Workplace Stress Estimation Method based on Multivariate Analysis of Physiological Indices

In this research, we have been developing a new integrated analysis method of multiple physiological signals to estimate stress in daily life, which is important in depression screening and life-style related diseases prevention. Experiments have been carried out on 100 participants, measuring electrocardiogram, pulse wave, breath rhythm, and skin temperature in four patterns of psychological states; relax state, normal stress state, monotonous stress state, and nervous state. The newly developed stress state estimation method relies on the integrated analysis of nine physiological indices related to stress that have been extracted from the four measured physiological signals. Because variation range of each index is different between individuals and types of stress, we divided estimation process into three steps. For each step, we performed cross-validation using various classification schemes to select the most relevant set of indices that enable estimation of stress state with few influences of individual variations. Through this method we could achieve 87% accuracy for stress detection, and 63% accuracy for stress type classification. Finally a validation study was performed to confirm this method can be an effective solution to estimate various types of stress state regardless of individuals.