The prototype of detecting risky behavior of drowsiness from video

Abstract

This article presents the prototype which was able to detect the behaviours that were at risk of dozing off by calculating eye ratios by applying Euclidean distance techniques and applying the calculated eye ratios to predict doze off. The functions of the prototype were as following: 1.) Face detection from video footage 2.) Dozing off detection from drivers’ risky behaviours including 2.1) Open eyes, 2.2) Normal Blink, 2.3) Blink frequency, 2.4) Drowsy manner, 2.5) Eye closure manner, 3.) Sound alerts to drivers which notified when drivers started to doze off while they were driving. The alerts were created to increase drivers’ security and reduce road accidents which result from dozing off and the results obtained from the detection of sleep disorders are as follows 1) Open eyes has Recall 1 2) Normal Blink has Recall 0.1 3) Blink frequency has Recall 0.4 4) Drowsy manner has Recall 0 5) Eye closure manner has Recall 0.7667.