We study the characteristics of infants’ spontaneous movements, based on data obtained from a markerless motion tracking system. From the pose data, the set of features are generated from the raw joint-angles of the infants and different classifiers are trained and evaluated using annotated data. Furthermore, we look at the importance of different features and outline the most significant features for detecting spontaneous movements of infants. Using these findings for further analysis of infants’ movements, this might be used to identify infants in risk of cerebral palsy