Automatic detection of drowsiness in EEG records based on time analysis

Driver drowsiness has been identified as a root cause of more than 20% of crashes resulting in injuries and deaths globally. Drowsiness leads to deterioration in driver's performance and vehicle-handling abilities. This research aims at using time analysis techniques to automatically detect the level of drowsiness in EEG records. Noise from the EEG signals were removed by subjecting it to band-pass filtering with cut-off frequency 0.5 Hz to 60 Hz and then segmented into 5 second intervals. 1 EEG channel was used to compute 3 features s to differentiate the awareness and drowsiness stages. This method gets up to 85.7% of alertness and drowsiness correct detections rates. Both the stages can be distinguished based on the parameters and acquired results. Real time outputs can be obtained making it easily accessible. An automatic drowsiness detection system in vehicles can be employed using these parameters thereby decreasing the risk posed by drowsy driving. The algorithm developed here has been tested on samples obtained from The MIT-BIH Polysonographic Database on physionet.org.