CLASSIFICATION OF FATIGUE FROM DRIVING BASED ON ELECTROENCEPHALOGRAPHY (EEG) CHANGES CHARACTERISTICS : STUDY ON DRIVING SIMULATION

• Main Author: Zuraida, Rida
• Format: Dissertations
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/44130
• Institution: Institut Teknologi Bandung

This research is motivated by the relationship between fatigue and the risk of accidents. In transportation sectors, the high number of road accidents in the world including in Indonesia affected by driver’s fatigue. This linkage puts fatigue management as a very important part of the works system in the transportation sector to minimize the risk caused by fatigue. Fatigue evaluation is part of fatigue management. The evaluation is currently not only utilizing subjective measurement tools such as questionnaires, but also objective measuring instruments such as Electroencephalography (EEG) as a tool that record brain waves. The brain wave can represent changes in fatigue state. Fatigue research in driving activity based on EEG still has gaps as follow: changes in EEG parameters that indicate fatigue, EEG parameters that are most sensitive for detecting fatigue, origin of brain areas for the EEG parameters evaluated, and classification of fatigue based on EEG.. This study aims to assess fatigue in driving activity based on changes in EEG signals. Through the design of fatigue classification, it is hoped that EEG parameters the area of the brain that is most sensitive to detection, and the fatigue state boundary can obtained. Furthermore, the results of the study are expected to be useful for the detection of fatigue that is simpler by utilizing certain parameters and brain areas only. To achieve these objectives, experiments were conducted using a driving simulator to obtain observable fatigue conditions. The factors considered affect fatigue were driving duration (time on task), lack of sleep, and time of day. The study that assess time on task factors and sleep deprivation was studied in Study-1. For this reason, a 5 hour driving experiment was designed in the morning which was divided into 2 sessions of 2.5 hours each with 30 minutes of rest between and afterwards. Experiments conducted under sufficient and lack of sleep. Study-1 is an experiment with within subject design and same participant following two conditions of sleep adequacy. This study also examined changes in EEG signals that are affected by time of day through comparison a morning and evening driving tasks. For this reason, a driving experiment at night (study-2) was designed using a 2.5 hour simulator followed by a 30 minute break. The results were compared with the results of the experimental driving the first 2.5 hours of adequate sleep in Study-1. Participants in the morning and evening consisted of 11 people. All participants had experience as commercial drivers, and were accustomed to driving long distances or long durations (more than 2.5 hours) and having a Driving License (SIM) A. Fatigue in driving activity is generally characterized by an increase in the average power of alpha, beta and theta waves and the ratio of brain waves. The increases were successively: alpha 7.2% and beta at 1.96% in the morning study when sleep deprived and theta at 9.1% under sufficient sleep and 9.3% under lack of sleep. For the results of the study at night obtained an average increase of 9% in alpha, and 12.5% in theta and beta. An increase in fatigue score was obtained by 210% and 300% (sleepiness and fatigue) between the beginning and end of the study session in the morning and evening, and by 62.5% between study-1 under sufficient and lack of sleep. The effect of sleep factor characterized by changes in alpha wave power (61.62%) and theta (17.9%) compared to beta waves (1.96%). A subjective indicators resulte showed that fatigue influenced by sleep factors, time on task and time of day (p-value <0.01). The changes in reaction time indicator were more influenced by the time on task and time of day (p-value <0.05). Changes in EEG parameters approached by subjective sleepiness and fatigue showed a tendency to rise sharply in conditions between alert and moderate fatigue (3-7%) and decrease when conditions turn out to be severe tired (1-% - 7%). The proposed an EEG-based classification of fatigue was resulting that fatigue can be classified to moderate and severe fatigue. Using the Support Vector Machine (SVM) as a method, it resulted in sensitivity 86,7%, the specificity of 78.9% and accuracy of 84.5% for moderate fatigue; and sensitivity 80,1%, the specificity of 79.5%, and accuracy of 80.0% for severe fatigue. Using the subjective sleepiness and fatigue as the approach on an EEG-based classification, alpha and theta waves found as the sensitive EEG parameters for fatigue detection. Those brainwaves suggested being recorded from Fontal and Occipital area. Through the subjective indicators approach and the use of the Receiver Operating Characteristics (ROC), the classification resulted in the cut-off values of the alert, mild fatigue, and severe fatigue parameters of the EEG, KSS, and F-VAS. These results can be used as a basis of fatigue detection from the driving task. The findings of this study can be utilized to evaluate current driver’s duty hours and confirmed that it is important for drivers to have adequate sleep to preventing fatigue. Obtained the characteristics of EEG changes that indicate fatigue can be utilized in the development of driver fatigue interference technology. This result can be used as a basis for future similar research. Hopefully an EEG simple fatigue detection could minimize the risk that are affected by fatigue.