Mental Stress Assessment using Wavelet Transform Features of Electrocardiogram Signals

Mental stress is an unavoidable part of our daily life now-a days and its presence in long term causes adverse affects on mental and physical health of person. In the presence of a stress event, there is an unbalance in activities of autonomic nervous system (ANS) that further results in a irregular heart function. The variations in resulting heart function in the presence of mental stress can be measured as bioelectric signals using electrocardiogram (ECG) and its temporal and morphological features used as a significant marker of stress. In this work, we used discrete wavelet decomposition to extract frequency components of ECG signal and calculated standard deviation (SD), entropy, and total energy for selected frequency components significant to the variations caused by stress events. The feature set is formed using calculated parameters and a multiclass logistic regression (MLR) model is trained to classify the mental stress in three different levels. The proposed method is validated with classification accuracy = 90.8% using Physionet data base containing ECG recording under different stress events. The presented work demonstrates the use of ECG signal as a significant marker for automatic assessment of mental stress.