From Alert to Doze off: Insights from ECG-Based Heart Rate Variability among Drivers

Road accidents are the major cause of fatality and disability across the globe. One of the major reasons attributed to this is the drowsiness while driving. It is necessary to have a system that promptly warns the driver at the onset of drowsiness and ensures the driver's alertness. The transitions from the alert phase to the sleepy phase are evident in heart rate variability (HRV) parameters derived from Electrocardiography (ECG) physiological signals, which serve as an index of the Autonomous Nervous System (ANS). This work involves meticulously analyzing the HRV patterns of each subject to identify and differentiate three states viz., non-drowsy, drowsy, and very drowsy phase. This is achieved by acquiring the ECG signal using a Holter recorder from the subjects in a medium-fidelity simulator driving environment. Eight HRV metrics in both time and frequency domains were calculated from the raw ECG data and were used to interpret the drowsy phase while driving. From the video recording module, behavioral parameters like yawning, eyelid closure, etc. and subjective assessment were taken as Ground truth references. Further, the ECG signals are synchronized with timestamped video. The experimental results clearly indicate the patterns to differentiate the drowsy from non-drowsy phase among all the subjects in the frequency domain HF (High Frequency), LF (Low Frequency), and LF/HF ratio parameters. Time domain metrics patterns were not consistent among all the subjects. Incorporating a drowsiness assessment based on HRV frequency metrics could greatly enhance the effectiveness of current automobile safety mechanisms.