Differential Sensitivity of EEG, ECG, Respiration, And GSR To Mental States Classification: A Comparative Machine-Learning Analysis

Open Access
Article
Conference Proceedings
Authors: Francis Logah, Younho Seong
Abstract

Understanding a pilot's cognitive state is critical for improving flight safety, optimizing human performance, and enabling intelligent adaptive cockpit systems. Although electroencephalography (EEG) has long been considered the benchmark for measuring cognitive workload, there has been a growing interest in the ability of other physiological signals, such as electrocardiography (ECG), respiration, and galvanic skin response (GSR) to reliably detect changes in mental state either individually or combined. This study uses four supervised machine learning algorithms: Logistic Regression, Support Vector Machine (SVM), Random Forest, and Multilayer Perceptron (MLP) to compare the sensitivity of these physiological modalities for classifying pilot mental states and examines whether multimodal physiological fusion provides superior performance. Model performance was assessed using Macro-F1 score, confusion matrices, and state-specific recall. The study showed that fusing multiple physiological signals consistently produced the most accurate mental-state classification. Multimodal fusion achieved the highest overall performance with Logistic Regression (Macro-F1 = 0.973); outperforming every individual single modality. Among the individual physiological modalities, MLP achieved the highest EEG performance (Macro-F1 = 0.930). More importantly, the fusion of ECG, Respiration, and GSR substantially improved classification accuracy with a maximum Macro-F1 score of 0.902 using the SVM classifier. This represents an approximately 18-37% point increment over the individual peripheral modalities. Overall, the study highlights the value of multimodal physiological sensing for mental state assessment. This work supports the development of robust, real-time pilot monitoring systems capable of enhancing adaptive cockpit technologies, decision-support systems, and human-centered aviation operations.

Keywords: Electroencephalography, Multimodal, Multilayer Perceptron, Algorithm, Galvanic, Physiological, Machine-Learning, Fusion, Respiration

DOI: 10.54941/ahfe1008333

Cite this paper
Downloads
0
Visits
1
Download PDF

More from this volume

← The Impact of Visual UI Elements on Information Comprehension in WebsitesEvidence-Informed Modular Design for Restorative Campus Open Spaces: A Case Study at Harbin Institute of Technology →
View all articles in Human Factors in Design, Engineering, and Computing →