An Optical Flow Based Approach for Facial Expression Recognition
Shivangi Anthwal, Dinesh Ganotra · 2019 International Conference on Power Electronics, Control and Automation (ICPECA) · 2019
Facial expressions provide pertinent cues helpful in understanding one's emotional state. Interpreting emotions through facial expressions has garnered overwhelming interest in the last few decades due to its prospective applications in smart environments and automated systems. In this work, a method is presented where facial expressions are identified by analyzing motion pattern of facial features arising when the face transforms from neutral state to emotional state. Motion patterns are categorized into distinct classes with feature descriptors giving the global spatial distribution of the dense horizontal and vertical optical flow between emotional and neutral facial images. Classification scheme based on multi-label support vector machine has been employed for categorization. Promising results are obtained from the comparative analysis of the technique presented on the Extended Cohn Kanade database with other previously proposed methods.