MACHINE LEARNING BASED DRIVER MONITORING SYSTEM

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 10 Issue: 12 | Dec 2023 www.irjet.net p-ISSN: 2395-0072
© 2023, IRJET | Impact Factor value: 8.226 | ISO 9001:2008 Certified Journal | Page 986
MACHINE LEARNING BASED DRIVER MONITORING SYSTEM
Arun Binoy1, John Pius Thayiparampil2, Shijina B3, Alby Alphonsa Joseph4, Athira R Kurup5,
Safna Sainudeen6
1-2BTECH UG Students, Department of Computer Science and Engineering, TOMS college of Engineering
3Head of the department, computer science and engineering, TOMS college of Engineering
4-6Assistant professor, computer science and engineering, TOMS college of Engineering
APJ Abdul Kalam Technological University, Kerala, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - The creation of efficient driver monitoring
systems has become essential due to the rise in traffic
accidents and driver-related safety issues. In order to improve
road safety, this project introduces a machine learning-based
Driver Monitoring System (DMS) that continuously evaluates
and warns drivers of potential hazards. The DMS analyses
driver behaviour in real-time, including gaze tracking, head
movement, eye closure, and facial expressions, by utilizing
cutting-edge computer vision and machine learning
techniques. The system uses a range of sensors, such as
cameras and in-cabin sensors, to gather information on the
driver's movements and environment. Machine learning
models are utilised to identify andcategorisearangeofsafety-
critical occurrences, including fatigue, inattention, andlackof
focus. To avoid collisions, the technology promptly warns the
driver when it detects a risk. Facial expressions are used inthe
proposed study to implement a Support Vector Machines
(SVM) based emotion identification algorithm. When
evaluated in situations of varying brightness, the algorithm
performed more accurately than existing research.
Key Words: Machine learning, Gaze tracking ,Real time
analysis, machine learning model, Eye closure, Driver
monitor system, computer vision
1.INTRODUCTION
The use of machine learning has brought about
revolutionary changes in our daily livesinthisdayandageof
swift technology advancement and increasing focus on road
safety. One noteworthy application is in the field of
intelligent transportation systems, whereDriverMonitoring
Systems (DMS) based on Machine Learning have become an
essential part. These technologies, which continuously
monitor and warn drivers of possible problems like
weariness,sleepiness,anddistraction,representa significant
advancement in improving road safety.Road accidents
continue to be a major problem despite advancementsin car
safety, with a significant percentageofthecausebeingdriver
behaviour. Globally, drowsiness and attention have been
identified as the main causes of traffic accidents. It is
essential to create technology that can both detect and react
to the nuances of human behaviour when driving. A creative
approach to this problem is machine learning-based DMS,
which provides proactive interventions and real-time
monitoring to reduce risks and improve driving enjoyment.
The significance of road safety cannot be overstated, and
cutting-edge driver monitoring and support systems are
essential to creating safer roadways. Significant
advancements in modelling and improving these systems
have been made in the last ten years, which have enhanced
driver performance and decreased the frequency and
severity of accidents. One prominent example is the
Advanced Driver Assistance System (ADAS), which was
developed to improve driving conditions while making a
major contribution to traffic safety overall. Among themany
capabilities included in ADAS are automated parking, road
sign recognition, pedestrian identification, and—most
importantly—driver tiredness detection. As one of the main
factors contributing to traffic accidents,driver wearinesshas
sparked a lot of study into mitigation and assessment
techniques. In contrast to other factors that contribute to
accidents, driver weariness has a progressive effect that
results in both detectable psycho-physiological indicators
and a noticeable deterioration in driving performance. The
progress in technology has led to the development of novel
methods for assessing and calculating driver fatigue.
These methods include measuringheadandeyemovements,
using electroencephalography to monitor brainactivity, and
examining a range of driver performance metrics, including
lane monitoring, steering wheel movements, and blinking.
The creation of an affordable drowsiness detection system
based on the PERcentage of eye CLOsure (PERCLOS)
approach is presented in this research. Crucially, the
PERCLOS approach has proven to be accurate in identifying
tiredness and is non-intrusive while being pleasant for the
driver. It can identify microsleeps based on a preset
threshold value and, unlike some other approaches, is
resistant to environmental influences like road
conditions.The positioning of a camera to record a livevideo
feed of the driver's face is part of the PERCLOS method's
operational architecture.
1.1 Problem Definition
In order to solve this crucial issue for road safety, the
focus of the current project is on the early identification and

