An Effectiveness Comparison between the Use of Activity State Data and That of Activity Magnitude Data in Chronic Stress Recognition

Abstract

Our aim is to improve the performance of the early recognition of chronic stress, through more effective monitoring of physiological signals produced as people live their daily lives (as opposed to monitoring during brief examination periods when physical activity is controlled). Physiological signals are influenced not only by responses to stress but also by physical activities, and it is necessary to distinguish between these two types of influence. There are basically two approaches to doing this. One is to separate the signals in terms of states of physical activity, such as sitting, walking, or running (the “Activity State” approach), and the other is to separate the signals in terms of the magnitude of physical activity (the “Activity Magnitude” approach). To determine which approach leads to better stress recognition performance, we performed evaluations using a database of 64 subjects and compared results for the two approaches. Results showed that the “Activity State” approach was, to a statistically significant degree, superior to the “Activity Magnitude” approach in the recognition of chronic stress.