Regresar

Development of a System to Detect Stress Using Electrocardiographic Signals and Machine Learning Models

Abstract:

Due to the modern lifestyle, environment, and global events of the century, mental health has become a significant concern worldwide, increasing people's awareness about emotions that can evoke psychological and physiological diseases, such as stress. Stress can lead to severe diseases; therefore, detecting and predicting it is vital to avoid health-related problems. This study develops a monitoring system using physiological signals and machine learning models to detect stress responses. Electrocardiographic data has been obtained from datasets recorded during different stress conditions. The signals are filtered, and the heart rate variability has been analyzed to get the time and frequency domain features. Processed data feed the different models and perform stress binary classification. After comparing results on five machine learning models, the random forest classifier obtained an accuracy of 82.4%. Additional signals were recorded in a no laboratory environment and processed, and the stress detection protocol suggested could successfully determine the presence of stress.