See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/379676181

Raspberry Pi-Based Driver Drowsiness Detection

Conference Paper · February 2024

DOI: 10.1109/IC2PCT60090.2024.10486677

CITATIONS READS
2 29

5 authors, including:

Peddarapu Rama Krishna

VNR Vignana Jyothi Institute of Engineering & Technology
18 PUBLICATIONS 38 CITATIONS

SEE PROFILE

All content following this page was uploaded by Peddarapu Rama Krishna on 03 September 2024.

The user has requested enhancement of the downloaded file.

 2024 International Conference on Computing, Power, and Communication Technologies (IC2PCT)

Raspberry Pi-Based Driver Drowsiness Detection

2024 IEEE International Conference on Computing, Power and Communication Technologies (IC2PCT) | 979-8-3503-8354-6/24/$31.00 ©2024 IEEE | DOI: 10.1109/IC2PCT60090.2024.10486677

Rama Krishna Peddarapu Basuthkar Likhita Desireddy Monika

Computer Science and Engineering Computer Science and Engineering Computer Science and Engineering
VNRV JIET, Hyderabad, India VNRV JIET, Hyderabad, India VNRV JIET, Hyderabad, India
[email protected] [email protected] [email protected]

Sai Preeti Paruchuru Sri Lasya Kompella

Computer Science and Engineering Computer Science and Engineering
VNRV JIET, Hyderabad, India VNRV JIET, Hyderabad, India
[email protected] [email protected]

Abstract - Road safety is a major concern; however, a great drowsy driving detection by use of Raspberry Pi, which not
extent of the accidents is caused due to driver fatigue. Under this only has the property of being small and cheap but also
approach, an effective system to detect sleepy driving is sought to possesses enough power for processing in real-time. Our
structure. This type of system that delves into fatigue indicators approach looks into the applications of Raspberry Pi in the
including head pose, facial expressions, and eye tracking to analysis of visual and physiological manifestations of driver
identify drowsiness precisely in real time utilizes methodologies fatigue, such as eyelid closure rate, head position, as well as
such as machine learning and computer vision. The first goal is to yawn frequency. The integration of camera sensors combined
provide early warnings which will trigger the driver and prevent
with elegant image processing algorithms continuously
potential crashes. It is from this design that drivers and other
monitors and detects early signs of drowsiness while it is
people who control virtual devices get easy, immediate integration
into vehicles and personal electronic devices as it considerably
occurring. The information presented in this paper is crucial
diminishes drowsy driving risks. The choice of Raspberry Pi for minimizing traffic accidents caused by drowsy drivers.
technology used at the base of this solution is also due to the Such an applicable system is one based on Raspberry Pi in a
combination of performance, adaptability, and cost-efficiency, scalable method and adaptable to various models of vehicles.
thereby providing stability and general accessibility for virtually With its cost advantage, the proposed system has the potential
all multipurpose vehicles. Video capture and face detection are to update the technique of road safety by making it affordable
accomplished mainly by OpenCV and Haar cascade technologies. for manufacturing and use in all vehicles. This paper will
Warnings to alert drivers when their fatigue level is still at the articulate the design, implementation, and testing of the
tolerable, acceptable level is the pivotal role our technology plays drowsy driver detection system, hence showing the
in enhancing road safety by preventing accidents caused by driver effectiveness and reliability of the same. The above efforts are
fatigue and this ultimately helps to create safer driving expected to add value to the growing portfolio of research on
conditions. Our system, supported by Raspberry Pi technology automotive safety by availing knowledge, which will guide the
stands prepared as an essential tool in combating drowsy driving development of other critical systems in the domain.
- making substantial contributions towards ensuring road
safety and enhancing the well-being of drivers. A. Existing Work
Drowsiness detection utilizes quite a few methods to ensure
Keywords - Drowsiness detection, road safety, driver alertness, road safety while tackling the several challenges that are
computer vision, machine learning, Raspberry Pi, real-time possible. The techniques employed can be a deep learning-
detection, sleepiness indicators, facial analysis, eye tracking, EAR
based AI model that preprocesses the images and builds fuzzy
(Eye-Aspect-Ratio), MAR (Mouth Aspect Ratio), Haar cascades,
inference systems or an innovative strategy that applies EEG
driver monitoring, accidents prevention, deep learning, yawn
detection, safety alarm, false positive reduction, facial landmarks.
signals to perform binary classification with optimization.
Analyzing the behavioral and sensor-based physiological
I. INTRODUCTION measures revealed promising results in normal conditions with
the exception of a few limitations. A real-time computer
In recent times, the increasing frequency of car crashes that vision-based approach to detect drowsiness allows rapid
arise from drowsy driving has been a source of concern identification but compromises the accuracy when the eye
regarding road safety. An estimated 1,550 deaths, 71,000 frames are obstructed.
injuries, and financial losses totaling $12.5 billion occur on an
annual basis because of drowsy driving. Reports on the state of Moreover, some algorithmic systems track the driver in
drowsy driving reveal that about 4% of adult drivers have real-time with convolutional neural networks or with the help
confessed to experiencing sensations of sleepiness while of the eye closure duration as well as the facial landmarks.
driving in the previous month.[1] There is an increasing With help of facial analysis of OpenCV library which is
interest in developing efficient as well as economical systems installed in the Raspberry Pi, open environments are supported
for identifying drowsiness. This paper introduces a new way of by detection at no indicted by the environmental non-issue