detection of driver fatigue. To this end, a hybrid machine
learning techniquewillbeutilised.Driverfatiguecontinuesto
be a major contributing factor in collisions,endangeringboth
individual motorists and the general public's safety. Drivers'
reaction times and general drivingskillsdegradewithfatigue
and distraction, raising the possibility of collisions. The main
obstacle is creating a reliable and accurate system that can
detect drivers' symptoms of tiredness in real time. This
entails examining a variety ofphysiological,behavioural,and
environmental elements, including eye movements, facial
expressions, vital signs, car information, and ambient
elements including lighting and road quality.
1.2 Objective
Our machine learning driver monitoring system's main
goal is to increase road safety by continually seeing and
evaluating driver behaviour in real-time, identifying
problematic behaviours includingdiversionsandsleepiness.
It will build precise models for behaviour detection while
enabling customisable alerts and user-friendly interfacesby
utilising a variety of data sources,includingvideoandsensor
data. Our goals include adhering tosafetylaws,beingflexible
enough to work with various car kinds and sectors,
improving continuously by retraining models, and placing a
high priority on data security and privacy. Another
important factor is cost-effectiveness, which enables the
system to be used for a variety of purposes and, in the end,
promotes safer and more responsible driving practises as
well as fewer traffic incidents.
Support Vector Machine (SVM) methods and PERCLOS
(Percentage of Eye Closure) analysis are two novel
components that work together to form the hybrid machine
learning driver monitoring system.PERCLOSisusedtotrack
how much a driver is sleeping by examining how much of
their eyes are closed, which is a good sign of exhaustion. The
system is therefore better able to identify and categorise
different driving behaviours once this data is fed into SVM, a
potent machine learning algorithm. SVM is trained to
identify not just other important factors like awareness,
distractions, and possibly risky actions, but also tiredness.
When PERCLOS and SVM are used together, they provide a
reliable and flexible driver monitoring solution that
enhances road safety by precisely recognising and reacting
to different driver states and behaviours.
2. LITERATURE SURVEY
A technique based on Adaboost and Haar-like features
was developed by the authors in [9] to train a cascade
classifier that demonstrated excellent face detection
performance. The fundamental idea behind this feature-
based method is that the face in the picture is recognised
using a set of fundamental traits, independent of the
surrounding lighting, the face's direction, or the subject's
posture.
Over time, machine learning has shown to be a reliable
foundation for classifying eye states and detecting faces.
Convolution neural network (CNN)-based deep learning has
quickly become a potent technique in face detection,
especially for drowsiness detection [10]. In order to identify
driver tiredness, a multitasking Convolutional Neural
Network model was created in [11]. Toidentifytiredness,the
model makes use of alterations in the driver's mouth and
ocular characteristics. The authors of [12] suggested a CNN-
based spatiotemporal strategy for real-time driver state
monitoring, in which action recognition is derived from
temporal data in addition to geographical data.
A combination of atwo-levelattentionbidirectionalLSTM
networkanda3Dconditionalgenerativeadversarialnetwork
forms the basis of a deep learning model for accurate
sleepiness prediction that is shown in [13]. In order to
retrieve short-term spatial-temporal features with a variety
of fatigue-related data, a 3D encoder-decoder generator was
created to raise high-resolution fake image patterns and
implement a 3D discriminator to predict fatigue incidents
from the spatial-temporal domain. The use of a two-level
attention strategy to direct the bidirectional LSTM to
recognise the importance of temporal data for long-term
spatial-temporal fusion and memory data for brief intervals
was also examined by the researchers.
A machine learning-based real-time image classification
and sleepiness system was put into place by Altameem et al.
[14]. An emotion recognition method based on Support
Vector Machines (SVM) is developed using facial features.
High performancewasshownbythealgorithmundervarious
brightness conditions.Becausethesystemisconnectedtothe
vehicle'scircuitry, it can track the vehicle's data and produce
more accurate results. Sensors in autonomous vehicles must
be able to identify whether a driver is weary, agitated, or
going through strong emotional swings like rage. These
sensing devices need to continuously track the driver's face
and identify facial landmarks in order to determine the
driver's condition and determine if they are driving safely.
Through interactive simulations, Esteves et al. explored
the application of machine learning and signal processing
techniques in connected automobiles to detect driver
weariness. Using the electrocardiogram and facial
recognition, extensive biometric research has made it
possible to continue developing subject-specific tiredness
frameworks for accurate prediction. Cheng et al. proposed a
multi-pronged method to identify whether the driver's facial
landmark sequence changes from alert to tired. The
simulation tool was used in the design and execution of the
investigation. The eye aspect ratio and mouth aspect ratio
patterns are generated based on face landmarks. We
gathered data on blink rate, average blink duration, closing
and reopening frequencies, percentage of closed eyes over a
predetermined period of time, and percentage of yawning. A
sleepiness evaluation methodology is proposed after the
features that were extracted. Several machine learning

