Title:Affect-driven Ordinal Engagement Measurement from Video

Abstract:In education and intervention programs, user engagement has been identified as a major factor in successful program completion. Automatic measurement of user engagement provides helpful information for instructors to meet program objectives and individualize program delivery. In this paper, we present a novel approach for video-based engagement measurement in virtual learning programs. We propose to use affect states, continuous values of valence and arousal extracted from consecutive video frames, along with a new latent affective feature vector and behavioral features for engagement measurement. Deep-learning sequential models are trained and validated on the extracted frame-level features. In addition, due to the fact that engagement is an ordinal variable, we develop the ordinal versions of the above models in order to address the problem of engagement measurement as an ordinal classification problem. We evaluated the performance of the proposed method on the only two publicly available video engagement measurement datasets, DAiSEE and EmotiW-EW, containing videos of students in online learning programs. Our experiments show a state-of-the-art engagement level classification accuracy of 67.4% on the DAiSEE dataset, and a regression mean squared error of 0.0508 on the EmotiW-EW dataset. Our ablation study shows the effectiveness of incorporating affect states and ordinality of engagement in engagement measurement.