International Journal of Latest Technology in Engineering, Management & Applied Science (IJLTEMAS)

Volume IX, Issue XI, November 2020 | ISSN 2278-2540

A Survey on Driver Drowsiness Detection

Techniques
Reshma J1, Aishwarya B2, Farheen Khanam Z M3, G Sai Vennela4, Lekhana A V5
1
Associate Professor, Department of Computer Science & Engineering, B N M Institute of Technology, Bangalore, India
2,3,4,5
Student, Department of Computer Science & Engineering, B N M Institute of Technology, Bangalore, India

Abstract: The developments in technology over the years bring 4) Physiological measurement- Requires body sensors to
the support to drivers using smart vehicle systems. In past few measure parameters like brain activity, heart rate, skin
years, there has been substantial increase in road accidents in conductance, muscle activity [2]
India and worldwide as well. The most significant reasons for the
same are drowsiness and fatigue. Therefore, driver drowsiness According to WHO, road accidents are rapidly increasing.
and fatigue detection is major possible area to prevent a large Since the drivers are often tired due to environmental
number of sleep induced road accidents. Considering this conditions or mental stress, if preventive actions are not taken
problem, this article proposes different methods for Driver before an accident occurs then it would cause serious deaths.
Drowsiness Detection System applicable in motor vehicles. The A driver drowsiness detection (DDD) system plays a vital role
system employed various applications using blink rate, eye
closure and yawning to effectively and quickly identify the
to prevent road accidents. There has been extensive research
drowsiness of a driver while driving the vehicle and alarm the and a number of research papers have put forth possible
driver accordingly. methodologies to detect inattentiveness and drowsiness in a
driver in the last two decades. The traditional techniques are
Keywords: Drowsiness Detection, Prevention, Road Accidents, elaborated which are based on physiological measurements
Eye State, Alarm
including brain waves, heart rate, pulse rate and respiration.
I. INTRODUCTION However, these techniques are intrusive in nature.

D riving fatigue is a common phenomenon due to long time

of driving or lack of sleep. It is a significant potential
hazard in traffic safety. As many as 100,000 traffic accidents
Drowsiness in humans is characterized by a few very specific
movements and facial expressions - the eyes begin to close,
mouth opens in a yawn, the jaw goes slack and the neck tilts.
caused by driving fatigue which led to 400,000 people injured This paper focuses on tracking the eyes and mouth to detect
and 1550 people killed happened in the United States each drowsiness and detects whether the driver is drowsy. For real
year [1]. time application of the model, the input video is acquired by
mounting a camera on the dashboard of the car and can
Driver Drowsiness Detection is a car safety technology which accommodate the driver’s face, hands, upper body and
helps prevent accidents caused by the driver getting drowsy. occlusions such as non-tinted spectacles.
Various studies have suggested that around 20% of all road
accidents are fatigue-related, up to 50% on certain roads. According to literature, artificial intelligence and computer
Some of the current systems learn driver patterns and can vision algorithms are mostly utilized in the past systems to
detect when a driver is becoming drowsy. Various detect driver fatigue. In practice, there are several vision-
technologies can be used to try to detect driver drowsiness. based techniques proposed in the past to detect fatigue.
According to past literature, change in weather conditions,
1) Steering pattern monitoring- Primarily uses steering input environmental parameters and long-drive are some of the
from electric power steering system. Monitoring a driver this factors which contribute to drowsiness. Although, state-of-
way only works as long a driver actually steers a vehicle the-art Driver Drowsiness Detection systems are considered
actively instead of an automatic lane-keeping system. effective and reliable during day-time but provide subtle
2) Vehicle position in lane monitoring- Uses lane monitoring results during night time driving. In previous works, it has
camera. Monitoring a driver this way only works as long as a been observed that there are several problems presented.
driver actually steers a vehicle actively instead of an Those problems are natural light, head center-off position and
automatic lane-keeping system. the detection of eyes due to large sunglasses were not perfect.
Moreover, most modern methods tested on video data where
3) Driver eye/face monitoring- Uses computer vision to people simulate fatigue behaviour but less focused on real-
observe the driver's face, either using a built-in camera or on time detection of driver fatigue. They did not experiment with
mobile devices. real-time data, where people are naturally stressed. Since the
last few years, many researchers were working on developing

www.ijltemas.in Page 13
 International Journal of Latest Technology in Engineering, Management & Applied Science (IJLTEMAS)
Volume IX, Is
Issue XI, November 2020 | ISSN 2278-2540

