IOT based Driver Drowsiness and Alcohol/Smoke Detection System

Using Deep Learning Algorithm

Prof. S. Saraswat, Ashutosh Bhoite, Aparna Gavhane, Aditya Sasne, Pradhunya Shendge, Sasmit
Waghmare

Dept. of Information Technology, PDEA’s College of Engineering, Pune, India

ABSTRACT —
stage of drivers with the aid of measuring their
Due to extended driving times and exhaustion, driver peculiar behaviours.[3] We propose a manifold
drowsiness is one of the main causes of traffic accidents, learning algorithm that can learn a mapping from
especially for drivers of large vehicles (like buses and a low-resolution eye image (e.g., 32 ×24 pixels)
heavy trucks). Driver drowsiness can also be found due to to a 1-dimensional continuous level of eye-
consumption of alcohol openness. There are two advantages of this
approach. First, there is no need to detect eye
Under operational circumstances. In this study, we offer an feature points for symptom estimation which often
easy-to-deploy and flexible vision-based tiredness fails on low-resolution face images. Second, it
detection system for bus driver monitoring in big vehicles, avoids classifying the ambiguous partially closed
including detection, face detection, eye detection, eye eyes into open or close states for binary state (0/1)
openness estimation, and alcohol. The following are the classification. This improves the accuracy of
primary novel techniques: 1) A spectral regression- identifying the “drowsy” state between “normal”
based method to estimate the continuous degree of and “sleepy” states. To obtain an accurate and
eye openness; 2) A fusion algorithm that uses adaptive robust estimate of eye openness, a novel fusion
integration on the multi-model detections of both eyes to algorithm was proposed which adaptively
estimate the eye state. 3) An alcohol detection sensor that integrates the results of eye openness estimations
will detect the presence of alcohol molecules. When a on the multi-model eye detections for both eyes.
camera with an oblique viewing angle to the driver's face Based on the innovative techniques, the system
is employed for driving status monitoring, the achieves robust performance on challenging
experimental results demonstrate the benefits of the system scenarios where the existing approaches often fail.
in terms of accuracy and robustness for difficult conditions
2. ARCHITECTURE
Key Words: Arduino Uno, Camera, Machine
learning, Video Analysis, MQ 3 sensor.

1. INTRODUCTION

Driver’s driving is a major problem. No one knows the

exact movement when the body goes into state of asleep.
This makes the driver less able to pay attention to the road.
It affects the driver’s ability to make good decision.

Behavioural Measures: A drowsy man or woman displays

several characteristic facial actions, including fast and
regular blinking, nodding or swinging their head, and
common yawning. Computerized, non-intrusive,
behavioural approaches are extensively used for
determining the drowsiness
The architecture of an IoT-based driver drowsiness and drowsy face is detected, the driver's
alcohol detection system employing deep learning is a drowsiness is identified when the graphically
sophisticated integration of hardware and software projected EAR value is less than 0.25. The
components designed to enhance road safety. At its core, eyelids' regular motions cause the EAR levels
the system relies on a network of sensors strategically to fluctuate. When the driver exhibits signs of
placed within the vehicle to capture real-time data. For fatigue, an email notification is sent to the
drowsiness detection, facial recognition cameras and owner or relevant authority and the driver is
sensors monitor the driver’s facial expression, eye alerted with a repeated voice message. In this
movements, and blinking patterns. Concurrently, an case, a speech speaker is used in place of a
alcohol detection module utilized breathalyzer sensors or buzzer to increase awareness; in the event that
touch-based alcohol sensors to measure the driver’s blood it malfunctions, the owner will issue any
alcohol content. The collected data is then fed into a deep warnings after getting the message from its
learning model, typically a convolutional neural network mail as represented.
(CNN) for facial analysis and a recurrent neural network
(RNN) for time-series data like breathalyzer readings. The
trained module evaluates the input data, identifying signs
of drowsiness or alcohol impairment. The system can issue
timely alerts or take preventive measures, such as
activating alarm or notifying authorities, to mitigate
potential risks and enhance overall road safety. The
interconnected nature of this IoT architecture ensured
seamless communication between the various
components, providing a comprehensive and reliable
solution for monitoring driver well-being.

3. Methodology Fig 1: EAR

On the top-left, we have a fully open eye and the

eye's facial landmarks plotted. Then on the top-
A. Software right, we have an eye that is closed. The bottom
Overview: then plots the eye-aspect ratio over time. As we can
There are six main steps in the process. First, a head- see, the eye aspect ratio is constant (indicating that
shoulder detector is applied to detect the presence of a the eye is open), then rapidly drops to close to zero,
driver and locate roughly the position of the driver’s head. then increases again, indicating a blink has taken
Then, two models of face detectors are used to detect a place.
front-view face or an oblique-view face within the region
In our drowsiness detector case, we’ll be
of the head. Third, two eye detection methods are employed monitoring the eye aspect ratio to see if the value
to locate the potential eye positions and scales in the image. falls but does not increase again, thus implying that
In the fourth step, our proposed eye openness estimation the driver/user has closed their eyes.
method is applied to the located eyes by the two eye
detectors. Next, a fusion operation is proposed to obtain an B. Hardware
accurate and robust estimate of the driver’s eye openness
based on adaptive integration on multi-model eye Raspberry Pi
detections for both eyes. The score of driver’s fatigue, i.e.,
PERCLOS, is computed on the recent records of eye Raspberry Pi 3 model B is a Master card estimated
openness over a specified period. singleboard PC (SBC) planned in the UK. This
turns into the primary model to include remote
Detection of Drowsiness availability. Wi-Fi 802.11n and Bluetooth 4.1.this
can likewise boot from a USB. The Pi 3 model B+
Open eyes are indicated by an EAR value greater than and pi 3
0.25. This test demonstrates that the driver has not model A+ are presented with 802.11ac and
succumbed to the visually depicted tiredness and that Bluetooth 4.2 in 2018 these interfaces, coming up
the face is not identified as drowsy. Similar to how a next are available on Pi 3 Model B+:Gigabit
Ethernet, simple sound/video 3.5mm jack,4x USB 2.0, This module is made using Alcohol Gas Sensor
HDMI, Camera sequential interface(CSI), Display MQ3. It is a low cost semiconductor sensor which
sequential interface(DSI) and a 40-stick GPIO header pins can detect the presence of alcohol gases at
can associate with a camera. concentrations from 0.05 mg/L to 10 mg/L. The
sensitive material used for this sensor is SnO2,
whose conductivity is lower in clean air. It’s
conductivity increases as the concentration of
alcohol gases increases. It has high sensitivity to
alcohol and has a good resistance to disturbances
due to smoke, vapor and gasoline. This module
provides both digital and analog outputs. MQ3
alcohol sensor module can be easily interfaced
with Microcontrollers, Arduino Boards,
Raspberry Pi etc.