approaches were used in the development of the fatigue
assessment system. Koohestani et al.'s main goal is to use a
range of machine learning techniques to analyse the driving
experience. There are two main stages in the optimisation
component of the proposed system. During the first stage, K-
nearest neighbours, support vector machines, and naïve
Bayesalgorithmsareoptimisedformaximumefficiencyusing
bagging, boosting, and votingensemblelearningapproaches.
Subsequently, four innovative optimisation methods (the
grey wolf optimizer, particle swarm optimisation, whale
optimisation algorithm,andantlionoptimizer)areappliedto
increase the system's overall functionality by enhancing its
variables. In [8], a sleepiness detection system for driver
monitoring is proposed that may be customised for use in
trucks and buses.
Several sub-systems of the design include head-pose
estimation, face identification, eye detection, eye-state
classification, and fatigue estimate. There are two primary
steps involvedinimplementingthesuggestedsystem.Foreye
state classification, a fusion model is employed after spectral
regression has been applied for eye tracking. The next stepis
to use PERCLOS to identify if the eye is open or closed.
3. METHODOLOGY
The suggested solution comprises of a camera mounted in a
bus so that it can monitor the driver's face. The driver's face
is detected by the camera, which also localises their eyes,
which are then classed as open or closed. The camera
records the live-stream footage and feeds the data into the
machine learning programmed. The driver monitoring
system alerts the driver with an alarm when it detects that
the eyes have been closed for a longer duration than
predetermined.
The system receives a live-stream video of the driver's face
as input; it outputs the driver's state—alert ordrowsy—and
sounds an alarm if it detects drowsiness. The movie was
divided into frames, face detection was applied to each
frame, and the ocular region of each discovered face was
localised. The driver's alertness or drowsiness was then
determined by applying eye state analysis to the eye area.
The project's algorithms are described in more detail below.
The Viola-Jones technique, which makes use of scalar
products between the image and a few Haar-like templates,
was used to recognise faces. The four primary steps of the
Viola-Jones approach are: choosing Haar-like features,
building an integral image, doing AdaBoost training, and
building classifier cascades. The analysis of eye state
encompasses the duration and rate of blinks, as well as the
open and closed states of the eyes. Blink and open/close eye
state detection were implemented using a convolutional
neural network. The driver's eye was closed for a certain
number of consecutive frames, which resulted in a high
enough departure from the threshold value to cause an
alarm. This method of detecting drowsiness was then
applied.
. Fig -1: System Model Flow Diagram
3.1 System Description
An advanced technology called a Driver Monitoring System
(DMS) monitors and evaluates a driver's behaviour and
condition while operating a vehicleusinga varietyofsensors
and data processing techniques. A DMS's major objective is
to improve road safety by making sure that drivers are
focused, aware, and in a condition that allows for safe
driving. This is a driver monitoring system description. By
collecting data from drivers in real-time under various
settings using in-car sensors and cameras, the proposed
driver drowsiness monitoring systemseekstoimprove road
safety. Machine learning models, such as Convolutional and
Recurrent Neural Networks,aretrainedtopredictsleepiness
levels after preprocessing and feature extraction. When
abnormal behaviour is noticed, ongoing monitoring sets off
haptic, aural, and visual alerts. Data logging and adaptive
alert levels help drivers and monitoring systems alike, and
an intuitive user interface gradually enhances system
performance. A reliable and efficient method for averting
mishaps and sparing lives is ensured by privacyandsecurity
precautions, regulatory compliance, frequent model
upgrades, and possible vehicle integration.
Fig -2: four stages of driver drowsiness system.