Driver Drowsiness Detection systems using many diverse  Input the image and clearly identify the face using
techniques. The driver's attentiveness
ess can also be driver
driver- Viola Jones Algorithm
driven with a focus on vehicle manner. Other technical  Clearly detect 68 points to identify (x,y) (x
methods are based on the driver's state. The best ways for coordinate various region of the face.
correct and relatively accurate detection is based on human  Localize the landmarks to detect eye, mouth
drivers such as brain waves and heartbeats. region etc.
Various methodological
logical techniques for the detection and  Change in the shape of said region reports various
prediction of driver fatigue have been developed through expressions for drowsiness and fatigue.
advanced image ge processing, machine learning and
computational intelligence algorithms. The purpose of this High vision cameras are embedded to monitor, capture and
study is to first investigate the underlying intell intelligent extractt frames one by one and generate the alerts accordingly.
algorithms that are applied to detect driver fatigue and then Each extracted frame is analyzed to study the pattern of facial
use advanced techniques to accurately detect driver fatigue features; using Haar Cascade Classifiers and determined Eye
suitable in the day and night-time driving. Aspect Ratio (EAR) and Mouth Aspect Ratio (MAR) for each
frame. EAR and MAR values exceed their respective
All the state-of-the-art
art classification techniques for Driver threshold values, a blink and a yawn is considered
Drowsiness Detection systems are error-prone.
prone. Hence, they respectively. The system alerts the driver by playing an alarm
have some limitations and problems regarding their use in if eye blinking rate and yawns are suspected for a certain
real-life
life scenarios. In practice, the authors focused more on number of consecutive frames. The alarm is activated
activate to grab
developing Driver Drowsiness Detection assistance systems the driver’s attention and keep on ringing until driver wakes
and they aimed at helping drivers for detecting
ecting their moments up [3].
of distraction and generate warning alarm.

Fig. 2.. Stepwise process Real Time Drowsiness Detection System

Fig. 1.. A Visual Example of Driver Fatigue Detection (DDD) Systems based
on different Approaches

II. LITERATURE SURVEY

Drowsiness Detection of a Driver using Conventional
Computer Vision Application (2020)
In this paper, pre-existing
existing features for facial landmark
detection is used. The stepwise process of the system is shown
in Figure 2. The methodology uses 68- facial landmark (a
predefined landmark) for shape prediction in order to identify
various
ous regions of the face like eye brows, eye, mo
mouth region
etc as shown in Figure 3.
The facial landmark recognition is carried out as follows:
Fig. 3. Visualizing the 68 facial landmarks coordinates using Dilb facial plot

www.ijltemas.in Page 14
 International Journal of Latest Technology in Engineering, Management & Applied Science (IJLTEMAS)
Volume IX, Is
Issue XI, November 2020 | ISSN 2278-2540