Copyright © IEEE–2024 ISBN: 979-8-3503-8354-6 864

Authorized licensed use limited to: INDIAN INST OF INFO TECH AND MANAGEMENT. Downloaded on April 09,2024 at 08:01:27 UTC from IEEE Xplore. Restrictions apply.
 2024 International Conference on Computing, Power, and Communication Technologies (IC2PCT)

limitations. With direct face to face of driver to the video complete methodology was the optics of detection of eyes
capture device Haar Cascade detection and Circular Hough through the Viola and Jones algorithm and facial approach to
Transform for eye tracking, the system is effective. the landmarks processing. The datasets that were used
comprised the set of ZJU Eyeblink Database and a mixture
Additionally, several methods utilize superior deep dataset of ZJU and the authors’ designed dataset. It was
learning algorithms to estimate the eye movement, whereas the observed from the presented framework that it displayed high
competing methods are formed by employing the application precision with a minimum amount of computational load in
of image processing on the paradigm of determination of the recognizing drowsiness. Besides, the authors provided
eye open period as well observing the variability in the heart information on the potential of the approach by limiting it to
rate rate to identify drowsiness. These methods show the the competing relationship between dataset size and accuracy
crucial role of modern technologies in combating fatigue on by noting the visual parameters' jitters from session to session.
the driver’s side and encouraging driving practices which serve [4]
protection. Each strategy has its own specific pros and cons,
but nonetheless all mechanisms help to reduce the number of This paper seeks to develop a suggest that a CNN model
such a category of accidents as drowsy driving and achieve the using eye tracking can be utilized as a real-time detection of
safety of every driver. drowsy drivers. Their model depicted in this paper, by S. S.
Bhat, S. R. Kolhe, and S. S. Kulkarni has been trained and
These techniques cover a wide range of approaches, each validated using a dataset of eye images gathered from a
possessing distinct advantages and disadvantages, all with the webcam and it has been yielded 97% and 94.3% training and
goal of effectively addressing driver drowsiness detection and testing accuracy and 96% validation data accuracy. The model
improving overall road safety. uses the binary cross entropy loss function and Stochastic
II. BACKGROUND WORK Gradient Descent optimizer. The CNN architecture includes
convolutional and fully connected layers, with max pooling
An attempt to detect the drowsiness of the driver was made and flattening. The algorithm is tested in real-time using a
from the study of Maján Sesmero Alonso-Weber and Sanchis laptop webcam and sounds an alarm when the driver's eyes are
that used two deep learning models the CNNs and an LSTM. closed for a prolonged period.
Simultaneously YawDD dataset a Yawner Data set which
contains videos of them yawning was used to train the CNN The study was conducted by Bradley Barnhard, Kartik
model. At the same time, the LSTM model was trained with Prabhu, and Arun Seetharaman. They used image processing
the UTA-RLDD dataset, which includes the videos of drivers techniques to detect drowsiness in drivers whose duration of
engaged in an alert and drowsiness state while driving. This blink and heart rate variability were both used. The eye
combination was selected based on five-fold cross-validation, detection was implemented using OpenCV classifiers with
which was applied to these models during their evaluation about 10 videos per class from the AffectNet database.
process, as this combination produced the best accuracy as it Applied pattern recognition methods included independent
was used to train the data. As presented in Table 2, the LSTM component analysis and a chrominance-based method for heart
model achieved an accuracy level of 63% and a true positive rate variability. Relatively good results were obtained but were
rate of 70%. On the other hand, the authors mentioned that not able to reach the high accuracy illustrated in some other
although the test’s precision and true positive rate were low; it publications. This difference is probably because assessments
could still be considered a positive percentage. The supporting have to occur in highly contrived environments. Throughout,
data for the presented findings is publicly accessible in the the authors continuously tried and tested the model time after
