Advanced forecasting of driver drowsiness events: Non-intrusive data and multimodal BiLSTM-based modeling

Key Points

• Driver drowsiness forecasting achieved a significant accuracy improvement of 15% over traditional methods.
• Key metrics include detection accuracy and response time in real-world driving scenarios over a six-month period.
• This observational analysis utilized bi-directional long short-term memory (BiLSTM) modeling with various non-intrusive data inputs.
• These findings highlight the need for advanced systems to prevent accidents caused by driver fatigue.