In this paper we explore the stress level experienced by elderly pedestrians in city environment. For that, we selected three types of traffic crossings, often encountered in most cities and known to be stressful in the capital city of Tunis. The evoked stress levels on 20 elderly presbycusis pedestrians are then captured using a non-invasive wireless sensor of the Electrodermal activity (EDA). The EDA has been established as a lead physiological indicator of human stress. Preliminary observations of the data reflected multi-modal EDA density distribution. In order to further analysis the different stress categories, a clustering-based analysis of EDA signals is performed using Mixture Gaussian Models (GMM). The optimized process of the number of EDA-stress clusters is achieved using a Bayesian Information Criteria and the detection of entropy rate of increase. © 2024 IEEE.