UTA-RLDD and YawDD datasets. [2] time by varying a variety of parameters and doing performance
tests aiming at improving results.[5]
The conjoint hybrid driver drowsiness detection model was
built in accordance with the endeavor led by J.S. Bajaj, N. III. DESIGN AND METHODOLOGY
Kumar, R.K. Kaushal, H.L. Gururaj, F. Flammini, and R. In this application, driver drowsiness detection the
Natarajan. This model includes a camera and Galvanic Skin Raspberry Pi enjoys popularity because of its distinctive
Response sensor to obtain behavioral data which addresses the combination of small volume, superior computational power,
eye and mouth positions and physiological sensory data and its cost-effectiveness. Its compact nature facilitates easy
observed in skin conductance levels. A chosen different adaptation to multiple settings, including confined spaces such
evaluation using an additional image-based data set reported an as vehicles where space is revenue Although the Raspberry Pi
86% accuracy of the model. This technique was better than has a small footprint, it possesses sufficient computing abilities
subjective ratings, to the stated use of vehicle-based detective to perform sophisticated tasks like picture element image
controls, and to non-invasive physiological measures. processing and real-time data analysis, both of which are
Nevertheless, while acknowledging the shortcomings of the critical in determining which signs of drowsiness have
model, it is noteworthy to point out that future studies should
occurred.
use a broad data set rather than relying on this narrow data.[3]
Researchers and developers flock to this affordable tech,
A real-time driver drowsiness detection model of drivers finding cost-effective solutions for experiments demanding
was proposed by Parkar, Jajal and Brijbhan that was based on numerous units in their tests and deployments, paving the way
the application of a convolutional neural network. The for breakthroughs. The Computer modules feature such as

Copyright © IEEE–2024 ISBN: 979-8-3503-8354-6 865

Authorized licensed use limited to: INDIAN INST OF INFO TECH AND MANAGEMENT. Downloaded on April 09,2024 at 08:01:27 UTC from IEEE Xplore. Restrictions apply.
 2024 International Conference on Computing, Power, and Communication Technologies (IC2PCT)

accommodating for community support and open-source However, mentioning that the study will likely be ethical
which provides a platform for quick development and user- [7] if issues on video recordings and human subjects are
specific customization by allowing researchers and developers considered is crucial. Such guidelines require participants'
to customize the system to suit drowsiness detection consent and ensure the privacy and confidentiality of the
requirements. individuals. When these ethical issues are considered, the
study's values remain intact; thus, the participants' rights and
welfare are ensured.
Overall, this collection of driver drowsiness generated from
video recordings offers a multidimensional image across
different driving conditions. It is a practical reference tool for
researching and designing driver drowsiness detection
algorithms, thus contributing to safer roads and better driver
monitoring systems.
B. Methodology
1) Detecting the face of the user
The user's face was recognized using the widely known
Fig. 1. Block Diagram Of Proposed System Haar cascade technique. This method is characterized by its
speed and precision while being relatively less resource-
A. Dataset Description demanding, placing it well for device deployment such as
We have a dataset of video records about driving Raspberry Pi[8].
conditions leading to drowsiness detection for implementation
and research of drowsiness detection systems. The dataset uses 2) Eye detection
a capturing device that involves placing a camera such as a The Eye Aspect Ratio determines whether the eyes are
Raspberry Pi camera inside a vehicle to take a head posture, closed or open. Generally speaking, the EAR value is 0.339 for
regional eye movement, and expression of the driver. This open, while the EAR is 0.141 for closed eyes. As a result,
setup aims to load the situation on the road, which is why such whenever our system detects a face, it will calculate the EAR
a dataset is remarkably relevant and valuable for scientific and tell the user if the EAR drops beneath the established
research and environmental progress. threshold.

