Classification of Driver Cognitive Load based on Physiological Data: Exploring Recurrent Neural Networks
DOI: 10.1109/icarm54641.2022.9959588
archive: archived pipeline: cataloged verified
Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)
Summary
This study addresses the challenge of classifying driver cognitive load using physiological data, specifically focusing on the generalizability of machine learning models across different individuals. Previous research primarily utilized "within-driver" partitioning, where models were trained and tested on data from the same participants, raising concerns about practical feasibility and generalizability in real-world applications. The authors aimed to evaluate the performance of widely used machine learning models (SVM, ANN, KNN, DTC, NB, LDA) and compare them with Recurrent Neural Networks (RNN), which are better suited for temporal data, under both within-driver and "across-drivers" partitioning schemes. The researchers utilized a driving simulator dataset comprising physiological signals (EEG, ECG, GSR) from 33 participants. Participants completed three drives: a baseline with no secondary task, and two drives involving auditory-verbal n-back tasks of varying difficulty (1-back and 2-back) to induce cognitive load. Data processing involved extracting 12 features: 10 from EEG power bands, one average heart rate from ECG, and one average galvanic skin response from GSR. To mitigate individual differences, data was normalized using baseline conditions. The study employed a binary classification approach (no-task vs. n-back task). Models were trained and tested using two partitioning methods: within-driver (80/20 split of each participant's data) and across-drivers (training on ~80% of participants and testing on the remaining ~20%). Results indicated that all models performed better with within-driver partitioning than with across-drivers partitioning. However, the RNN model, specifically a 2-layer LSTM architecture, outperformed all other models in both scenarios. The RNN achieved mean accuracies of 88.1% for within-driver partitioning and 85.6% for across-drivers partitioning. It also demonstrated greater robustness, evidenced by a narrower range of accuracy values across different data splits compared to traditional models like SVM or LDA, which had lower accuracy and higher variability. Precision-recall curves further confirmed that RNN maintained high precision without significant loss in recall, despite data imbalance. The study concludes that RNNs are superior for classifying driver cognitive load from physiological time-series data, particularly when generalizing across different drivers. This suggests that RNNs are more viable for practical in-vehicle systems that must adapt to new users without retraining. However, the authors note that RNNs are "black-box" models with lower interpretability than traditional methods. Future research should focus on larger, more diverse datasets and real-world in-vehicle tasks to validate these findings and explore ensemble methods that might balance accuracy with interpretability.
Key finding
Recurrent neural networks outperformed traditional machine learning models in classifying driver cognitive load, achieving mean accuracies of 88.1% for within-driver partitioning and 85.6% for across-drivers partitioning.
Methodology
simulator
Sample size: 33
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.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | — | — | — | 1 | 2026-05-28 |
| archive | success | canonical_url | — | — | 1 | 2026-06-06 |
| extract | success | cached | — | — | 3 | 2026-06-10 |
| clean | success | clean | — | — | 1 | 2026-06-07 |
| chunk | success | chunk | — | — | 1 | 2026-06-07 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-07 |
| enrich | success | semantic_scholar | — | — | 4 | 2026-06-15 |
| 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
- Theoretical Contribution: computational model