Occlusion-aware Dynamic Human Emotion Recognition Using Landmark Detection

Human emotion analysis depends upon the accurate identification of key frames and balanced representation of facial features. Accurate identification and tracking of facial features while including variations in pose, face shape, illumination, and image resolution is very much essential for serving the purpose. The complexity improves when the task has to be performed from videos since tracking human emotions frame-wise is cumbersome. Face alignment issues and occlusions need to be taken into account in a robust manner. This paper proposes novel methods based on ACNN for dynamic emotion recognition from human videos.