An Application of Optical Flow - Extraction of Facial Expression -

This report discusses a method that uses optical flow to estimate facial muscle movements which could then be recognized these movements as facial expressions. The. human face has several features such as eyes, mouth and nose. Their movements and deformations are due to the contraction and/or relaxation of the facial muscles. Facial skin has the texture of a finegrained organ and this helps in extracting the optical flow. We estimate the movement of major facial muscles from the optical-flow field. We evaluate the muscle movement in each muscle window which defines one primary direction of muscle contraction. Finally, we get velocity patterns in terms of time. We will use the feature patterns to recognize facial expressions by means of pattern matching. Our experiments show that the features acquired by this approach are sufficient to recognize certain expressions.