Improvement in Chronic Stress Level Recognition by Using Both Full-term and Short-term Measurements of Physiological Features

Abstract

In this paper, we propose improvement in chronic stress level recognition by using both full-term and short-term physiological features. In our proposed method, we employ the characteristic of PSS (Perceived Stress Scale), a widely used chronic stress measure. PSS scores are known to be influenced by mental states caused by high-stress experiences that could occur over shorter terms. So we added new stress features calculated on a weekly basis to the conventional stress features calculated on a monthly basis for which PSS questionnaires recognizes stress level. With weekly-basis feature calculations, we are able to recognize high-stress experiences over shorter terms. To evaluate our proposed method, we performed experiments using a 33-employee, 1-month database of physiological signals. Results have shown the Pearson’s correlation coefficient to improve from 0.66 to 0.72 with use of our method.