Fig -3: Flowchart of the drowsiness detection system.
3.2 Image and Video Input Stream
The technology would require input from photos andvideos
in order to detect faces and compare them to a database of
recognised individuals. CCTV cameras, cameras worn while
moving, and other sources may provide this information.
3.3 Face Recognition
The proposed driver sleepiness monitoring system relies
heavily on face recognition, which uses sophisticated
algorithms to reliably identify registeredusers.Inparticular,
the system's facial recognition capabilities is based mostly
on the Haar Cascade algorithm.
For the purpose of identifying faces in pictures or video
frames, the Haar Cascade algorithm is a trustworthy
technique. Its use in this system entails identifying facial
features including the mouth,nose,and eyes, whichhelpsthe
system locate and identify faces with accuracy. By using this
technique, the system is able to accurately identify
registered users by differentiating them based onfacetraits.
Moreover, the face recognitioncomponentplaysa majorrole
in the system's overall functionality. It enables tailored
monitoring and analysis of individual sleepiness levels by
precisely identifying drivers. This customised method
improves the efficacy of the sleepiness detection and alert
system by enabling it to better adjust alerts and actions
based on the driving habits of individual drivers.
Furthermore, the system's emphasis on user-friendliness is
in accordance with the face recognition integration. Drivers
can have a smooth and customised experience with the
system by enrolling individuals and linking their facial data
to certain profiles. This guarantees that every registered
user's unique demands and features are taken into account
by the monitoring and alarm system. Furthermore, the
integration of facial recognition technology enhances the
privacy and security protocols built into the system. By
limiting access to the monitoring functions to justthose who
are authorised, it protects confidential driver data and
complies with privacy laws.
In general, the incorporation offacial recognitionalgorithms,
specifically the Haar Cascade algorithm, enhances the
system's capacity to precisely identify persons who have
registered, allowing for customised surveillance while
upholding strong privacy and security protocols.
After the text edit has been completed, the paper is ready for
the template. Duplicatethe templatefile by using the SaveAs
command, andusethenamingconventionprescribedbyyour
conference for the name of your paper. In this newly created
file, highlight all of the contents and import your prepared
text file. You are now ready to style your paper.
3.4 Facial landmark detection
A key element in the creation of machine learning-based
driver sleepiness monitoring systems is facial landmark
detection. The training of a model to identify important face
characteristics including the lips, nose, and eyes allows real-
time tracking of a driver's movements and expressions. The
indicators of drowsiness, such as closed eyes, head position,
blink rate, yawning, and pupil dilation, can then be evaluated
using this data.
Following the identification of these signs,analertingsystem
is triggered by a predetermined set of rules and thresholds,
warning the driver and eventually improving road safety by
preventing accidents caused by intoxicated driving.
Fig -4: facial landmark system
3.5 Face Segmentation
An essential part of a machine learning-based driver
sleepiness monitoring system is face segmentation. In order

