ID3RSNet: cross-subject driver drowsiness detection from raw single-channel EEG with an interpretable residual shrinkage network
DOI: 10.3389/fnins.2024.1508747
archive: archived pipeline: cataloged verified
Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)
Summary
This paper addresses the challenge of developing accurate, calibration-free driver drowsiness detection systems using single-channel electroencephalography (EEG). While EEG is considered the gold standard for monitoring brain activity related to fatigue, single-channel setups are difficult to utilize due to signal non-stationarity, individual heterogeneity, and low signal-to-noise ratios. Furthermore, existing deep learning models often lack interpretability, functioning as "black boxes" that hinder trust and practical deployment. To resolve these issues, the authors propose ID3RSNet, a novel interpretable residual shrinkage network designed for cross-subject drowsiness detection from raw single-channel EEG data. The ID3RSNet framework comprises three primary modules: a base feature extractor (BaseFE), a residual shrinkage building unit (RSBU), and a classification layer. The BaseFE utilizes a shallow 1-D convolutional neural network to extract essential frequency features from 3-second EEG segments. The RSBU integrates an attention mechanism with soft thresholding to automatically denoise signals and recalibrate features, enhancing discriminative power while suppressing redundant information. To prevent overfitting and enable interpretability, the model employs global average pooling (GAP) and a fully connected layer with weight freezing (FC-WF). Additionally, the authors introduce an EEG-based Class Activation Map (ECAM) method, which visualizes the specific neurophysiological patterns the model uses for classification decisions. Experiments were conducted using the Sustained-Attention Driving Task (SADT) dataset, which contains single-channel EEG data (Oz channel) from 11 subjects sampled at 128 Hz. The study employed a leave-one-subject-out cross-validation (LOSO-CV) protocol to evaluate cross-subject generalization. The proposed ID3RSNet was compared against conventional machine learning classifiers (e.g., SVM, Random Forest) and state-of-the-art deep learning baselines, including EEGNet and ShallowConvNet. The results demonstrated that ID3RSNet achieved superior classification performance compared to these baselines. Crucially, the ECAM interpretation method provided reliable neurophysiological evidence, highlighting that the model correctly identified drowsiness-related patterns in the theta and alpha frequency bands, thereby validating the biological plausibility of the learned features. The significance of this work lies in its dual contribution to performance and interpretability in wearable EEG applications. By achieving high accuracy with a single channel, the method supports the development of cost-effective, portable drowsiness monitoring devices. The inherent interpretability of the residual shrinkage network addresses the critical limitation of black-box models in safety-critical applications, offering transparency into how neural features correlate with driver mental states. This approach facilitates the deployment of trustworthy, calibration-free systems for real-time fatigue detection, potentially reducing traffic accidents caused by drowsy driving.
Key finding
The proposed interpretable residual shrinkage network achieves superior cross-subject driver drowsiness detection performance using single-channel EEG while providing visualizable, neurophysiologically reliable classification evidence.
Methodology
lab_experiment
Sample size: 11
Provenance
The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via topic_sweep_doaj on 2026-06-01.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | — | — | — | 1 | 2026-06-01 |
| archive | success | canonical_url | — | — | 1 | 2026-06-06 |
| extract | success | cached | — | — | 3 | 2026-06-10 |
| clean | success | clean | — | — | 1 | 2026-06-04 |
| chunk | success | chunk | — | — | 1 | 2026-06-04 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-04 |
| enrich | success | — | — | — | 1 | 2026-06-01 |
| promote | success | — | — | — | 1 | 2026-06-04 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 2 | 2026-06-10 |
| tag | success | vector_similarity | — | — | 15 | 2026-06-11 |
| verify | success | — | — | — | 2 | 2026-06-10 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.
Topics
Ranked by relevance to this paper. Hover a topic for its definition.
Information type
What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).
- Empirical Findings: physiological data
- Methodological Resource: tool software