Scribd
Upload
30 views2 pages

Drowsiness Detection Abstract Revised

The document presents a mobile application for real-time driver drowsiness detection using AI and a smartphone's front-facing camera, addressing the issue of drowsy driving which contributes to a significant number of accidents. The system classifies driver states into Alert, Low Vigilance, and Drowsy, with a classification accuracy of 73.2% under optimal conditions, and employs tiered responses to alert the driver. Privacy is prioritized through on-device processing, and future enhancements may include differential privacy and federated learning for better personalization.

Uploaded by

thouqanlujain
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
30 views2 pages

Drowsiness Detection Abstract Revised

The document presents a mobile application for real-time driver drowsiness detection using AI and a smartphone's front-facing camera, addressing the issue of drowsy driving which contributes to a significant number of accidents. The system classifies driver states into Alert, Low Vigilance, and Drowsy, with a classification accuracy of 73.2% under optimal conditions, and employs tiered responses to alert the driver. Privacy is prioritized through on-device processing, and future enhancements may include differential privacy and federated learning for better personalization.

Uploaded by

thouqanlujain
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 2

SRC25

Real-Time Driver Drowsiness Detection System

Lujain Thouqan, Hala Kahwajy, Asma Luai, Judi Chbli


Supervised by Dr. Salam Fraihat
Ajman University, Ajman, UAE

ABSTRACT

In recent years, drowsiness detection systems have gained significant attention due to the rising number of
fatal accidents caused by drowsy drivers. Fatigue is frequently cited as a major contributing factor, present
in approximately 20% of all crashes and 16% of near-crashes [1]. To address this critical issue, we propose
an innovative mobile application that leverages artificial intelligence (AI) to monitor and assess a driver’s
alertness in real-time using only the smartphone’s front-facing camera—without requiring specialized
hardware. The system employs a Convolutional Neural Network (CNN) trained on a dataset of 15,000
annotated images sourced from public repositories (e.g., NTHU-DDD) and custom-simulated driving
scenarios. It classifies driver states into three categories: Alert, Low
Vigilance, and Drowsy. Empirical testing yielded 73.2% classification
accuracy under optimal lighting, dropping to 70.1% in low-light
conditions. To reduce false positives caused by natural behavior like
yawning or glancing, the system incorporates temporal analysis using
eye closure duration and head pose estimation via MediaPipe.The
application initiates tiered responses: for Low Vigilance, it plays
stimulating music via Bluetooth; for Drowsiness, it escalates to loud
alarms and engages the driver with conversational prompts using
Whisper-AI. If the driver remains unresponsive, an emergency message
is sent to pre-registered contacts. Privacy and security are prioritized
through on-device data processing—no facial images or audio
recordings are stored or transmitted externally. All sensitive operations
occur locally, and the system employs memory-safe architecture with
transient in-memory buffering to ensure compliance with user data
protection standards. Future versions may integrate differential privacy
techniques and federated learning to enhance personalized model
adaptation without compromising data ownership.In terms of energy
efficiency, the app was observed to consume moderate battery during continuous use. While no advanced
optimizations were implemented, strategies like adjusting processing frequency and using hardware
acceleration are considered for future improvements to reduce power consumption.
The solution is built using TensorFlow, OpenCV, Flutter, and MediaPipe, and was tested across mid-range
Android and iOS devices. Compared to existing solutions like Eye-Tracker and DROZY, our system
achieves a 7% improvement in latency, greater cross-platform compatibility, and independence from
proprietary sensors or wearables. Limitations include dependence on stable camera positioning and
reduced reliability in cases of facial occlusion. Overall, the proposed system offers a practical, privacy-
aware, and hardware-efficient solution to drowsiness detection, aiming to reduce road accidents by
enabling real-time, accessible driver monitoring.

Zayed University April 9-10, 2025, Dubai, UAE


16th Student Research Conference on Applied Computing
SRC25
Reference
[1] EHS Today, "Wake up and drive: Fatigue causes 20 percent of crashes," Apr. 15, 2013. [Online]. Available:
https://www.ehstoday.com/safety/article/21917988/wake-up-and-drive-fatigue-causes-20-percent-of-crashes . [Accessed: Mar. 25, 2025].

Zayed University Figure 1- General Architecture of the proposed solution April 9-10, 2025, Dubai, UAE
16th Student Research Conference on Applied Computing

You might also like