Head movement-based driver drowsiness detection:

A review of state-of-art techniques

Ajay Mittal Kanika Kumar, Sarina Dhamija, Manvjeet Kaur
Department of Computer Science and Engineering, Department of Computer Science and Engineering,
UIET, Panjab University, PEC University of Technology,
Chandigarh, INDIA Chandigarh, INDIA
Email: [email protected] Email: {kanika.277, sarinadhamija18}@gmail.com, [email protected]

Abstract—Driver fatigue is one of the most common reasons for

deadly road accidents around the world. Continuous monotonous
driving for long hours without rest causes drowsiness and
consequently fatal road accidents. Automatic driver drowsiness
detection can prevent a vast number of sleep persuaded road ac-
cidents, and hence can save precious lives. Number of techniques
for driver drowsiness detection has been examined in the recent
past. This paper presents a survey of these techniques. These
techniques detect the driver drowsiness by observing the driving
pattern. Abnormalities in driving pattern is hypothesized as a
drowsiness state of driver. Various measures such as subjective,
behavioral, physiological, and vehicular have been used for this
purpose. The comparative analysis of these techniques indicates Fig. 1. Road Accidents in India in the year 2013
that behavioral measures are easy to acquire and does not disturb
the driver as they are non-invasive. Among various behavioral
measures, head movement measure is found to be most precise from the line of collision, or braking. The fact is corroborated
and effective. by Figure 2, which indicates that the majority of road accidents
Index Terms—Vehicular Safety; Accident Prevention; Drowsi- occurred due to driver’s drowsiness are fatal.
ness detection; Head movement-based techniques

I. I NTRODUCTION
Road network is the backbone for economic growth in any
country. India has second largest road network in the world,
which spans more than 4, 689, 842 kilometers in length [1].
Better road infrastructure has improved connectivity between
the cities and has facilitated mobility. This ease has in-
turn increased the vehicular traffic and the number of road
accidents. According to a statistical report [2], in India, a total
of 4, 86, 476 road accidents took place in the year 2013 in
which 1, 37, 572 people lost their lives. The critical analysis
of factors responsible for these accidents reveals that driver’s
fault is the most important factor which accounts for 78% of
the total road accidents. The statistics of Ministry of Transport, Fig. 2. Severity of accidents caused due to drowsiness as percentage of total
Government of India (GoI), indicates that driver fatigue is a road accidents reported in the years 2012-2014.
major cause behind driver’s fault. Other factors responsible for
road accidents and their percentage share is depicted in Figure Thus, to reduce the number of road accidents, to make
1. driving safe and to save precious lives, a real-time non-
Driving with fatigue on long monotonous routes, where not intrusive day-night driver drowsiness detection system is de-
much driving input is required, generally leads to drowsiness. sired which can alarm the driver, and if required can take
Drowsiness can be described as a state of reduce alertness the control of vehicle to avoid accident. One such earliest
along with a tendency to fall asleep. Its occurrence during driver assistant system based on pro-prioceptive sensor with
driving generally leads to severe accidents since driver is Anti-locking Braking System (ABS) was developed Bosch
unable to take avoiding action such as steering off the vehicle in 1978 [3]. Since then, the advancements in technology
have facilitated the availability of fast on-board processors,
978-1-4673-9916-6/16/$31.00 c 2016 IEEE cheaper memories, day-night vision cameras, sensors such
2nd IEEE International Conference on Engineering and Technology(ICETECH), 17th − 18th March 2016, Coimbatore, India.

