A Decision Tree Optimised SVM Model for Stress Detection using Biosignals

In our work we propose a machine learning model based on human bio signals to detect human stress. Detecting stress properly can help in preventing a large number of mental and physical scenarios which lead to abnormalities in cardiac rhythm or depression and more. In our work we selected ECG as the bio signal and extracted its features. The advantage of taking ECG as the bio signal is, information about respiratory signals EDR (ECG Derived Respiration) feature can be easily derived without any extra sensors. Among those unique features we chose ECG derived Respiration, Respiration Rate, QT interval. For training and validation of our new model we used Physionet’s “drivedb” database. Our proposed model uses Optimised Support Vector Machines (SVM) using decision trees. Our experimentation results show better accuracy in detecting stress