Fig 2: Raspberry pi 3 Model

Features of Raspberry pi

Low Cost: Raspberry Pi offers a cost-effective computing

solution, making it accessible for educational purposes,
hobbyists, and various projects.

Compact size: The small form factor of the Raspberry Pi

allows it to fit into compact spaces, making it suitable for
embedded systems and portable projects.
Fig 5: MQ3 Alcohol sensor
GPIO Pins: General-Purpose Input/Output (GPIO) pins
This alcohol sensor is suitable for detecting alcohol
enable interaction with the physical world. These pins
concentration on your breath, just like your
allow users to connect the Raspberry Pi to sensors, common breathalyzer. It has a high sensitivity and
actuators, and other hardware components. fast response time. Sensor provides an analog
resistive output based on alcohol concentration.
Processing Power: Despite its small size, Raspberry Pi The drive circuit is very simple, all it needs is one
models offer varying levels of processing power, capable of resistor. A simple interface could be a 0-3.3V
handling a range of tasks, from simple projects to more ADC.
complex applications.

I/O Ports: Pi boards come equipped with various ports,

including USB ports, HDMI ports, audio jack, Ethernet
ports, and more. These ports enhance connectivity and
make the Raspberry Pi adaptable to different peripherals.

Wireless Connectivity: Many Raspberry Pi model include

built-in wireless connectivity options such as Wi-Fi and
Bluetooth, providing flexibility for communication with
other devices.
Fig : MQ3 Board Schematic
Camera and Display Support: Raspberry Pi boards have
interfaces for connecting a camera module and displays,
allowing users to create projects involving image and video
processing.

MQ3 sensor
 4. Conclusion Paper on Vehicle Drowsiness Detection System”
Dept. of Computer Engineering, HSBPVt’s
In conclusion, the development of an IoT-based Drowsiness Parikrama Gol, COE, Kashti (JETIR) ISSN-2349-
and Alcohol/Smoke Detection System using a Deep 5162, Vol. 7, issue 4, April 2020
Learning algorithm holds great potential for enhancing road
[4] S. N. Sawalkar, Ankita Bhoyar, “A Review
safety and preventing accidents caused by impaired
Paper on Visual Analysis of Eye State using Image
driving. The integration of advanced technologies such as
Processing for Driver Fatigue Detection”
IoT devices, sensors, and deep learning models allows for
Computer Science and Engineering, Sipna
real-time monitoring and timely intervention. This system
C.O.E.T., (IRJET) vol. 6, issue 4, April 2019
addresses critical issues associated with driver drowsiness,
alcohol consumption, and smoke presence, contributing to [5] Bappaditya Mandal, Liyuan Li, Gang Sam
a safer driving environment. Wang and Jie Lin, “Towards Detection of Bus
Driver Fatigue Base on Robust Visual Analysis of
The IoT-based Driver Drowsiness and Alcohol/Smoke Eye State”, IEEE Transactions on Intelligent
Detection System represents a significant advancement in Transportation Systems, vol. 18, No. 3, March
leveraging technology for road safety. By combining IoT, 2017
sensors, and deep learning, the system contributes to the
overarching goal of reducing accidents and promoting [6] Anupriya R., Binishamol Joby, Liya Joy,
responsible driving behavior. As technology continues to Neetha K. Nataraj, “Driver Drowsiness Detection
evolve, there is potential for further enhancements and And Alert System Based On Behaviour Measures”
widespread adoption, ultimately making roadways safer for Dept. of C.S., Adi Shankara Institute of
all users. Engineering Technology, Kalady, (JRJET) vol. 7,
issue 6, June 2020.
5. References [7] Ch Venkata Rami Reddy, U. Srinivasulu Reddy,
[1] Jang Woon Baek, Byung-Gil Han, Kwang-Ju Kim, Yun- D Mahesh Babu, “An Automatic Driver
Su Chung, SooIn Lee. Real-time Drowsiness Detection Drowsiness Detection System using DWT and
Algorithm for Driver State Monitoring System. (2018) RBFNN”, International Journal of Recent
Technology and Engineering (IJRTE) ISSN: 2277-
[2] K. Srijayathi and M. Vedachary, “Implementation of 3878, Volume-7, Issue-5S4, February 2019.
The Driver Drowsiness Detection System,” International
Journal of Science, Engineering and Technology Research
(JSETR), vol. 2, pp. 1751-1754, September 2013.

[3] Prof. Hiranwale SachinMiss. Tagad Sonali N., Miss

Randhave Mansi R., Mr. Khan Mohsin M. “Implementation