Author Year KSS reading Comparative Signal

as accelerometer, gyroscope, electro-mechanical actuators and interval
has exponentially increased the research in the field of Intel- Hu et al.[11] 2009 5 minutes EoG
ligent Transportation System (ITS). A number of techniques Portouli et al.[12] 2007 5 minutes EEG
Sommer et al.[13] 2010 2 minutes Variation of Lane Posi-
for drowsiness detection has been proposed in the recent past tion (VLP)
[4-6]. A survey of these techniques is presented in Section II. Ingre et al.[14] 2006 5 minutes Eye Blink Duration
Section III exclusively focuses on state-of-art techniques for (EBD)
driver drowsiness detection based on head movement analysis. TABLE II
D RIVER D ROWSINESS D ETECTION S YSTEMS USING KSS
II. D RIVER D ROWSINESS D ETECTION S YSTEMS
Driver drowsiness detection systems (DDDS) continuously
scans for various parameters/ features of vehicle [7], driver Hence, the measure is subjective and cannot give precise
[8] or the surrounding environment to detect abnormality in readings like physiological or behavioral measures.
the pattern of driving [8], [9]. On the basis of entity under
observation, the DDDS are classified as shown in Figure 3. B. Behavioural measures
Behavioral measures detects the driver drowsiness by ob-
serving features such as eye state, eye blinking rate, head
movement and yawning. The general process followed in the
behavioral measure based system is shown in the Figure 4.

Fig. 3. Classification of Driver Drowsiness Detection Systems.

Another commonly prevalent classification of DDDS is on

the basis of measure adopted to estimate driver drowsiness[8].
The measures commonly used to estimate driver drowsiness
are: (i) Subjective Measures, (ii) Behavioural Measures, (iii)
Physiological Measures, and (iv) Vehicular Measure.
A. Subjective Measures
Subjective measures rely on the driver’s feedback to evalu-
ate the level of drowsiness. Karolinska Sleepiness Scale (KSS)
[10] is a nine point scale that uses verbal descriptions to
measure the drowsiness level. Various verbal descriptions used
in KSS are listed in Table I.

Scale Verbal Description

1 Extremely Alert
2 Very Alert
3 Alert
4 Fairly Alert
5 Neither Alert nor Sleepy
6 Some signs of sleepiness
7 Sleepy, but no effort to keep alert
8 Sleepy, some effort to keep alert
9 Very Sleepy, great effort to keep alert
TABLE I Fig. 4. General process followed by behavioral measure based systems
K AROLINSKA S LEEPINESS S CALE (KSS) [8]

The representative methods that use KSS measures for de- The behavioral measure based systems observe and determine
tection driver drowsiness are listed in Table II. Subjective the abnormal patterns in driving on the basis of the following
measures perform well in simulated environments but does not features [15]:
give accurate results in real-time environments. Also, frequent 1. Eye state analysis: Eye state is an important feature that can
questioning alerts the driver and alters the drowsiness level. help in driver drowsiness. The methods based on this feature

2
2nd IEEE International Conference on Engineering and Technology(ICETECH), 17th − 18th March 2016, Coimbatore, India.

Author Year Metric Visual Features Methods Classifiers Accuracy

G.M. Bhandari et 2014 Yawn frequency Yawn YCbCr Color Space, Canny edge Haar 80
al.[16]
M. Saradadevi, P. Ba- 2008 Mouth and Yawning Yawn Viola-Jones Algorithm, Adaboost, RDF SVM 81
jaj[17] Kernel
C. Sun et al.[18] 2014 PERCLOS Eye state Viola-Jones Algorithm, Adaboost, PERC- SVM, 80
LOS LBP
J. Jo et al.[19] 2014 PERCLOS Eye state AdaBoost, PCA and LDA, PERCLOS SVM 95
M. Chutorian et al.[6] 2010 Head pose, Lip corner, Head Movement AdaBoost, LGO Histogram SVR N/A∗
Eye state
A.Tawari et al.[20] 2014 Facial features Head movement Constrained local model and pictorial struc- N/A∗ 89-96
tural matching
M. Lew et al.[21] 2007 Blink rate, Yawn mo- Facial Action Gabor Filter, Wavelet decomposition SVM 96
tion, Head motion
TABLE III
D RIVER D ROWSINESS D ETECTION S YSTEMS BASED ON BEHAVIORAL MEASURES .