to preciselyanalyse facial features and expressions, it entails
detecting and isolating the driver's facial region from the
entire video feed or image. The technology can efficiently
track and monitor important signs of drowsiness, like eye
closure, head position, and facial muscle movements, by
segmenting the face. For the purpose of precisely
determining the driver's level of attentiveness and
guaranteeing prompt notifications or interventions to
improve road safety, this segmentation procedure is
essential.
Fig -5: facial landmark system
3.6 Face Segmentation
A two-stage method is commonly used by machine learning-
based driver drowsiness monitoring systems to identify and
distinguish between a driver's normal and drowsy states.
The systemcontinuouslyexaminesavarietyofdriver-specific
cues during the normal state stage, including head position,
eye movements, steering wheel behaviour, and facial
expressions. Machinelearning modelsare trained to identify
patternslinked to alertnessand attentiveness. These models
arecommonlybuiltonconvolutionalneuralnetworks(CNNs)
and recurrent neural networks (RNNs). The system stays
inactive and the driver is free to operate the car as usual
when these indicators point to the driver being in a normal
state.
Fig -6: face detection during normal state.
During the sleepy state phase, the system assumes a
proactive role and initiates notifications upon identifying
indicators of inattention, such drooping eyelids or sluggish
reaction times.
Now that these signs have been trained to recognise them,
the ML models can inform the driver to take a break or do
something to regain consciousness by displaying visual or
audible warnings. The driver sleepiness monitoring system
can balance convenience and safety with this two-stage
method, interfering only when necessary to maintain driver
autonomy and promote safer driving.
Fig -7: face detection during drowsy state.
3.7 Blink detection
By determining the coordinates of the eye landmarks on the
face, the distance between the vertical and horizontal eye
landmarks may be calculated . The eye aspect ratio (EAR) is
calculated using the two sets of distances between the
various eyes. In this study, the blink detectionalgorithmuses
facial landmark detectors for the localization of eyelid
contours and eyes. The eye state is then identified as open or
closed using the eye aspect ratio, which is obtained from the
eye contours.Theocularlandmarksareidentifiedandlocated
for every frame in the film as shown in Figure
Fig -8: Segmenting image of eye

