Skip to content
SDGs Center UDINUS

Evidence record · UDINUS · SDGs

Microsleep detection on drowsy-driver using convolutional neural networks (CNN)

Rektorat UDINUS 2024 Published 2026-10-30

Evidence record · UDINUS · SDGs

Unit
Rektorat UDINUS
Period
2024
Goals
Record
15bcffd3-62a2-4f1c-b1e6-e6e1aa6047cf
Published
2026-10-30

Universitas Dian Nuswantoro conducted research on microsleep detection among drowsy drivers using convolutional neural networks (CNN), with Ilmadina, Hepatika Zidny; Muhammad Naufal; and Dyah Apriliani identified as activity participants. The activity addresses the condition in which a person may fall asleep while moving without realizing it. The source describes microsleep as an event in which an individual sleeps for four seconds while travelling at 100 kilometres per hour, resulting in 111 metres of driving in an unconscious state.

The research is situated in the context of traffic accidents and human factors. Data from the Traffic Directorate of the Polda Metro Jaya are reported as 7,565 accident cases during 2020, including 1,565 seriously injured people and 559 deaths. The source states that 61% of the cases were caused by human error, including sleepiness, lack of focus, and fatigue. Vehicle factors accounted for 9%, while infrastructure and environmental factors accounted for 30%.

UDINUS examined microsleep detection through eye blinks. The described approaches include eye aspect ratio with manual thresholding, based on setting the minimum duration of eye closing or blinking, as well as Support Vector Machine and Convolutional Neural Network methods. The best model reported in the source was ResNet50, which achieved 99% accuracy. The system was implemented with a Haar cascade classifier to detect closing eyes as a sign of microsleep. The research is expected to help reduce and prevent accidents caused by microsleep based on a person's eye blinking. This activity is presented in relation to the stated goal of health research and records UDINUS's institutional engagement with the issue.

Reference https://www.scopus.com/pages/publications/85190551364?origin=resultslist