Predicting emotions induced by music using system identification theory

Modeling the emotional content of music is of great importance, since it is believed that music is capable of inducing different emotions. In this study we present an Autoregressive with Exogenous Input (ARX) model based on system identification theory for modeling the emotional content of music in a two dimensional emotion space and also a nonlinear Autoregressive with Exogenous Input (NARX) model to capture the nonlinear characteristics of the system. We also investigate the causal relationship between musical features and the induced emotions by removing the autoregressive terms from the developed model. Finally A brief discussion about the most important features is presented.