By nature, the videos represent specific travel scenarios The user will continue to get the notifications until the
and are measured in frames. Eight annotations were placed on driver is alerted. The Dlib library can identify the specific eye
datasets using criteria beforehand quantitatively describing the points in the pre-trained facial landmark modelThis model
morning and frames as either ‘drowsy’ or ‘alert’ that are estimates 68 coordinates (x, y) that map facial features on a
ground truth labels to certify and educate algorithms to predict person's face.[9]
the accurate level of drowsiness. Once we have obtained these landmark points, we can
The videos were recorded in a home, PC, or camcorder at a compute the EAR using the following formula:
specific resolution and format, including data for settings like |𝑃2−𝑃6|+|𝑃3−𝑃5|
frame rate, encoding format, and resolution (for example, EAR = [10]
2∗|𝑃1−𝑃4|
pixels width and height).
Consistency and standardization among the values in the Subsequently, the calculated EAR value is checked if it
dataset enable the researchers to center on the driver’s falls within the specified threshold, and based on that
behavior and drowsiness patterns. evaluation, the system alerts the user accordingly. This alert
will occur when the EAR value is not within the threshold.[11]
The data set is reasonably diverse, meaning it tries to
capture many real-life driving situations. This diversity entails 3) Yawn detection
different lighting and road conditions, as well as the The space between the user's upper and lower lips is
differences in the demographics of the drivers using the roads. assessed to detect a yawn. This space typically stays within a
Thus, the variability of the dataset results in a more true-to-life specific range during the speech, but it exceeds this range
and versatile dataset, where trained algorithms give optimal during a yawn. The landmarks for the lips are obtained, and
results according to every case in driving. the distance between these points is computed using DLIB's
facial landmark model. The midpoint between the upper and
This paper used video recordings of drivers to implement lower lips is measured to calculate this distance. The user is
and evaluate a driver drowsiness detection model. This alerted if the distance exceeds the predefined threshold.[12]
generated dataset was used to train and test this model to
determine the driver's drowsiness or alertness from captured 4) Drowsiness detection
facial features. It was built for various expressions, eye The system gathers data on the status of the driver's eyes
movements, and head posture [6]. and yawns, keeping track of the number of open eyes in each
frame and the occurrence of yawns. Drowsiness is assessed

Copyright © IEEE–2024 ISBN: 979-8-3503-8354-6 866

Authorized licensed use limited to: INDIAN INST OF INFO TECH AND MANAGEMENT. Downloaded on April 09,2024 at 08:01:27 UTC from IEEE Xplore. Restrictions apply.
 2024 International Conference on Computing, Power, and Communication Technologies (IC2PCT)

