An Accurate Detection of Drowsiness Using a Graph-Based Neural Network

According to the findings of this research, a low-cost solution for identifying driver drowsiness might be the use of microsleep patterns. In contrast to the standard method, we collected pictures by putting the camera on the extreme left side of the driver and suggested two algorithms that allow re...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:2023 International Conference on Applied Intelligence and Sustainable Computing (ICAISC) s. 1 - 6
Hlavní autoři: Agarwal, Varsha, Rajneesh
Médium: Konferenční příspěvek
Jazyk:angličtina
Vydáno: IEEE 16.06.2023
Témata:
Cameras
Drowsiness
Eye State Detection
Graph Neural Network
Graph neural networks
Lighting
Measurement
Real-time systems
Sleep
Whale optimization algorithms
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:According to the findings of this research, a low-cost solution for identifying driver drowsiness might be the use of microsleep patterns. In contrast to the standard method, we collected pictures by putting the camera on the extreme left side of the driver and suggested two algorithms that allow reliable face and eye detections regardless of whether the driver is gazing straight at the camera or has closed his or her eyes. This was accomplished by obtaining images by positioning the camera on the driver's extreme left side. It has been recommended that a Graph Neural Network, often known as a GNN, be used in order to determine whether or not the right eye is open. The whale optimization (WO) approach was developed in order to determine which qualities are considered to be the most desirable. Eye states are utilized to identify patterns of microsleep, and an alert is then transmitted to the driver of the vehicle when one is detected. Our data set consisted of a large number of male and female participants, each of whom had unique physical traits and had been exposed to a range of lighting conditions. The suggested method achieves an accuracy of face detection of 99.9% and an accuracy of eye detection of 98.7%. The overall accuracy and precision levels of WO-GNN are, on average, 96.4 and 95.4 percent, respectively, across the board for all subject areas.
DOI:10.1109/ICAISC58445.2023.10201129