Author Year Sensor Methods Classifiers Accuracy

R.N.Khushaba et 2011 EEG, ECG, EoG Optimal Wavelet Packet, Fuzzy Wavelet Packet LDA,SVM,KNN 95-97
al.[22]
Vicente et al.[23] 2016 Heart Rate Variability Estimated time-varying mean HR signal,SPWVD of LDA N/A∗
the modulating signal
Sampei et al.[24] 2015 Pupil movement, Wearable eye detection system,dye-sensitized photo- N/A∗ N/A∗
Blinking, EEG voltaic cells
Li, Ming-ai et al.[5] 2010 EEG Power spectrum analysis, FastICA
Yeo, Mervyn VM.et 2008 EEG EEG signal pattern recognition, FFT with Hann win- SVM 99.3
al.[25] dow
C.L. Fu et al.[26] 2012 EEG Independent component analysis decomposition, FFT Self-organizing 96.7
Neural-Fuzzy
inference network

TABLE IV
D RIVER D ROWSINESS D ETECTION S YSTEMS BASED ON PHYSIOLOGICAL MEASURES .

observes the eye states. The metric used to detect drowsiness Infra-red (IR) cameras have also been developed [21]. These
is PERCLOS. PERCLOS is percentage of eye closure over a system can perform well both during day and night. The major
period of time. The classification is done as open eye or closed advantage of the systems based on behaviorial analysis is their
eye based on value of PERCLOS. non-invasiveness. These systems in no way disturb or irritate
2. Eye blinking analysis: The methods based on eye blinking the driver.
analysis uses blink frequency as a measure to detect driver
drowsiness. Fatigue decreases the eye blink frequency, i.e., C. Physiological Measures
the eyelid closure duration increases.
3. Yawning Analysis: Fatigue and boredom leads to yawning, Physiological measures are invasive method for detecting
which is described as uncontrolled wide opening of mouth drowsiness and requires the electrodes to be in contact with
to increase the oxygen supply in the body. Due to wide driver’s body. During fatigue, the physiological indexes of
opening of mouth during yawning the geometric shape of the driver gets changed and can be observed to determine
mouth changes, which can be observed and can be used for drowsiness level. The general process followed by driver
driver drowsiness detection. The systems based on this feature drowsiness detection system based on physiological measures
uses mouth shape, position of lip corners and degree of mouth is shown in Figure 5. Some of the physiological signals that
openness to determine yawning or driver drowsiness. are used as efficient drowsiness indicators are:
4. Facial action analysis: A mix of more than one facial 1. Electroencephalogram (EEG): EEG is a very common
feature can also be used to detect level of drowsiness in driver. and efficient measure to detect the fatigue. EEG sensors are
The facial feature set includes blink, yawning, nose wrinkle, attached to the driver’s head record brain activity over a period
lip corner puller. As more than one facial feature is used in of time with the intend to detect the alertness level of driver.
detection of drowsiness, artificial neural network and SVM 2. Electrocardiogram (ECG): ECG is the most convenient
can be used to optimize the detection. and effective measure of detecting drowsiness. ECG use the
electrodes placed on drivers body to record the variation in
Various driver drowsiness detection systems based on be- heart rate over period of time. These reading are used to detect
havioral measures are listed in Table III. The main limitation driver drowsiness.
of the systems based on visual feature and behavioral measures 3. Electrooculogram (EoG): EoG is the most accurate measure
is their inability to perform under poor lighting conditions.
Some of the lightening invariant systems that make use of ∗ N/A: Not Applicable/Not Reported

3
2nd IEEE International Conference on Engineering and Technology(ICETECH), 17th − 18th March 2016, Coimbatore, India.

