PENENTUAN KRITERIA DAN PENGEMBANGAN ALGORITMA ALAT DETEKSI KANTUK BERBASISKAN PARAMETER OKULOMOTOR

• 主要作者: Florence Andersen, Hareliz
• 格式: Final Project
• 語言: Indonesia
• 在線閱讀: https://digilib.itb.ac.id/gdl/view/83037
• 機構: Institut Teknologi Bandung

Railway accidents remain a critical issue in Indonesia, with 156 incidents reported between 2015 and 2022, resulting in significant material damage and serious threats to human safety. This alarming statistic has raised concerns among the management of PT KAI, Indonesia's state-owned railway company. Research indicates that sleepiness negatively impacts attention, memory, alertness, reaction time, and coordination, thereby increasing the likelihood of railway accidents. One effective metric for assessing sleepiness is through oculomotor parameters. However, there is a lack of specific criteria for sleepiness threshold levels among railway operators in Indonesia. This study aims to establish sleepiness threshold values and develop a sleepiness detection algorithm based on oculomotor parameters. The research objectives were achieved through laboratory experiments involving 8 participants whose physical characteristics resemble those of train drivers. These participants operated a train simulator for 60 minutes after normal sleep (8 hours) and under sleep-deprived conditions (no sleep for 24 hours). Eye tracker technology was utilized to objectively measure sleepiness based on blink duration and frequency metrics, supplemented by subjective assessments using the Karolinska Sleepiness Scale (KSS). The findings of this study indicate that sleep deprivation significantly affects subjective sleepiness scores, blink frequency, and blink duration. Among all oculomotor indicators, blink duration demonstrated the highest accuracy and sensitivity in detecting fatigue. The results revealed threshold values for classifying low sleepiness or fit conditions, with a blink frequency per hour of less than 1071 and a blink duration of less than 164 ms; moderate sleepiness with a blink frequency between 1071 and 1262 and a blink duration between 164 ms and 280 ms; and severe sleepiness with a blink frequency per hour of more than 1262 and a blink duration of more than 280 ms. These threshold values were utilized to design an algorithm aimed at identifying the number of blinks and calculating the average blink duration, subsequently classifying the level of sleepiness of train drivers into these three categories. The sleepiness detection algorithm is expected to accurately detect the sleepiness condition of train drivers, thereby assessing their fitness for duty for the next shift. The findings are anticipated to contribute to enhancing railway safety in Indonesia. Future research should consider increasing the number of participants and involving actual train drivers to improve the statistical robustness and representativeness of the study results.