Abstract: Agitation is a serious problem in sedated intensive care unit (ICU) patients. In this work, standard machine learning techniques working on wearable accelerometer data have been used to classifying agitation levels achieving very good classification performances.