Brain Dynamics in Predicting Driving Fatigue Using a Recurrent Self-Evolving Fuzzy Neural Network
DOI: 10.1109/tnnls.2015.2496330
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 critical safety issue of driving fatigue by proposing a generalized prediction system capable of estimating driver drowsiness levels using electroencephalography (EEG) signals. While EEG-based brain-computer interfaces (BCIs) are effective for monitoring cognitive states, existing systems often suffer from low resolution and a lack of generalizability across different subjects. Furthermore, traditional neural networks struggle with the temporal continuity of cognitive states. To overcome these limitations, the authors introduce a Recurrent Self-Evolving Fuzzy Neural Network (RSEFNN). This model combines the memory capabilities of recurrent neural networks with the adaptive learning of fuzzy neural networks, allowing it to capture nonlinear dynamics and temporal dependencies in EEG data for real-time fatigue prediction. The experimental design utilized a virtual-reality (VR) driving simulator equipped with a six-axis motion platform to create a realistic long-term driving environment. Twenty healthy young adults participated in a 90-minute simulated night drive on a four-lane highway. The study employed an event-related lane-departure paradigm, where random perturbations caused the vehicle to drift, requiring participants to steer back to the center. The primary metric for drowsiness was the response time (RT) to these deviations. EEG signals were recorded from 33 electrodes, with specific focus on the occipital lobe (O1, O2, Oz). The researchers extracted power spectrum features from the delta, theta, alpha, and beta frequency bands using the 5 seconds of EEG data preceding each deviation event. These features served as inputs to predict the normalized RT, which represented the driver's fatigue level. The proposed RSEFNN architecture features a six-layer structure with local feedback connections that allow the network to memorize past and current EEG states without complex external registers. The model employs simultaneous structure and parameter learning via an on-line gradient descent algorithm, enabling it to adaptively generate fuzzy rules and update parameters in real-time. This approach treats fatigue as a continuous dynamic process rather than isolated events. The system was evaluated using a generalized cross-subject approach, meaning models trained on some participants were tested on others to assess generalizability. The RSEFNN was compared against several benchmark systems, including support vector regression, self-organizing neural fuzzy inference networks, and various recurrent fuzzy and wavelet neural networks. The results demonstrated that the RSEFNN significantly outperformed competing models in terms of prediction error within the generalized cross-subject framework. The proposed system required fewer feature dimensions, which reduced computational time and enhanced its suitability for real-time applications. By effectively capturing the temporal continuity of brain dynamics, the RSEFNN provided a more accurate and robust estimation of driving fatigue than static or non-recurrent architectures. The study concludes that this recurrent fuzzy neural network approach offers a promising solution for developing practical, real-time BCI systems for monitoring driver cognitive states and improving road safety.
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 | OpenAlex-citations | — | — | 1 | 2026-06-19 |
| archive | success | unpaywall | — | — | 2 | 2026-06-25 |
| extract | success | cached | — | — | 2 | 2026-06-26 |
| clean | success | clean | — | — | 1 | 2026-06-19 |
| chunk | success | chunk | — | — | 1 | 2026-06-19 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-19 |
| promote | success | — | — | — | 1 | 2026-06-19 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 1 | 2026-06-26 |
| tag | success | vector_similarity | — | — | 6 | 2026-06-19 |
| verify | success | — | — | — | 1 | 2026-06-26 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; 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
- Theoretical Contribution: computational model