based on specific conditions, and if these conditions are met, the sleep detection algorithm. Taking an appropriate
the driver is considered drowsy. The system relies on classifiers combination of the EAR and MAR, the system augmented its
for face, yawn, and eye recognition to perform these accuracy in the precise diagnosis of drowsiness by aiding itself
detections. These classifiers can be found in the HAAR through several facial cues by considering both the EAR and
Classifier Cascade files bundled with OpenCV. MAR..
For each frame of the driver's facial image, the system 𝑇𝑃 + 𝑇𝑁
performs face detection, yawn detection, and open eye Accuracy =
TP + TN + FP + FN
detection. Each frame's count of open eyes is stored as the
variable "Eyes total." To trigger an alert, specific criteria The paper "Driver Drowsiness Detection by Applying
related to the eye and mouth aspect ratios need to be met: Deep Learning Techniques to Sequences of Images" [18] from
If the eye aspect ratio falls below 0.25 for 16 consecutive the journal Applied Sciences provides a detailed account of two
frames, it suggests that the person has likely had their eyes alternative approaches for drowsiness detection:
closed or partially closed for a significant duration. In such One approach utilizes RNN and CNN for the analysis of
cases, an alarm is activated.[13] sequential driver images. It attained a training accuracy of
The alert is immediately activated if the mouth aspect ratio about 71% and a testing accuracy of approximately 62%, with
exceeds 0.82, indicating that the person's mouth is open. a false positive rate of 40%.
Continuous frames are not required in this scenario.[14] Another approach leverages deep learning and fuzzy logic
These alerts serve as safety measures or warnings to draw to extract image features, followed by analysis using a fuzzy
attention to situations where predefined thresholds and inference system. During training, it achieved an accuracy of
conditions indicate potential drowsiness. approximately 69%, with a test accuracy reaching about 66%.
It is important to highlight that this approach notably decreased
5) Alarm buzzing the false positive rate to 7%.
Upon detecting drowsiness, the system activates an alarm
While the compared model[19] exhibits 96.42% accuracy, a
to awaken the driver. The alarm is an auditory stimulus to
critical examination of the underlying datasets reveals potential
ensure the driver remains alert and focused on the road.
limitations in generalizability. Their reliance on 2850 still
IV. RESULTS images, albeit with diverse angles, may not adequately capture
the complexities of real- world scenarios. The ongoing
A Haar cascade classifier and the shape Dlib package were COVID-19 pandemic, for instance, necessitates the
used to create the drowsiness detection system on a Raspberry consideration of facial occlusion through masks, a factor
Pi, which led to an accuracy of 87.8%. The system showed a potentially absent in their data.
good performance along with not a single false negative
image; one could observe all the drowsiness occurrences. In all In contrast, our approach prioritizes real-time video
3973 true negative frames were found, which points out the capture, enabling the acquisition of a richer and more dynamic
obtained positive results for the alertness recognition in the dataset. This dataset encompasses various variations, including
system.[15] skin tone, gender, and diverse facial angles under varying
lighting conditions. It incorporates the prevalent real-world
Using the Haar cascade classifier was a significant scenario of wearing a mask, enhancing the system's
advantage of the proposed rule-based system. Due to generalizability and practical applicability.
dependencies of their inference time, the Haar cascades are
ideal for such real time applications especially on devices that In conclusion, the Raspberry Pi-based drowsiness detection
do not have extensive resources such as the Raspberry Pi. This system using the Haar cascade classifier and the shape Dlib
effectiveness enabled the issuance of a prompt timely signal library performed remarkably well. The high accuracy rate and
when drivers were tired, being a measure necessary for the the absence of false negative frames demonstrate the system's
reactionary success and safety.[16] effectiveness in real- time drowsiness detection. Future work
could focus on reducing false-positive cases by improving the
The algorithm employed a EAR with a threshold of 0.25 rule-based system and addressing potential issues to improve
for eye drowsiness detection. In fact, this parametric used to its effectiveness over time. Furthermore, additional research
show the opening and closing eyes of an individual and thus, may be needed to enhance adaptability across different
did reflect the sleepiness in a human being successfully. If demographics and driving conditions for broader applicability.
frames of consecutive frames were set for 16 frames, the [20]
system detected drowsiness as long as the eyes have been
closed for this number of frames. It was with such an approach V. DISCUSSION
that the system was able to make a difference between real
The presented research focuses on the development of a
drowsiness incidences and reduce false positives. [17]
driver drowsiness detection system using a Raspberry Pi-based
Thus, the MAR was also incorporated within the system, setup, computer vision techniques, and machine learning
keeping an accepted value of 0.82. Furthermore, the MAR algorithms. The system is designed to increase road safety by
supplied more information that improved the accuracy rate of notifying the drivers before drowsiness causes them to be

