Monitoring pilots’ mental workload in real flight conditions using multinomial logistic regression with a ridge estimator

Haseeb, Muhammad; Nadeem, Rashid; Sultana, Nazia; Naseer, Noman; Nazeer, Hammad; Dehais, Frédéric · 2025 · Crossref

DOI: 10.3389/frobt.2025.1441801

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 pilot mental workload, a primary contributor to aviation accidents, by developing a robust monitoring system for real flight conditions. While previous research often relied on cumbersome wet-electrode EEG systems or controlled laboratory settings, this work aims to improve classification accuracy using advanced machine learning techniques on data collected from pilots operating under visual flight rules (VFR). The motivation stems from the need for practical, non-invasive brain-computer interfaces (pBCIs) that can function effectively in noisy cockpit environments without hindering pilot performance. The researchers utilized EEG data recorded from 22 pilots using a six-channel dry-electrode Enobio Neuroelectrics system. Five subjects were excluded due to data synchronization issues, leaving 17 participants for analysis. The experimental design involved two traffic patterns: a low-load condition where pilots monitored an instructor-controlled flight, and a high-load condition where pilots flew the aircraft themselves. Data preprocessing included high-pass filtering and artifact removal using the Riemannian artifact subspace reconstruction (rASR) filter. From the six channels, 72 features were extracted, comprising power spectral bands (delta, theta, alpha, beta, gamma) and statistical metrics (mean, variance, skewness, etc.). Feature selection was performed using Information Gain (IG) and Correlation Attribute Evaluator (CAE) methods within the WEKA environment. Finally, 15 different classifiers were evaluated to determine the optimal model for workload detection. The results demonstrated that multinomial logistic regression with a ridge estimator achieved the highest performance, yielding a mean classification accuracy of 84.6% with a mean classification time of 0.03 seconds. Feature selection analysis revealed that the Information Gain method with 25 selected features provided the best trade-off between accuracy and computational efficiency, outperforming the Correlation Attribute Evaluator, which suffered from higher processing times and greater accuracy degradation as feature counts decreased. The selected feature set consisted of approximately 50% power-based and 50% statistical features. Statistical validation via Student’s t-test confirmed the significance of these results (p < 0.05). The study concludes that multinomial logistic regression with a ridge estimator is an effective and efficient method for detecting pilot mental workload in real-world scenarios. By achieving competitive accuracy with a minimal six-electrode dry-EEG setup, the research establishes a new benchmark for simplified, operational-ready monitoring systems. This approach offers a practical solution for enhancing aviation safety by providing real-time insights into pilot cognitive states, thereby potentially reducing human error in high-stakes flight environments.

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.

• workload measurement

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