Facial Expression Learning Simulation：面部表情的学习与仿真.pptVIP

Facial Expression Learning Simulation Company Logo Catalog What we do? How we did it? What can we do to improve? What we do? Our project can separate into five phases 1) build a 3-D face model-Avatar 2) Face Landmark Extraction 3) attached the human face image on to Avatar(Now Avatar has a face!) 4) change Avatars facial expression by altering prameters(Now avatar has feeling!) 5) Avatar mimics Peter’s face. How we do it? build a 3-D face model-Avatar CANDIDE-3 Model, which has 113 Vertexes and 184 triangle surfaces How we do it? Face Landmark Extraction use Violas harr-like featureadaboost for face detection use ASM model to build human face model and locate landmark feature points ASM Active shape models (ASMs) are statistical models of the shape of objects which iteratively deform to fit to an example of the object in a new image Build model manully record landmark feature-58 align all the model into same coordinate PCA 　　 ASM Testing roughly position of model is obtained, find good fitting point for every current model point multi-resolution ASM(3 level) minimize: update the transformation parameter to fit the model to the new point set let b remains to be constrained see if b converged ASM How we did it? attached the human face image on to Avatar(Now Avatar has a face!) Two Catalogs: 1) Nose and Chin Determine the position and size of head Use Similarity transform + Least Square Global Changing 2) Mouth, Eye, and Eyebrow Determine the characteristic of face Use Box Matching --- mean location,width,height Local Changing Test Result How we did it? change Avatars facial expression by altering prameters(Now Avatar has feeling) Mesh Model --- pre-store the way of changing vertex in file (Action Unit Vertex, AUV) 11 different Expression controlling by 11 AUV 1)----Upper lip raiser 2)----Jaw drop 3)----Lip stretcher 4)----Brow lower 5)----Lip corner depressor Etc. How we did it? Avatar mimics Peter’s face We try