment scene：Using dull beltway as experiment scene, the dullness beltway is two-way 6 vehicle of an annular
The suburb dullness high-grade highway in road, lane width 3.75m, road overall length 20km, evenness of road surface, weather are set as fine day；
3) experimenter：Health, rule of sleeping is interior for 24 hours before test not allow to drink, and does not allow to drink in 12h before testing
Coffee, tea；
4) experimental program：The speed of 100km/h is kept to be travelled in middle lane, staff primary subject personnel of record per 5min
Subjective fatigue grade, wherein level of fatigue by sleep scale table Karolinska sleepiness scale, KSS demarcate,
Until travelling 60min, experiment terminates.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610869724.9A CN106251583B (en) | 2016-09-30 | 2016-09-30 | Fatigue driving discrimination method based on driving behavior and eye movement characteristics |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610869724.9A CN106251583B (en) | 2016-09-30 | 2016-09-30 | Fatigue driving discrimination method based on driving behavior and eye movement characteristics |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106251583A CN106251583A (en) | 2016-12-21 |
| CN106251583B true CN106251583B (en) | 2018-09-25 |
Family
ID=57612122
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610869724.9A Active CN106251583B (en) | 2016-09-30 | 2016-09-30 | Fatigue driving discrimination method based on driving behavior and eye movement characteristics |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106251583B (en) |
Families Citing this family (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106611169B (en) * | 2016-12-31 | 2018-10-23 | 中国科学技术大学 | A kind of dangerous driving behavior real-time detection method based on deep learning |
| CN107301497A (en) * | 2017-05-27 | 2017-10-27 | 西南交通大学 | The method and system of working memory ability evaluation |
| CN107239895A (en) * | 2017-05-27 | 2017-10-10 | 西南交通大学 | The method and system examined for Sustained attention power |
| CN107316436B (en) * | 2017-07-31 | 2021-06-18 | 努比亚技术有限公司 | Dangerous driving state processing method, electronic device and storage medium |
| CN109901446B (en) * | 2017-12-08 | 2020-07-07 | 广州汽车集团股份有限公司 | Intersection passage control method, device and system |
| CN108312964A (en) * | 2018-03-21 | 2018-07-24 | 华南理工大学 | A kind of fatigue driving state identification early warning system and method |
| CN108711016A (en) * | 2018-06-05 | 2018-10-26 | 合肥湛达智能科技有限公司 | A kind of driving behavior methods of marking based on BP neural network |
| CN111292548B (en) * | 2020-02-06 | 2021-02-05 | 温州大学 | Safe driving method based on visual attention |
| CN111353636A (en) * | 2020-02-24 | 2020-06-30 | 交通运输部水运科学研究所 | Multi-mode data based ship driving behavior prediction method and system |
| CN111932829B (en) * | 2020-08-14 | 2022-04-05 | 北京智华通科技有限公司 | Fatigue driving prevention facility utility testing method and system |
| CN112052905B (en) * | 2020-09-11 | 2023-02-03 | 重庆科技学院 | Method for extracting multi-operation fatigue features of driver based on recurrent neural network |
| CN112750533A (en) * | 2020-11-03 | 2021-05-04 | 中国人民解放军63919部队 | Exoskeleton assistance effect evaluation method based on neural network |
| CN114170588B (en) * | 2021-12-13 | 2023-09-12 | 西南交通大学 | Eye feature-based bad state identification method for railway dispatcher |
| CN114424941A (en) * | 2022-01-26 | 2022-05-03 | 广东电网有限责任公司 | Fatigue detection model construction method, fatigue detection method, device and equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101872171A (en) * | 2009-04-24 | 2010-10-27 | 中国农业大学 | Driver fatigue state recognition method and system based on information fusion |
| CN104207791A (en) * | 2014-08-26 | 2014-12-17 | 江南大学 | Fatigue driving detection method |
| CN204332007U (en) * | 2015-01-12 | 2015-05-13 | 江苏省交通规划设计院股份有限公司 | A kind of driving behavior early warning car-mounted terminal based on information fusion |