Hybrid Fatigue: A Real-Time Driver

iver Drowsiness Detection frequencies of gradient direction of an image in localized
using Hybrid Features and Transfer Learning (2020) regions are used to form histograms. It is used to map the
coordinates of the facial landmarks of the input video and
A systematic flow diagram of the Hybrid Fatigue system is
drowsiness detected by monitoring aspect
aspe ratios of eyes and
given in Figure 4. The Hybrid Fatigue system is based on
mouth.
integrating visual features through PERCLOS measure and
non-visual features by heart-beat
beat (ECG) sensors. A hybrid The following steps are followed for the testing of the model:
system combines both visual and non-visual
visual features. Hybrid
features are extracted and classified as driver fatigue by  Input video (pre-recorded
recorded or real-time)
real is fed into
advanced deep-learning-based
based architectures in real
real-time. A the model. Individual frames are resized and
multi-layer based transfer learning
earning approach by using a converted to grayscale.
convolutional neural network (CNN) and deepdeep-belief network  Dlib’s HOG based face detector is initialised.
initia The
(DBN) is used to detect driver fatigue from hybrid features. location of the face is pinpointed.
 The facial landmarks for the face region are
The proposed Hybrid Fatigue system is developed based oon determined by the predictor and mapped onto the
the following stages: face.
 The first step is to use 68-different
ifferent landmark points  Left eye, right eye and mouth coordinates are
to track driver fatigue from the background video extracted, which are then used to compute aspect
frames. ratio for both
h eyes and mouth based on Euclidean
 After detecting and segmenting the landmarks points distance respectively.
from two-mounted
mounted cameras on vehicle, the next step  The calculated aspect ratios are compared with
is to extract different visual features along with ECG fixed threshold values to determine signs of
non-visual features. drowsiness. If the average aspect ratio of left and
 After collecting those hybrid parameters thorough the right eye falls below the threshold, it is
CNN model at the collection site, this intelligent data recognized d as a sign of drowsiness. Similarly, if
is then passed on to automated and operator
operator-assisted the mouth aspect ratio exceeds the set threshold,
applications, which is developed through the DBN there is a possibility for it to be a yawn.
model for classification [4].
 When continuous signs of drowsiness is detected over a
longer duration,
ration, the driver is alerted [5].
[5

Fig. 5. Block diagram of proposed real-time

time drowsiness detection system

Driver Drowsiness Detection using

g Percentage Eye Closure
C
method (2020)
One of the main reasons for accidents is drowsiness or fatigue
in driver. Preventing the drowsiness prevents accidents and it
employsys road safety. There are many types of analysis done
Fig. 4. A Systematic Diagram of Develop Systems of Hybrid Fatigue
Algorithm
regarding the drowsiness which are seen in driver such as
yawning, eye closure and head movement. The proposed
Design of Real-time
time Drowsiness Detection System usin
using Dlib system focuses on eye closure by the effective method named
(2019) Percentage Eye Closure (PERCLOS).
(PERCLOS The PERCLOS
Drowsy driver detection system is designed using Python and establishes a parameter level to detect the drowsiness. The
Dlib model. This model is trained to identify 64 facial detection is carried out by Viola-Jones
Viola detector, which
landmarks. As shown in Figure 5,, the drowsiness features are segments the driver face and image of the eye from the
extracted and the driver is alerted in case of drowsiness. The detector. The main concept of this method is that it rejects the
Dlib library is used to detect and localize facial landmarks non-face
ace quickly and spends time on the face region.
using Dlib’s pre-trained
trained facial landmark detector called The proposed flow of this system is given in Figure 6.
Histogram of Oriented
riented Gradients (HOG). In this method,

www.ijltemas.in Page 15
 International Journal of Latest Technology in Engineering, Management & Applied Science (IJLTEMAS)
Volume IX, Is
Issue XI, November 2020 | ISSN 2278-2540

 Eye closure and yawning detection

The main concept is detecting the driver’s face and set it to
Region off Interest (ROI). Next use ROI to find targets as eye
and mouth. This process starts by getting input from infrared
2D camera and processing by MATLAB. First step author
will set the infrared camera at the top of the vehicle facing the
driver and records the driver’s random action like blink,
closing eyes and yawn randomly. When the program detects
the symptoms of driver drowsiness, ss, program will print red
rectangle around symptom area such as yawning will print red
rectangle around driver’s mouth and eyes closure will print
with red rectangle
angle around driver’s eyes area [7].
[

