Non Invasive Stress Detection Method Based on Discrete Wavelet Transform and Machine Learning Algorithms

Stress can lead to many serious health problems as it can cause long-term chemical changes if our body is exposed to it for a lengthened period. Thus, it is important to develop a stress detection system to provide early warning for individuals. This is a study about the stress detection in students during their real time activities using the more convenient wearable smartwatch (Raqib) with the complete access to the raw ECG data. We have proposed a Discrete Wavelet Transform (DWT) method to preprocess the raw data along with the use of symlet4 filter to generate 4 levels of Daubechies coefficients. This study identifies 9 different time as well frequency-based features that show significant dependence on stress or simply directly related to stress. We have applied 7 different classifiers to our ECG data features obtained from 30 students and Naïve Bayes classifier produces the highest accuracy of 96.67% under 10-fold cross validation procedure.