Open AccessDriver’s Drowsiness Detection Based on Facial Multi-Feature Fusion
Author(s) -
V. Venkata Sai Vardhan,
N K Sagar Reddy,
K. Surya,
J. Uday Kiran,
Ashwani Kumar
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1998/1/012034
Subject(s) - computer science , thresholding , artificial intelligence , histogram , histogram of oriented gradients , support vector machine , feature (linguistics) , computer vision , frame rate , local binary patterns , consistency (knowledge bases) , face detection , face (sociological concept) , frame (networking) , pattern recognition (psychology) , facial recognition system , image (mathematics) , telecommunications , social science , linguistics , philosophy , sociology
Among the leading causes of traffic accidents and deaths is drowsy driving. As a result, detecting and indicating driver drowsiness is an important research subject. The majority of existing approaches are based on automobiles, behavioral-based, or physiologically based. A few approaches are invasive and so the driver been interrupted, while others necessitate the use of pricey sensors and data processing. As a result, real time driver sleepiness detection system with adequate consistency and minimized cost has been established in this work. A webcam captures the video in the proposed system, and image processing techniques are used. Histogram of oriented gradients (HOG) which is available in Dlib toolkit, is used to recognize the driver’s face in every frame. The landmarks on the face are identified by State Vector Machine, and the eye aspect ratio, mouth aspect ratio, are computed, and drowsiness is recognized using created adaptive thresholding based on their values. Offline implementations of machine learning techniques were also been made. In terms of accuracy and speed, we show that our algorithm outperforms existing fatigue methods. In Support, a sensitivity of 95.28 percent and a response rate of 100 percent were attained in Support Vector Machine.