Fig. 6. Process flow of the Proposed System

It starts with capturing the original image and then it is

Fig. 7. Flowchart of the proposed method
observed by Viola-Jones
Jones method if it is face or not, which is
done by AdaBoost. If it is face then by the PERCLOS method Alert System forr Driver’s Drowsiness using
usin Image Processing
it detects the eye part, and it checks whether it is open or (2019)
closed. PERCLOS is directly proportional to time for which
Accidents occur all over the world cause of not being able to
the eyelids remain closed more than the required time period.
concentrate on the road while driving. The concentration is
It is searched and related with the help of database images if
missed due to driving the car without resting or if consumed
the eye is closed then the alert message is sent to the driver. If
alcohol which makes the he person drowsy. This problem is
eye is open then the driver is alert and no me measure is taken
overcome by developing the various system in detecting the
regarding this [6].
sleep. The architecture of the proposed system is given in
A Method of Driver’s eyes closure and yawning detection for Figure 8. The system uses the Raspberry Pi and various
drowsiness analysis by Infrared cameraa (2019) sensors like Gas Sensor, Vibration Sensor for the detection of
the type of drowsiness. The driver is been monitored by
A challenge in the area of the driver drowsiness detection is to
placing a camera which captures the vital sign. If the eye is
detect the drowsiness in low light condition. The proposed
closed for a longer period of time then the image of the person
system is a method to detect driver’s eye closure and yawning
is sent to cloud.
for drowsiness analysis by infrared camera. The flowchart of
the proposed system is given in Figure 7. The advantage of The accident is detected using a vibration sensor and the
this method is that it can detect eye closure and yawning in server is notified by sending latitude and longitude. The
low light condition. location of the car is sent by the IOT modem which is
embedded in the car. This system goes one step ahead and
This method consists of four steps, namely,
detects if the driver has consumed alcohol with the help of gas
 Face detection sensor. If the driver is found unfit to drive then an alert
 Eye detection message is sent to the driver. This system detects not only
 Mouth detection

www.ijltemas.in Page 16
 International Journal of Latest Technology in Engineering, Management & Applied Science (IJLTEMAS)
Volume IX, Is
Issue XI, November 2020 | ISSN 2278-2540

drowsiness but also detects if the car is met with slight classify and detect the eye closure levels like facial landmark
accident and sends
ends information to the cloud [8
[8]. recognition, PERCLOS measure, Dlib model, Viola-Jones
Viola
method and sensor.
REFERENCES
[1] S. R. Paul, NHTSA's Drowsy Driver Technology Program,
Progra 2005.
[2] https://en.wikipedia.org/wiki/Driver_drowsiness_detection
[3] Hitendra Garg Drowsiness Detection of a Driver using
Conventional Computer Vision Application, 2020 International
Conference
onference on Power Electronics & IOT Applications in
Renewable Energy and its control (PARC), Mathura, Uttar
Pradesh, India.
[4] Qaisar Abbas Hybrid Fatigue: A Real-Time
Real Driver Drowsiness
Detection using Hybrid Features and Transfer Learning, Learning
International Journal
urnal of Advanced Computer Science and
Applications, Vol. 11, No. 1, 2020
202
[5] Shruti Mohanty, Shruti V Hegde, Supriya Prasad, J. Manikandan
Design of Real-Time
Time Drowsiness Detection System using Dlib,
2019 5th IEEE International WIE Conference on Electrical and
Computer Engineering, 15-16 16 Nov 2019, Bangalore, India.
[6] K. Sakthidasan, N. Vasudevan, V. Nagarajan Driver Drowsiness
Detection using Percentage Eye Closure method, 2020
International Conference on Communication and Signal
Processing, Chennai, India.
[7] Wisaroot
aroot Tipprasert, Theekapun Charoenpong, Chamaporn
Chianrabutra, Chamaiporn Sukjamsri A Method of Driver’s Eyes
Closure and Yawning Detection for Drowsiness Analysis by
Fig. 8. Architecture of the proposed
ed system Infrared
[8] Camera, 2019 1st International Symposium on Instrumentation,
III. CONCLUSIONS Control, Artificial
al Intelligence and Robotics, Bangkok, Thailand.
[9] Ratna Kaavya M, Ramya V, Ramya G Franklin Alert System for
Drowsiness
owsiness in drivers while travelling is the major cause for Driver’s Drwosiness using Image Processing, 2019 International
the accidents. An elaborate literature survey is carried out on Conference on Vision towards Emerging Trends in
drowsiness detection of drivers while travelling and a Communication and Networking, Vellore
Vellor India, March 2019.
comparison is made on the various methods of detectin detecting
drowsiness. This paper surveys the various methods to

www.ijltemas.in Page 17