Abstract
Stress is associated with significant behavioral and physiological changes, including decreased heart rate variability (HRV) at rest. Environmental factors such as air pollution are increasingly recognized as potential triggers of physiological stress responses, especially in highly polluted cities such as Almaty, Kazakhstan. However, the relationship between air quality and HRV as a physiological stress marker has not been sufficiently studied. This study explores the development of an IoT system for assessing physiological stress levels based on HRV under various environmental conditions, with a particular focus on air pollution. The study was conducted in three contrasting locations in Almaty, Kazakhstan: a green vegetation area (Botanical Garden), a busy urban area (Al-Farabi Avenue), and an enclosed space with regulated conditions (laboratory). HRV data were synchronously recorded from 10 healthy volunteers using both an optical photoplethysmography (PPG) sensor and an electrocardiographic (ECG) sensor, while air quality parameters (PM2.5, PM10, CO₂) were measured simultaneously. The results showed that sympathetic nervous system activation was most pronounced in the Botanical Garden, where elevated levels of particulate matter (PM2.5 and PM10) were detected. Fine PM2.5 particles had the most significant impact on HRV, followed by PM10 and CO₂, leading to a reduction in overall HRV and an increase in the low-frequency to high-frequency (LF/HF) ratio, indicating heightened physiological stress. Machine learning models, including DNN, XGBoost, Random Forest, and TabNet, were developed and trained to assess stress levels based on air quality data. Among them, the XGBoost model achieved the highest classification accuracy of 91.92%. This research provides valuable insights for evaluating disease risks and analyzing the potential impact of long-term exposure to polluted air on the cardiovascular system.