A Multi-Stage, Multi-Feature Machine Learning Approach to Detect Driver Sleepiness in Naturalistic Road Driving Conditions

Driver fatigue is a contributing factor in about 20% of all fatal road crashes worldwide. Countermeasures are urgently needed and one the most promising currently available approaches for that in-vehicle systems driver detection. The main objective this paper to present video-based sleepiness detection system set up as two-stage model with (1) generic deep feature extraction module combined (2) personalised module. approach was designed evaluated using data from 13 drivers, collected during naturalistic driving conditions on motorway Sweden. Each performed 90-minute session daytime (low condition) night-time (high condition). outputs continuous output representing Karolinska Sleepiness Scale (KSS) scale 1–9 or binary decision alert (defined KSS 1–6) sleepy 7–9). Continuous modelling resulted mean absolute error (MAE) 0.54 units. Binary classification showed an accuracy 92% (sensitivity = 91.7%, specificity 92.3%, F1 score 90.4%). Without personalisation, corresponding 72%, while standard PERCLOS-based baseline method reached 68% same dataset. developed real-time can be used management sleepiness/fatigue by detecting precursors severe fatigue, ultimately reduce sleepiness-related alerting drivers before high levels reached.