Multi_class Stress Detection Through Heart Rate Variability

Humans naturally experience stress when facing pressure or demanding situations, especially when these demands are perceived as harmful or overwhelming. Continuous and excessive stress can increase the risk of both mental and physical health problems. Long-term exposure to stress may lead to issues such as anxiety, depression, and sleep disorders. One common method used to evaluate stress is the analysis of physiological signals such as Heart Rate Variability (HRV). However, achieving very high accuracy in stress detection using HRV data remains a challenging task. HRV is different from heart rate; heart rate represents the average number of heartbeats per minute, whereas HRV measures the variation in the time interval between consecutive heartbeats. HRV is calculated from the variation in RR intervals, where RR represents the time between two successive peaks in a heartbeat signal. In this study, a machine learning model is developed for multi-class stress detection by analyzing HRV parameters as biological indicators of stress. Specifically, a convolutional neural network (CNN) model is used to classify three types of stress conditions: no stress, interruption stress, and time pressure stress. The model utilizes both timedomain and frequency-domain features of HRV signals for accurate classification. Experimental results using the publicly available SWELL-KW dataset demonstrate that the proposed model achieves very high performance, reaching an accuracy of 99.9% with Precision = 1, Recall = 1, F1- score = 1, and MCC = 0.99. In addition, the study shows that Analysis of Variance (ANOVA) is an effective feature selection technique for identifying important HRV parameters related to stress detection.