| CN105719431A (en) * | 2016-03-09 | 2016-06-29 | 深圳市中天安驰有限责任公司 | Fatigue driving detection system |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7839292B2 (en) * | 2007-04-11 | 2010-11-23 | Nec Laboratories America, Inc. | Real-time driving danger level prediction |
-
2016
- 2016-09-30 CN CN201610869724.9A patent/CN106251583B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101872171A (en) * | 2009-04-24 | 2010-10-27 | 中国农业大学 | Driver fatigue state recognition method and system based on information fusion |
| CN104207791A (en) * | 2014-08-26 | 2014-12-17 | 江南大学 | Fatigue driving detection method |
| CN204332007U (en) * | 2015-01-12 | 2015-05-13 | 江苏省交通规划设计院股份有限公司 | A kind of driving behavior early warning car-mounted terminal based on information fusion |
| CN105719431A (en) * | 2016-03-09 | 2016-06-29 | 深圳市中天安驰有限责任公司 | Fatigue driving detection system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106251583A (en) | 2016-12-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106251583B (en) | Fatigue driving discrimination method based on driving behavior and eye movement characteristics | |
| CN107092881B (en) | A kind of human pilot changes system and method | |
| Chen et al. | A graphical modeling method for individual driving behavior and its application in driving safety analysis using GPS data | |
| Daza et al. | Drowsiness monitoring based on driver and driving data fusion | |
| Rakha et al. | Characterizing driver behavior on signalized intersection approaches at the onset of a yellow-phase trigger | |
| CN106652445B (en) | A kind of road traffic accident method of discrimination and device | |
| Zhenhai et al. | Driver drowsiness detection based on time series analysis of steering wheel angular velocity | |
| CN103530540B (en) | User identity attribute detection method based on man-machine interaction behavior characteristics | |
| García et al. | Vision-based drowsiness detector for a realistic driving simulator | |
| Das et al. | Differentiating alcohol-induced driving behavior using steering wheel signals | |
| CN113743471B (en) | Driving evaluation method and system | |
| CN113753059B (en) | Method for predicting takeover capacity of driver under automatic driving system | |
| CN108803623B (en) | Method for personalized driving of automatic driving vehicle and system for legalization of driving | |
| CN108171361A (en) | Consider the Traffic Flow Simulation Models scaling method of traffic conflict index distribution problem | |
| Ābele et al. | Consistency between subjectively and objectively measured hazard perception skills among young male drivers | |
| CN110139231A (en) | A kind of car networking social contact method based on travelling data | |
| Koh et al. | Smartphone-based modeling and detection of aggressiveness reactions in senior drivers | |
| CN106408032A (en) | Fatigue driving detection method based on corner of steering wheel | |
| CN113521723B (en) | Cognitive function evaluation system and method based on virtual reality | |
| CN103268426A (en) | Method for evaluating safe driving level of automobile driver | |
| CN106446849A (en) | Fatigue driving detection method | |
| CN111657923A (en) | Method and system for testing danger perception capability of driver | |
| CN115904091A (en) | Method and system for determining front distance of warning sign in operation area | |
| CN108537161A (en) | A kind of driving of view-based access control model characteristic is divert one's attention detection method | |
| CN109649399A (en) | A kind of driver distraction's method of discrimination based on lateral running speed |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20180816 Address after: 214000 canal road 555-3908, Wuxi, Jiangsu Applicant after: SOL ELECTRONICS TECHNOLOGY CO., LTD. Address before: 538000 No. 2, Qiao Hing Street, Fangcheng District, Fangchenggang, the Guangxi Zhuang Autonomous Region. Applicant before: FANGCHENGGANG PORT DISTRICT GAOCHUANG INFORMATION TECHNOLOGY CO., LTD. |
|
| TA01 | Transfer of patent application right | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |