Assessment of Mental Workload Using a Transformer Network and Two Prefrontal EEG Channels: An Unparameterized Approach
archive: archived pipeline: cataloged
Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)
Summary
This paper addresses the challenge of assessing mental workload using wearable electroencephalography (EEG) devices, which typically offer fewer channels than research-grade systems. While previous studies have shown promise with consumer-oriented EEG headsets, most relied on single-database validation, limiting the generalizability of their findings. This research investigates whether a transformer network can effectively classify mental workload levels using only two prefrontal EEG channels, validated across two distinct commercial datasets to ensure portability and practical applicability. The study utilized 60 recordings from 46 participants performing a three-level n-back game (0-back, 1-back, and 2-back), corresponding to low, medium, and high mental workload. Data was collected using two different EEG devices: the Enobio system (Database E, 500 Hz sampling rate, Fp1/Fp2 channels) and the Muse headband (Database M, 256 Hz sampling rate, AF7/AF8 channels). The methodology involved denoising the raw signals via band-pass filtering and discrete wavelet transform (DWT) to remove eye-blink artifacts. The signals were then decomposed into sub-bands, from which Shannon entropy and wavelet log energy features were extracted. These features were fed into five classifiers: Support Vector Machine (SVM), k-Nearest Neighbors (kNN), Multi-Layer Perceptron (MLP), AdaBoost, and a Transformer Network (TN). The TN architecture processed sequences of feature vectors using self-attention mechanisms, requiring no hyperparameter calibration for the feature extraction stage. Results indicated that the Transformer Network outperformed the other four classifiers in both datasets. It achieved a mean accuracy of 88% for Database M and 85% for Database E. The consistent performance across two devices with different sampling rates and channel configurations demonstrates the robustness of the proposed unparameterized approach. The findings suggest that complex cognitive states can be accurately monitored using minimal EEG channels when combined with advanced deep learning architectures. The significance of this work lies in its demonstration that high-accuracy mental workload assessment is feasible with consumer-grade, two-channel EEG devices. By validating the method on two independent databases, the authors address the portability concerns that have limited previous wearable EEG studies. This approach offers a viable pathway for real-life, continuous monitoring of cognitive load in practical settings, such as driving or industrial tasks, without the complexity and cost associated with multi-channel research-grade EEG systems.
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 | Crossref | — | — | 1 | 2026-08-09 |
| archive | success | unpaywall | — | — | 2 | 2026-08-09 |
| extract | success | cached | — | — | 4 | 2026-08-23 |
| clean | success | clean | — | — | 1 | 2026-08-09 |
| chunk | success | chunk | — | — | 1 | 2026-08-09 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-08-09 |
| enrich | success | semantic_scholar | — | — | 1 | 2026-08-09 |
| promote | success | — | — | — | 1 | 2026-08-09 |
| summarize | success | llm | qwen3.8-27b-gittensor | summ-v5 | 3 | 2026-08-23 |
| tag | success | vector_similarity | — | — | 10 | 2026-08-11 |
| verify | success | — | — | — | 2 | 2026-08-09 |
Summary generated by qwen3.8-27b-gittensor on 2026-08-23; verification: pending re-verification.
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