Depression and Fatigue Symptom Recognition Based on Electrocardiogram Signals

The onset of depression is often marked by a range of physiological symptoms, including persistent fatigue, increased body weight in the morning, and a diminished appetite. Early detection of these symptoms can play a critical role in preventing the development of depression and enabling timely interventions for patients. This study proposes a method for early depression prevention based on real-world scenarios, specifically through the detection of depressive fatigue symptoms, allowing for targeted interventions. In this experiment, electrocardiogram (ECG) signals were collected from 77 college students between 22:30 and 23:00. The raw data were preprocessed, and a total of 26 autonomic nervous system features were extracted, which include 8 time-domain features, 10 frequency-domain features, and 8 nonlinear features. Five machine learning classifiers were then trained to classify depression and fatigue. Among them, XGBoost demonstrated the best performance, achieving an accuracy of 79.74 %, a recall of 71.89 %, and an F1 score of 74.19 %.