Analýza fyziologických dat řidičů při variabilních situacích

Abstract

The thesis focuses on the analysis and classification of drivers' physiological data collected during simulated drives under various stress-inducing situations. The main physiological parameters examined were heart rate (HR), heart rate variability (HRV), and galvanic skin response (GSR), which enable objective evaluation of stress factors and drivers' emotional reactions. The work focuses on the preprocessing of physiological signals during stress-inducing situations, the implementation of algorithms for identifying relationships between physiological responses and specific driving scenarios, and the quantitative evaluation of these relationships.The classification of stress-related situations based on physiological signals involved both traditional statistical methods and advanced approaches, such as Support Vector Machine (SVM) models and shallow neural network models, aiming to improve the accuracy of stress detection in different situations. The results demonstrated that the combination of HR and GSR parameters provides more reliable stress situation classification outcomes compared to the combination of HRV and GSR parameters for short 6–7 minute signal recordings.The highest achieved 95% accuracy in binary stress classification was obtained using shallow neural networks based on HR and GSR parameters.The proposed methods and algorithms were successfully validated on real-world data and demonstrated potential forimplementation in driver assistance systems.