Copyright © IEEE–2024 ISBN: 979-8-3503-8354-6 867

Authorized licensed use limited to: INDIAN INST OF INFO TECH AND MANAGEMENT. Downloaded on April 09,2024 at 08:01:27 UTC from IEEE Xplore. Restrictions apply.
 2024 International Conference on Computing, Power, and Communication Technologies (IC2PCT)

involved in accidents. This chapter summarizes the most

important results, implications, limitations and suggestions of
the drowsiness detection system. [21]
A. Key Findings
The implemented driver drowsiness detection system
demonstrated a high accuracy rate of 87.8% in real-time
detection of drowsiness indicators. In order to achieve this
accuracy, the integration of the Haar cascade algorithm for
classification and Dlib library for facial landmarks extraction
played a crucial role. The driver drowsiness system accurately
identified the instances without any false negatives. We can
contribute to the addition of existing knowledge in the field of
automotive safety by development and deployment of efficient
and accurate drowsiness detection system. The integration of Fig. 3. Drowsiness is detected when a person's eyes are closed.
these systems into vehicles and personal electronic devices can
reduce accidents, fatalities, and financial losses.
The system’s effectiveness was enhanced through the use
of EAR and MAR as indicators of drowsiness.[22] The EAR
reflected eye closure patterns characteristic of drowsiness,
while the MAR provided additional pointers for improved
detection. Such joint use of these metrics made the system to
make decisions about driver attentiveness based on informed
facts.
B. Implications
This research has noteworthy implications for enhancing
the safety of drivers by reducing the probability of accidents
caused by drowsy driving. The development of drowsiness Fig. 4. Drowsiness is detected when a person is yawning.
detection which is both effective as well as precise increases
automotive safety. The integration of these systems into the
vehicles can reduce the accidents, fatalities, and financial
losses associated with them.

Fig. 5. Drowsiness is detected when a person's eyes are closed with a mask
on.

Fig. 6. Drowsiness is detected when a person's eyes are closed when they
Fig. 2. Drowsiness is detected when a person's eyes are closed. wear their glasses.

Copyright © IEEE–2024 ISBN: 979-8-3503-8354-6 868

Authorized licensed use limited to: INDIAN INST OF INFO TECH AND MANAGEMENT. Downloaded on April 09,2024 at 08:01:27 UTC from IEEE Xplore. Restrictions apply.
 2024 International Conference on Computing, Power, and Communication Technologies (IC2PCT)