It is evident from the figure in Figure that the widthand
height of the coordinates are related. Equation, which is
derived from equation, describes the connection between
the width and height of the coordinates and is known as the
Eye Aspect Ratio (EAR). where the 2D face landmark
positions shown in Figure are P1, P2,..., P6. This equation's
numerator calculates the separation between vertical eye
landmarks, while its denominator calculates the separation
between horizontal eye landmarks. The studies' suggested
EAR threshold of 0.3 was used. A blink is detected if the EAR
drops below 0.3 and then climbs above it. Three was chosen
as the EAR consecutive frames threshold. This suggests that
for a blink to be recorded, three consecutive frames with an
EAR below the EAR threshold must occur. The output frame
displays the number of blinks.
Fig -9: analyzing graph
4. DRIVER MONITORING SYSTEM
A sophisticated technology called a Driver Monitoring
System (DMS) monitors and evaluates a driver's behaviour
and condition while operating a vehicle using a variety of
sensors and data processing techniques. A DMS's main
objective is to improve road safety by making sure that
drivers are focused, aware, and in a condition that allowsfor
safe driving. This is an explanation of a driver monitoring
system.
4.1 Haarcade Algorithm for face detection and
Recognition
In computer vision applications, the Haar Cascadealgorithm
is a reliable and popular technique forfaceidentificationand
recognition. Its efficacy stems from its capacity to precisely
identify faces in pictures or videoclips,whichmakesituseful
in a variety of industries including biometrics, surveillance,
and human-computer interaction. The algorithm's training
phase begins with a sizable dataset of faces in positive
photos and Faceless negative photographs are employed. In
this stage, the algorithm examines these pictures and
extracts characteristics that resemble Haar. Rectangular
patterns known as Haar-like features are used to represent
several aspects of images, including edges, lines, and
textures. After that, integral ages are computed to expedite
the feature extraction procedure. Calculating the sums of
pixel intensities inside rectangular sections may be done
quickly and efficiently with integral pictures.
The most illuminating Haar-like feature subset is then
chosen using the Adaboost method. Each characteristic is
given a weight by Adaboost depending on how well it can
distinguish between positive andnegativesamples.Faceand
non-face areas may be distinguished by the algorithm by
concentrating on the most discriminative attributes.
Fig -10: face detection framework
Due to its effectiveness and precision, the Haar Cascade
technique is frequently used for face detection applications.
It is appropriate for real-world applications due to its
capacity to manage changes in lighting conditions, partial
occlusions, and face positions. However, it might not work
well in situations involving significant position fluctuations
or low-resolution photos. However, for face identification
and recognition in computer vision, the Haar Cascade
method continues to be an essential tool.
In the context of a driver drowsiness monitoring system,
Support Vector Machine (SVM) is a supervised machine
learning method that may be used to categorise and identify
sleepy and alert states of a driver based on attributes taken
from pictures or video frames. The following explains how
SVM may be used to this system:
1. Data collection and feature extraction: The driver's face is
collected in pictures or video frames, and pertinent features
are retrieved for analysis, such as the length of the driver's
eye closure and facial landmarks.

2. Data Labelling: Depending on the driver's real state at the
time of data collection, the obtained data is categorized as
"alert" or "drowsy."
3. Data Splitting: To train an SVM model, the dataset is split
into two parts: a training set and a testing set. Thetesting set
is used to assess the model's performance.
4. Training the SVM: Using the training data, the SVM
algorithm learns to identify the optimal hyperplane that
divides the "alert" and "drowsy" classes.
5. Kernel Functions: To help identify intricate correlations
between characteristics, kernel functions may be used to
turn the data into higher-dimensional spaces.
Fig -11: Support Vector machine
6. Testing the SVM: To evaluate the trained SVM model's
accuracy in classifying sleepy and awake states, it is tested
on a different testing dataset.
7.Evaluation: To determine how well the model identifies
drivers, measures such as accuracy, precision, recall, F1-
score, and ROC curves are used to assess the model's
performance.
8. Real-time Inference: To generate predictions in real-time,
the SVM model is utilised. identifying whether the driver is
"alert" or "drowsy" based on real-time video frames.
9.Alert Mechanism: The systemsendsoutnotificationsto the
user based on the SVM's predictions. when a driver exhibits
signs of fatigue, such as haptic feedback, auditory alerts, or
visual input to improve the safety of the roads.
5. CONCLUSIONS
To sum up, the present condition of MachineLearningBased
Driver Monitoring Systems signifies a noteworthy
progression in augmenting traffic safetyandcustomising the
driving encounter. These devices provide in-the-moment
driver behaviour monitoring, which helps avert accidents
brought on by intoxication, weariness, or impairment.
Nonetheless, issues with accuracy and privacy must be
resolved. These systems are probablygoingtobecommonin
cars in the future because to legal regulations and customer
desire for safer and more convenient driving. According to
our study, real-time Drowsiness Detection Techniques
perform effectively in a range of illumination scenarios.
Hardware was used as input for our support vector machine
and image processing methods for video analysis. The ideal
camera distance and lighting conditions were ideal for the
algorithm's performance. With increased camera distance
and in poor light, accuracy dropped. It is possible to test this
proposed algorithmundervariousbrightnessconditions and
with an improvedcamera.Several datasetsandmoderndeep
learning techniques may be used to test this approach.
ACKNOWLEDGEMENT
The authors can acknowledge any person/authoritiesinthis
section. This is not mandatory.
REFERENCES
[1] A. Altameem, A. Kumar, R. C. Poonia, S. KumarandA.K. J.
Saudagar, ”Early Identification and Detection of Driver
Drowsiness by Hybrid Machine Learning,” in IEEE
Access, vol. 9, pp. 162805-162819, 2021, doi:
10.1109/ACCESS.2021.3131601.
[2] A. Altameem, A. Kumar, R. C. Poonia, S. KumarandA.K. J.
Saudagar, ”Early Identification and Detection of Driver
Drowsiness by Hybrid Machine Learning,” in IEEE
Access, vol. 9, pp. 162805-162819, 2021, doi:
10.1109/ACCESS.2021.3131601.
[3] World Health Organization, Road Traffic Injuries, 20 08
2021. [Online]. Available: https://www.who.int/news-
room/fact sheets/detail/road-trafficinjuries
[4] Security enhancement utilizing motiondetection,Signal
Processing,Communication, ComputingandNetworking
Technologies (ICSCCN), 2011 International Conference
on, T huckalay, 2011, pp. 552–557. T homas, Ashraf, Lal,
Mathew, Jayashree.
[5] Moghaddam, B., Pentland, A.P.: Face recognition using
view-based and modular eigenspaces. In: Automatic
Systems for the IdentificationandInspectionofHumans,
vol. 2277, pp. 12–22 (1994),Sawhney, S., Kacker,K.,Jain,
S., Singh, S. N., Garg, R. (2o19, January). Real-time smart
attendance system using face recognition techniques.In
2o19 9th international conference on cloud computing,
data science engineering (Confluence) (pp. 522-525).
IEEE
[6] T Kundinger, A Riener, N Sofra, and K Weigl,” Driver
drowsiness in automated and manual driving: insights
from a test track study”, 25th International
ConferChapter 6.

