Answers are grounded in this paper and cite their sources.
Driver drowsiness is a major contributor to road traffic crashes, highlighting the need for objective and practical monitoring systems. Electroencephalography (EEG) directly reflects neural dynamics associated with drowsiness, but cross‐subject estimation and few‐channel deployment remain challenging. This study systematically evaluated global field power (GFP), a low‐dimensional descriptor of scalp‐potential dispersion, as a compact EEG feature for cross‐subject driver drowsiness estimation using the public Sustained Attention Driving Task dataset. The evaluation included neurophysiological analysis, alert‐baseline‐based machine learning, subject‐independent deep learning, frequency‐band ablation, temporal smoothing, and reduced‐channel testing. With full‐channel EEG, GFP‐SVR achieved a Pearson correlation coefficient (PCC) of 0.6700 ± 0.2124, while Multi‐window‐EEGTCNet with GFP reached 0.7624 ± 0.1305. Frequency ablation identified θ‐ and α‐band GFP as the main contributors. Under reduced‐channel settings, GFP remained robust in deep learning, yielding mean PCCs of 0.6274–0.7043 and consistently outperforming raw EEG. In the GFP‐SVR analysis, a 24‐s causal moving average further restored substantial predictive performance, with mean PCC above 0.60. These results demonstrate that GFP is a compact and computationally efficient representation of drowsiness‐related EEG dynamics, supporting robust cross‐subject prediction and practical few‐channel monitoring.
No takes yet. Share an insight, caveat, or question.
Wu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: