Emotion recognition at a distance: The robustness of machine learning based on hand-crafted facial features vs deep learning models

•Comparative study of a Deep Learning technique versus a Machine Learning one is conducted.•Dynamic-based methods are applied to classify three different expression phases.•Expression stages for each basic emotion are exploited in an “at-a-distance” scenario.•Experiments were conducted on the well-known Cohn-Kanade dataset (CK+).