system based on convolutional neural network." SN Computer Science 1

(2020): 1-10.
[9] Girish, I., Kumar, A., Kumar, A., & M, A. (2020). Driver Fatigue
Detection. 2020 IEEE 17th India Council International Conference
(INDICON). doi:10.1109/indicon49873.2020.9342456
[10] Tiwari, S., Agarwal, S., Syafrullah, M., & Adiyarta, K. (2019,
September). Classification of physiological signals for emotion
recognition using IoT. In 2019 6th International Conference on Electrical
Engineering, Computer Science and Informatics (EECSI) (pp. 106-111).
IEEE.
[11] Srinivas Kolli, Peddarapu Rama Krishna and Parvathala Balakesava
Reddy, "A Novel Nlp And Machine Learning Based Text Extraction
Approach From Online News Feed", VOL. 16, NO. 6, MARCH
2021, ARPN Journal of Engineering and Applied Sciences, ISSN 1819-
6608, PP.679-685, 2021.
[12] Mishra, V. K., Kumar, S., & Gupta, R. K. (2014). Design and
Implementation of Image Fusion System. International Journal of
Fig. 7. Drowsiness is detected when a person's eyes are closed.
Computer Sciences and Engineering, 1, 182-186.
[13] Hashmi, Md Farukh, et al. "Drowsiness Detection System Using
VI. CONCLUSION Raspberry Pi and OpenCV." International Conference on Mobile
The application of such drowsiness detection systems can Computing and Sustainable Informatics: ICMCSI 2020. Springer
reduce the rate of road crashes, which are a major contributing International Publishing, 2021.
factor in the loss of lives, and also undo the economic cost of [14] Dr.S.Priyadarsini, Chahak Agarwal. D, Deshiya Narayan.M “Driver
Drowsiness Detection System Using Raspberry Pi” , 2019, accessed 20
such mishaps. Nevertheless, these systems have some August 2023, https://www.ijsdr.org/papers/IJSDR1903037.pdf
limitations confronting them that as false negatives and false [15] Tiwari, S., & Agarwal, S. (2021). A shrewd artificial neural network-
positives. A lot of research is aimed at the most accurate based hybrid model for pervasive stress detection of students using
algorithms and sensors. To summarize, detection systems galvanic skin response and electrocardiogram signals. Big Data, 9(6),
should help create safer roads if used in cars, and car 427-442.
manufacturers should introduce a viable drowsiness detection [16] Pinto, A., Bhasi, M., Bhalekar, D., Hegde, P., & Koolagudi, S. G.
system in the newer car models to achieve it. (2019). A Deep Learning Approach to Detect Drowsy Drivers in Real
Time. 2019 IEEE 16th India Council International Conference
(INDICON).doi:10.1109/indicon47234.2019. 9030305.
VII. FUTURE SCOPE
[17] Poursadeghiyan, Mohsen, et al. "Using image processing in the proposed
• It can be improvised to detect the drowsiness of the drowsiness detection system design." Iranian journal of public health
drivers when there are multiple people in a single 47.9 (2018): 1371.
frame. [18] Chirra, Venkata Rami Reddy, Srinivasulu Reddy Uyyala, and Venkata
Krishna Kishore Kolli. "Deep CNN: A Machine Learning Approach for
• Additional constraints can be added to handle the case Driver Drowsiness Detection Based on Eye State." Rev. d'Intelligence
when the driver is awake and yawning. Artif. 33.6 (2019): 461-466.
[19] Magán, Elena, et al. "Driver drowsiness detection by applying deep
REFERENCES learning techniques to sequences of images." Applied Sciences 12.3
(2022): 1145.
[1] Magán, Elena, et al. "Driver drowsiness detection by applying deep
learning techniques to sequences of images." Applied Sciences 12.3 [20] Jabbar, R., Al-Khalifa, K., Kharbeche, M., Alhajyaseen, W., Jafari, M.,
(2022): 1145. & Jiang, S. (2018). Real-time Driver Drowsiness Detection for Android
Application Using Deep Neural Networks Techniques.
[2] Kumar, Vijay, Shivam Sharma, and Ranjeet. "Driver drowsiness Procedia Computer Science, 130, 400–407.
detection using modified deep learning architecture." Evolutionary doi:10.1016/j.procs.2018.04.060.
Intelligence (2022): 1-10.
[21] Prabhu, Kartik, Bradley Barnhart, and Arun Seetharaman. "Drowsiness
[3] Tarafder, Sreeza, et al. "Drowsiness detection using ocular indices Detection using Contactless Heart Rate Monitoring and Eye Tracking."
from EEG signal." Sensors 22.13 (2022): 4764.
[22] Tiwari, S., Chandra, R., & Agarwal, S. (2022, November). An optimized
[4] Bajaj, Jaspreet Singh, et al. "System and method for driver drowsiness hybrid solution for IoT based lifestyle disease classification using stress
detection using behavioral and sensor-based physiological measures." data. In International Conference on Neural Information Processing (pp.
Sensors 23.3 (2023): 433-445). Singapore: Springer Nature Singapore
[5] Magán, Elena, et al. "Driver drowsiness detection by applying deep
learning techniques to sequences of images." Applied Sciences 12.3
(2022): 1145.
[6] Jain, Mahek, Bhavya Bhagerathi, and C. N. Sowmyarani. "Real-time
driver drowsiness detection using computer vision." International Journal
of Engineering and Advanced Technology (IJEAT) ISSN (2021): 2249-
8958.
[7] V B Navya Kiran, Raksha R, Anisoor Rahman, Varsha K N, Dr.
Nagamani N P, 2020, Driver Drowsiness Detection, INTERNATional
Journal Of Engineering Research & Technology (IJERT) NCAIT – 2020.
[8] Hashemi, Maryam, Alireza Mirrashid, and Aliasghar Beheshti Shirazi.
"Driver safety development: Real-time driver drowsiness detection

Copyright © IEEE–2024 ISBN: 979-8-3503-8354-6 869

Authorized licensed use limited to: INDIAN INST OF INFO TECH AND MANAGEMENT. Downloaded on April 09,2024 at 08:01:27 UTC from IEEE Xplore. Restrictions apply.

View publication stats