Author Year Parameter Methods Accuracy

Chen, Zhijun et al.[7] 2015 Yaw rate,Lateral Acceleration, Longitudinal Accel- Wilcoxon rank sum test N/A∗
eration
P.M. Forsman et 2013 Steering wheel variability, Lateral lane position vari- PCA, Psychomotor vigilance test N/A∗
al.[27] ability
J. Krajewski et al.[28] 2008 Steering wheel behavior SVM,KNN, Time domain features,Frequency 86.1
domain features
Morris et al.[9] 2015 Lane deviation, Vehicle heading difference Psychomotor vigilance task, Oculomoter fa- N/A∗
tigue measure
Tansakul et al.[29] 2016 Blink rate, Driving behavior PCA, Haar cascade classifier, PERCLOS, 89
Template matching

TABLE V
D RIVER D ROWSINESS D ETECTION S YSTEMS BASED ON VEHICULAR MEASURES .

Fig. 5. General process followed by physiological measures based systems.

to detect drowsiness. EoG are the electrical signals produced

by electro-physiology electrodes attached around the eyes of
the driver which record the eye movements.
Various driver drowsiness detection systems based on phys-
iological measures are listed in Table IV. Prolonged used of Fig. 6. General process followed by vehicular measures based systems.
the systems based on this measure can be irritate the driver.

D. Vehicular Measures
Abnormality in the driving pattern can be detected using 1. Steering wheel movement: Steering wheel movements are
parameters such as steering wheel movement, lane departure, measured by steering angle sensors which are placed on
acceleration pedal movement, braking, etc. The general pro- steering column. Different metrics of steering wheel movement
cess adopted by systems based on these measures is shown in include standard deviation of steering wheel angle, steering
Figure 6. wheel velocity, steering wheel action rate, high frequency
The commonly used vehicular measures that have been used component of steering wheel angle, etc.
in DDDS are: 2. Lane departure: An external camera is used to monitor
vehicle’s position with respect to the center lane of the road.
∗ N/A: Not Applicable/Not Reported Lane departure is dependent on the lane marking on the road,

4
2nd IEEE International Conference on Engineering and Technology(ICETECH), 17th − 18th March 2016, Coimbatore, India.

Author Year Parameters Methods-Classifiers Advantages Limitations

A. Tawari and 2014 Head pose esti- Random forest classier is used Predicts uneven eye gaze direc- It does not disambiguate con-
M. M. Trivedi mation to differentiate b/w dynamic tion using head pose cues fusion between adjacent gaze
[30] and static feature extraction zones
A. Tawari et 2014 Facial features Constrained local model Can also be adapted with mul- Need to improve computational
al.[20] (CLM) and the pictorial tiple participants and multiple cost and more head pose error
structure matching sensory cues
Martin et al.[31] 2013 Facial features Constrained local model (CLM) Deals with the issues like spa- Multi-perspective framework
is used for feature extraction tially large head turns, light- should be improved
ing conditions, and yaw rota-
tion angle
Xianping Fu et 2013 Facial location, 3-D face model for tting and Without need of the drivers co- It does not work in different
al.[26] Face feature, and tracking operation in a calibration pro- lighting conditions.
Face size cessit demonstrates robust esti-
mates of the drivers gaze direc-
tion
E. M. Chutoriam 2010 Eye state, Head Face detector - Three cas- Problem of varying lighting Single camera perspective is
et al.[7] Pose, Lip corners caded Adaboost, LGO his- conditions in a moving car is used
togram, SVRs solved
E. M. Chutoriam 2007 Head Pose Three cascaded Adaboost, LGO Tolerate deviations caused by It uses sensor and single camera
et al.[32] histogram, SVRs scale, position, rotation, and video
lighting
Qiong Wang et 2006 Eye state, Mouth PERCLOS, SVM Provide effective methods to More research and development
al. [33] shape, Head posi- improve driving safety is needed for effective fatigue
tion monitoring system
TABLE VI
H EAD MOVEMENT BASED DRIVER DROWSINESS DETECTION TECHNIQUES .

