Driver Drowsiness Estimation From EEG Signals Using Online Weighted Adaptation Regularization for Regression (OwARR)

Wu, Dongrui; Lawhern, Vernon J.; Gordon, Stephen; Lance, Brent J.; Lin, Chin-Teng · 2017 · Crossref

DOI: 10.1109/tfuzz.2016.2633379

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 generalizing brain-computer interface (BCI) algorithms to new subjects with minimal calibration data, specifically focusing on online driver drowsiness estimation from EEG signals. While transfer learning (TL) and domain adaptation (DA) have been extensively applied to BCI classification tasks, their application to regression problems remains underexplored. The authors propose a novel algorithm, Online Weighted Adaptation Regularization for Regression (OwARR), which extends previous classification-based methods by integrating fuzzy sets to handle continuous drowsiness indices. Additionally, they introduce a Source Domain Selection (SDS) approach to reduce computational costs by selecting only the most relevant source domains. The OwARR algorithm minimizes a composite objective function that includes the sum of squared errors for both source and target domains, the distance between marginal probability distributions (using Maximum Mean Discrepancy), and the distance between conditional probability distributions. To adapt conditional distributions for regression, the authors transform continuous outputs into three fuzzy classes (Small, Medium, Large) based on percentiles, allowing the use of classification-style distribution adaptation metrics. The algorithm also maximizes an approximate sample Pearson correlation coefficient to prevent constant regression outputs. The SDS method clusters source domains based on the distance between their fuzzy class mean vectors and the target domain, selecting the closest cluster to reduce the number of models trained. Experiments were conducted using a simulated driving dataset from 15 subjects, where drowsiness was induced via monotonous driving in a virtual reality environment. EEG signals were recorded, and drowsiness indices were derived from response times to lane-departure perturbations. The study compared OwARR and OwARR-SDS against baseline approaches, including Domain Adaptation with Model Fusion (DAMF). Results demonstrated that OwARR and OwARR-SDS achieved significantly smaller estimation errors than the baselines. The SDS approach reduced computational cost by approximately half by limiting the number of source domains processed, without compromising estimation accuracy. The authors also provided comprehensive analyses on the robustness of the proposed methods, confirming their effectiveness in handling individual differences and non-stationarity in EEG signals. The significance of this work lies in its contribution to making BCIs more practical for real-world applications by reducing the need for extensive subject-specific calibration. By effectively leveraging data from other subjects through domain adaptation and fuzzy set theory, OwARR offers a robust solution for continuous state estimation, such as driver drowsiness, which is a critical safety concern. The integration of SDS further enhances the feasibility of deploying these algorithms in resource-constrained environments. This research bridges the gap between laboratory-based BCI studies and real-life applications, providing a scalable framework for regression tasks in neurotechnology.

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.

Summary generated by qwen3.6-27b-nvidia on 2026-08-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• drowsiness detection algorithms

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