Investigating Association and Causal Relationships between Physiological Signals and Affective State

This paper offers two approaches to detecting the association and causal relationships between multi-modal physiological signals and the human subjects' affective state. In particular, we focus on the emotional state called here “stress”. The dataset used in this study is the WESAD which contains time series recorded using wearable devices. The first approach we studied is variance-covariance networks using a multivariate auto-regressive model. The second approach investigated causal relationships using the Granger causality test. The study concluded that respiratory, electrocardiogram, and accelerometer signals, measured using wearable sensors, have the strongest correlation with the state of emotional stress. In contrast, electrodermal activity, temperature and electromyogram signals show small or no correlation with such a state. The Granger causal test confirmed a strong causal relationship between the stress state and respiratory, electrocardiogram, and accelerometer.