and thus is not always effective. Moreover, this measure is movements. It is based on a single camera which operates
heavily dependent on lighting and weather conditions. the instrument board, the side mirrors, the rear-view mirror,
Various DDDS based on vehicular measures are listed in and different zones in the windshield as calibration points.
Table V. E. Murphy-Chutorian and M. Trivedi [6] presents a system
which detects drivers head, provide initial estimates of the
III. H EAD MOVEMENT- BASED DROWSINESS DETECTION heads pose, and constantly track its position and orientation
TECHNIQUES
in six degrees of freedom. To build a system three cascaded
Head pose is a robust indicator of driver’s field of view Adaboost face detectors are used to grayscale video images,
and current focus of attention. A number of methods for LGO histogram have been used to normalize the image, three
monitoring fatigue and attention of driver using head dynamics support vector regressors (SVRs) are trained for head pitch,
have been proposed in the recent past. A consolidated list of yaw, and roll. E. Murphy-Chutorian et al. [32] have also
head pose based methods is presented in Table VI. proposed a head pose estimation system which works online in
Ashish Tawari and Mohan M. Trivedi [30] worked on large daylight and nighttime lighting conditions. Localized Gradient
head movements by presenting distributed camera framework Orientation (LGO) histograms with support vector machines
for gaze zone estimation using head pose dynamics. Several for regression have been used in their approach to tolerate
experiments are conducted on naturalistic road driving dataset deviations produced by scale, position, rotation, and lighting.
to design dynamic feature. The performance of the proposed Qiong Wang et al. [33] presented a comprehensive review
method has been compared with the performance against static on driver fatigue detection and provides different structural
head pose based approach and has found to better. Tawari categorization of the methods. E. Murphy-Chutorian and M.
et al. [20] proposes system which tracks facial features and Trivedi [34] discussed the intrinsic difficulties in head pose
analyze its geometric conguration to estimate the head pose estimation and presented a precise survey on the progress in
using a 3-D model. They have also proposed two solutions the field of driver assistant systems. They have also discussed
that improve the accuracy of tracking and reduces the com- the benefits and drawbacks of each approach. Comparison of
putation time. The comparative analysis with different camera systems have been done by focusing on their ability to estimate
configuration has been performed. Martin et al. [31] presents head pose, highlighting approaches that are well matched in
multi-perspective framework based on a shape feature to unrestricted environments.
continuously monitor the drivers head dynamics. The proposed
approach uses distributed camera setup to monitor drivers IV. C ONCLUSION
head movements. A camera selection procedure is used to A survey various state-of-art techniques for driver drowsi-
select the suitable perspective based on head dynamics and ness detection has been presented in the paper. The funda-
a condence measure based on symmetry of facial features. mental working principle of all the techniques is to detect
Xianping et al. [7] proposes an automatic calibration method abnormality in the driving pattern. In order to do so, these
which is supported by self-learning algorithm to track head techniques uses different measures such as (i) Subjective, (ii)

5
2nd IEEE International Conference on Engineering and Technology(ICETECH), 17th − 18th March 2016, Coimbatore, India.

