Feedback based real time facial and head gesture recognition for e-learning system
Shubhada Deshmukh, Manasi S. Patwardhan, Anjali Mahajan · 2018
Lack of feedback 1, which an instructor receives in a classroom setting in terms of facial expressions of the listeners, serves as a deterrent to the advantages of the exiting e-learning systems. For such real-time application, there is a need for a facial expression recognition system which not only provides good accuracy, but also is time-efficient. It should not only work with traditional datasets, but should be invariant to the new subjects. In this paper, we have come up with our own set of features and an algorithm for non-traditional set of facial expressions, such as sleeping, yawning, smiling and head gestures, such as nodding, shaking and tilting. This setup not only provides the required real-time efficiency, but also gives very good accuracy for these new set of gestures. With distances between the action units acting as time efficient set of features, we have achieved 97% accuracy for facial expressions. Whereas, the vector movement of a single nose tip point serving as feature for head gestures, we have achieved 98% accuracy in real-time. We have applied a simple normalization technique to make it person-independent.