Mental Stress Assessment via Ultra-Short-Term Recordings of Photoplethysmographic Sensor

This paper investigates the potential of ultra-short term recordings of a low cost Photoplethysmographic (PPG) sensor to detect multilevel mental stress. For this purpose, we designed an experimental paradigm to induce different level of stress using Mental Arithmetic Tasks (MAT). Stress-related data was acquired with a single low-cost PPG sensor. After estimating pulse rate variability series from 60 seconds long segments of PPG signals, we computed different features based on their reliability for ultra-short term PRV analysis. In order to mitigate the issues of irrelevancy and redundancy among features, we employed a Sequential Forward Floating Selection (SFFS) algorithm to select an optimum feature set. We developed two classifiers based on Quadratic Discriminant Analysis (QDA) and Support Vector Machine (SVM). The results of the proposed stress detection system produced 92% accuracy with SVM for five level identification of mental stress. In conclusion, we proposed a multilevel stress detection system that has the potential to detect five different mental stress states using the ultra-short recordings of a low-cost PPG sensor.