ECG Based Stress Detection Using Machine Learning

Today, the endeavour of accomplishment and performance has increased the efficiency immensely, yet it comes with its own price. There has been a drastic increase in the diseases related to stress, especially in the past couple of decades. The plethora of diseases and disorders related to longterm effects of stress vary from muscle related disorders to nervous system related diseases. Stress can be defined as unrest in the normal homeostasis. Since this state of unrest is usually triggered by the sympathetic nervous system as a physiological response, stress can be captured by physiological signals. Though a variety of approaches such as the use of questionnaires, biochemical measures and physiological techniques are available to diagnose stress; physiological signals are the most reliable method. Therefore, we have analysed stress using Electrocardiogram which is a physiological signal to increase the accuracy rate by using machine learning algorithms. Here we propose a simple algorithm for the classification of ECG signal as stress or normal by the automatic detection of heart rate variability from R peaks through DWT method. Works includes ECG raw data extraction, wavelet de-noising, R peak detection and classification. Machine learning algorithm uses various parameters obtained from classification for finding the accuracy of the results. Short term ECG is needed for stress detection, which produces a reliable classification with high accuracy.