Behavioral, (iii) Physiological, and (iv) Vehicular. Subjective [19] J. Jo, S. J. Lee, K. R. Park, I.-J. Kim, and J. Kim, Detecting driver
measures alerts the driver and does not give exact estimate. drowsiness using feature-level fusion and user-specific classification,
Expert Syst. Appl., vol. 41, no. 4, pp. 11391152, Mar. 2014.
Physiological measures uses invasive techniques and disturbs [20] Tawari, Ashish, Sujitha Martin, and Mohan Manubhai Trivedi. ”Contin-
the driver. Vehicular measure requires complex and costly uous head movement estimator for driver assistance: Issues, algorithms,
infrastructure. Behavioral measures have been found to be and on-road evaluations.” Intelligent Transportation Systems, IEEE Trans-
actions on15.2 (2014): 818-830.
the easiest and cost effective to detect driver drowsiness. [21] Lew, M.; Sebe, N.; Huang, T.; Bakker, E.; Vural, E.; Cetin, M.; Ercil,
Among various behavioral measures such as eye state, eye A.; Littlewort, G.; Bartlett, M.; Movellan, J., ”Drowsy driver detection
blink rate, yawning and head movement, the head movement through facial movement analysis”, In Human-Computer Interaction;
Springer: Berlin, Germany, 2007; Volume 4796, pp. 618.
based techniques have found to be more precise in detecting [22] Khushaba, R.N.; Kodagoda, S.; Lal, S.; Dissanayake, G. Driver drowsi-
driver drowsiness. ness classification using fuzzy wavelet-packet-based feature-extraction
algorithm. IEEE Trans. Biomed. Eng 2011, 58, 121131.
R EFERENCES [23] Vicente, Jos, et al. ”Drowsiness detection using heart rate variability”,
Medical and biological engineering and computing (2016): 1-11.
[1] ”National Highways Development Project: An Overview (GoI)”, pp. 1-2, [24] Sampei, Kota, et al. ”Mental Fatigue Monitoring Using a Wearable
Accessed [20 Feb 2016] Transparent Eye Detection System” Micromachines 7.2 (2016): 20.
[2] ”Road Accidents in 2013”, Available [25] Yeo, Mervyn VM, et al. ”Can SVM be used for automatic EEG detection
”http://www.morth.nic.in/showfile.asp?lid=1465” Accessed [20 Feb of drowsiness during car driving?” Safety Science 47.1 (2009): 115-124.
2016] [26] Fu, C.L.; Li, W.K.; Chun, H.C.; Tung, P.S.; Chin, T.L. Generalized EEG-
[3] Bengler, K. and Dietmayer, K. and Farber, B. and Maurer, M. and Stiller, based drowsiness prediction system by using a self-organizing neural
C. and Winner, H., ”Three Decades of Driver Assistance Systems: Review fuzzy system. IEEE Trans. Circ. Syst 2012, 59, 20442055.
and Future Perspectives”, Intelligent Transportation Systems Magazine, [27] P. M. Forsman, B. J. Vila, R. A. Short, C. G. Mottd, and H. P. A.
Vol. 6, No. 4, pp. 6-22, 2014 Van Dongen, Efficient driver drowsiness detection at moderate levels of
[4] Wang, Xuesong, and Chuan Xu, ”Driver drowsiness detection based on drowsiness, Accident Anal. Prevent., vol. 50, pp. 341350, Jan. 2013.
non-intrusive metrics considering individual specifics”, Accident Analysis [28] J. Krajewski, D. Sommer, U. Trutschel, D. Edwards, and M. Golz,
and Prevention (2015) Steering wheel behavior based estimation of fatigue, in Proc. 5th Int.
[5] Li, Ming-ai, Cheng Zhang, and Jin-Fu Yang, ”An EEG-based method for Driving Symp. Human Factors Driver Assessment, Train. Veh. Des., 2008,
detecting drowsy driving state”, 2010 Seventh International Conference pp. 118124.
on Fuzzy Systems and Knowledge Discovery (FSKD), Volume 5, 2010. [29] Tansakul, Wasan, and Poj Tangamchit. ”Fatigue Driver Detection System
[6] Murphy-Chutorian, Erik, and Mohan Manubhai Trivedi, ”Head pose Using a Combination of Blinking Rate and Driving Inactivity.” Journal
