Electrogastrogram-Derived Features for Automated Sickness Detection in Driving Simulator

Jakus, Grega; Sodnik, Jaka; Miljković, Nadica · 2022 · Crossref

DOI: 10.3390/s22228616

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 challenge of objectively and continuously assessing simulator sickness, specifically nausea, in driving simulators used for evaluating automated driving systems. While subjective questionnaires are common, they are retrospective; physiological measures like the electrogastrogram (EGG) offer real-time assessment but are highly susceptible to noise and motion artifacts, particularly in dynamic simulator environments. The authors aim to identify EGG-derived features that are robust to noise contamination and suitable for automated nausea detection. To achieve this, they introduce novel parameters alongside traditional metrics, hypothesizing that measures of signal randomness and non-linearity can effectively distinguish between baseline gastric activity and nausea-induced dysrhythmias. The researchers utilized EGG data from 20 healthy participants recorded during a high-fidelity driving simulation involving autonomous vehicle scenarios. After excluding three subjects due to poor signal quality, data from 17 participants were analyzed. To evaluate feature robustness, the authors created semi-synthetic datasets by adding pseudo-random colored noise to the original signals at various Signal-to-Noise Ratios (SNRs), ranging from approximately −23 dB to 17 dB. They extracted a comprehensive set of features, including traditional metrics like dominant frequency and root mean square amplitude, as well as novel parameters: sample entropy, spectral entropy, autocorrelation zero-crossing, and features derived from Poincaré plots. These features were analyzed using both statistical linear methods and machine learning techniques, specifically Random Forest classifiers, to detect nausea incidence based on participant button presses. The results demonstrated that sample entropy exhibited outstanding robustness across different noise levels, making it a highly reliable indicator for nausea detection. Autocorrelation zero-crossing, dominant frequency, and median frequency showed moderate robustness. The machine learning model achieved an accuracy of 88.2% in classifying nausea states, identifying sample entropy as one of the most relevant features. In contrast, linear statistical analysis highlighted spectral entropy, spectral variation distribution, and the crest factor of the Power Spectral Density as significant indicators. The study confirms that nausea causes EGG signals to lose regularity and become more random, a change effectively captured by entropy-based and non-linear features. The significance of this work lies in its provision of a robust, automated framework for objective nausea assessment in noisy, real-world-like conditions. By validating the efficacy of sample entropy and other novel features, the study supports the integration of EGG-based monitoring into driving simulator evaluations. The findings advocate for a complementary approach combining machine learning and statistical analysis, emphasizing the need for customized feature selection to handle environmental noise. This contributes to the standardization of EGG techniques for assessing human factors in automated driving and broader brain-gut interaction research.

Key finding

Sample entropy emerged as the most robust and relevant EGG feature for automated nausea detection, enabling machine learning models to achieve 88.2% accuracy despite signal noise.

Methodology

simulator

Sample size: 17

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-prismaquant on 2026-06-10; verification: verified.

Topics

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

• simulator sickness
• passenger motion sickness comfort
• 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
• Methodological Resource: validation psychometrics, tool software