Multivariate synchrosqueezing transform and time-frequency attention for mental workload classification from EEG signals
DOI: 10.1038/s41598-025-34783-w
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 accurately classifying mental workload (MWL) from electroencephalogram (EEG) signals, a critical task in neuroergonomics and human-machine interaction. The authors identify two primary limitations in existing methods: the poor time-frequency resolution of traditional techniques like STFT and CWT, which suffer from energy smearing, and the neglect of inter-channel spatial dependencies in multichannel EEG modeling. To overcome these issues, the study proposes a multi-stage framework that integrates the Multivariate Synchrosqueezing Transform (MSST) with a novel Convolutional Neural Network (CNN) architecture featuring a Time-Frequency Attention Module (TFAN). The methodology involves four distinct steps. First, EEG signals from two public datasets, STEW and MAT, are preprocessed using dataset-specific filtering protocols to remove artifacts. Second, the MSST is applied to generate precise multichannel time-frequency representations by calculating joint instantaneous frequencies and amplitudes across channels, thereby preserving spatial dependencies. Third, deep features are extracted using a CNN augmented with the TFAN. This module employs a dual-branch attention mechanism: a time-attention branch using frequency-elongated kernels and a frequency-attention branch using time-elongated kernels. These branches generate a 2D attention map that re-weights features to highlight discriminative spectro-temporal regions. Finally, feature dimensionality is reduced using Semi-supervised Discriminant Analysis (SDA), optimized via Bayesian optimization, before classification with a Support Vector Machine (SVM). Experimental results demonstrate that the proposed framework significantly outperforms traditional time-frequency methods and deep learning models lacking attention mechanisms. The optimized combination of SDA and SVM achieved classification accuracies of 97.1% on the STEW dataset and 98.6% on the MAT dataset. The study confirms that MSST provides a sharper time-frequency representation than other methods, and the integration of the attention-enhanced CNN architecture substantially improves classification accuracy by adaptively concentrating on relevant features. The significance of this work lies in its robustness and effectiveness for mental workload analysis. By combining advanced multivariate time-frequency analysis with attention-based deep learning, the framework addresses the nonstationarity and inter-individual variability inherent in EEG signals. This approach offers a reliable tool for adaptive systems in high-stakes environments, such as driving and medical monitoring, where accurate real-time assessment of cognitive load is essential for safety and performance.
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 | canonical_url | — | — | 1 | 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.