Improved Driver Fatigue Detection using Blinking and Yawning Rate

Improved Driver Fatigue Detection using Blinking and Yawning Rate

Srijan Pratap Srivastava

Student, National Institute of Technology Kurukshetra, Haryana

Corresponding Author: Srijan Pratap Srivastava

ABSTRACT: Fatigue-induced driving is a prevalent cause of road accidents. Hence, multiple authors have proposed various methods to detect driver drowsiness at an early stage to prevent such occurrences. This scholarly piece introduces an efficient approach to detect exhaustion by examining the blinking and yawning rate of driver. The partitioning process generates a representation, which assists in identifying the pupil and mouth. To ensure greater accuracy and applicability in determining eye openness, a composite network comprising a classification network and an evaluation network is developed. The partitioning network employs a lightweight Linknet architecture to categorize pixels in eye and mouth images, which enables the accurate extraction of the pupil and mouth. The decision network utilizes the representation extracted to assess the extent of eye and mouth openness. Finally, the model is tested on a larger dataset, and it achieves state-of-the-art accuracy of 92%.

KEYWORDS: LinkNet, fatigue detection, facial features, PERCLOS(percentage of eyelid closure ),EOR(Eye opening ratio), MOR(mouth opening ratio)