[7] M. Turk and A. Pentland, “EigenfacesforRecognition,”in
Journal of cognitive neuroscience, vol 3, pp. 71-86, Jan
1991.
[8] [6] Roqueiro, Petrushin, Counting people using video
cameras, Department ofComputerScience,Universityof
Illinois at Chicago, Chicago, IL 60607, USA, 2006
[9] W. Kongcharoen, S. Nuchitprasitchai, Y.NilsiamandJ.M.
Pearce,” Real-Time Eye State Detection System for
Driver Drowsiness Using Convolu- tional Neural
Network,” 2020 17th International Conference on
Electrical Engineering/Electronics
[10] Chin, H.: Face recognition based automated student
attendance system. Diss. UTAR (2018)
[11] Ramanan, D., Zhu, X.: Face detection, pose estimation,
and landmark localization in the wild. In:Proceedingsof
the 2012 IEEE Conference on Computer Vision and
Pattern Recognition (CVPR), pp. 2879–2886 (2012)
[12] Yue, Wei, and Li. A Video Sequence Segmentation
Algorithm for ForegroundBackground.China’s Jiangnan
University.2015
[13] Chu Y, Ahmad T, Bebis G, Zhao L (2017) Low-resolution
face recognition with single sample per person. Signal
Process 141:144–15.
[14] C. Dewi, R.-C. Chen, X. Jiang, and H. Yu, “Adjusting eye
aspect ratio for strong eye blink detection based on
facial landmarks,” 2022, doi: 10.7717/peerj-cs.943. .

MACHINE LEARNING BASED DRIVER MONITORING SYSTEM

More Related Content

Similar to MACHINE LEARNING BASED DRIVER MONITORING SYSTEM (20)

More from IRJET Journal (20)

Recently uploaded (20)

MACHINE LEARNING BASED DRIVER MONITORING SYSTEM