estimation and augmented reality tracking: An integrated system and eval- of Automation and Control Engineering Vol 4.1 (2016).
uation for monitoring driver awareness”, IEEE Transactions on Intelligent [30] Tawari, Ashish, and Mohan Manubhai Trivedi. ”Robust and continuous
Transportation Systems, pp. 300-311, 2010. estimation of driver gaze zone by dynamic analysis of multiple face
[7] Chen Z., Wu C., Zhong M., Lyu N., HUang Z., ”Identification of common videos.”Intelligent Vehicles Symposium Proceedings, 2014 IEEE. IEEE,
features of vehicle motion under drowsy/distracted driving: A case study 2014.
in Wuhan, China”, Accident Analysis and Prevention, Vol. 81, pp. 251- [31] Martin, Sebastien, Ashish Tawari, and Mohan Manubhai Trivedi. ”Mon-
259, 2015. itoring head dynamics for driver assistance systems: A multi-perspective
[8] Sahayadhas, Arun, Kenneth Sundaraj, and Murugappan Murugappan, approach.”Intelligent Transportation Systems-(ITSC), 2013 16th Interna-
”Detecting driver drowsiness based on sensors: a review”, Sensors, Vol tional IEEE Conference on. IEEE, 2013.
12, pp. 16937-16953, 2012. [32] E. Murphy-Chutorian and M. Trivedi, Head pose estimation for driver
[9] Morris, Drew M., June J. Pilcher, and Fred S. Switzer III, ”Lane heading assistance systems: A robust algorithm and experimental evaluation, in
difference: An innovative model for drowsy driving detection using Proc. IEEE Conf. Intelligent Transportation Systems, 2007, pp. 709 714.
retrospective analysis around curves”, Accident Analysis and Prevention, [33] Qiong Wang, Jingyu Yang, Mingwu Ren, Yujie Zheng Driver Fatigue
Vol. 80, pp. 117-124, 2015. Detection: A Survey, IEEE Proceedings of the 6th World Congress on
[10] Otmani, S.; Pebayle, T.; Roge, J.; Muzet, A., ”Effect of driving duration Intelligent Control and Automation, June 21 - 23, 2006, Dalian, China
and partial sleep deprivation on subsequent alertness and performance of [34] E. Murphy Chutorian and M. Trivedi, Head pose estimation in computer
car drivers”, Physiology and Behavior, Vol. 84, pp. 715-724, 2005 vision: A survey, IEEE Trans. Pattern Anal. Mach. Intell., vol. 31, no. 4,
[11] Hu, S.; Zheng, G. Driver drowsiness detection with eyelid related pp. 607626, Apr. 2009.
parameters by support vector machine. Exp. Syst. Appl. 2009, 36,
76517658.
[12] Portouli, E.; Bekiaris, E.; Papakostopoulos, V.; Maglaveras, N. On-
road experiment for collecting driving behavioural data of sleepy drivers.
Somnology 2007, 11, 259267.
[13] Sommer, D.; Golz, M.; Trutschel, U.; Edwards, D. Biosignal based
discrimination between slight and strong driver hypovigilance by support-
vector machines. In Agents and Artificial Intelligence; Springer: Berlin,
Germany, 2010; Volume 67, pp. 177187.
[14] Ingre, M.; AKerstedt, T.; Peters, B.; Anund, A.; Kecklund, G. Subjective
sleepiness, simulated driving performance and blink duration: Examining
individual differences. J. Sleep Res. 2006, 15, 4753.
[15] S. Kaplan, M. A. Guvensan, A. G. Yavuz and Y. Karalurt, ”Driver
Behavior Analysis for Safe Driving: A Survey”, IEEE Transactions on
Intelligent Transportation Systems, Vol. 16 (6), pp. 3017-3032, 2015.
[16] G. M. Bhandari, A. Durge, A. Bidwai, and U. Aware, Yawning analysis
for driver drowsiness detection, Int. J. Res. Eng. Technol., vol. 03, no. 2,
pp. 502505, Feb. 2014.
[17] M. Saradadevi and P. Bajaj, Driver fatigue detection using mouth and
yawning analysis, Int. J. Comput. Sci. Netw. Security, vol. 8,No. 6, pp.
183188, Jun. 2008.
[18] C. Sun, J. Li, Y. Song, and L. Jin, Real-time driver fatigue detection
based on eye state recognition, Appl. Mech. Mater., vol. 457-458, pp.
944952, 2014.