Facial Expression Recognition Research Based On Elastic Graph Matching And Hausdorff Distance

Facial expression recognition is a challenging topic in psychology, physiology,image procession, machine vision,pattern recognition and other areas.In order to meetthe application system of facial expression recognition in real life,more and moreresearchers began to study the algorithm of facial expression recognition. Theaccuracy and effectiveness of emotional feature extraction and classification is the keyto the success of facial expression recognition system.Although there have been a lotof facial expression recognition technologies emerge,but for facial expressionrecognition under the target block, image noise and camera shake and otherconfounding factors not yet ripe.In order to solve the above problems,this papermakes a deep research on facial expression feature extraction and classification undermultiple confounders,proposes a classification algorithm of elastic graph matchingand improved Hausdorff distance similarity calculation combining.We have finishedthe following main task:Firstly,emotional feature extraction of facial expression image.Facial expressionsemotional feature extraction directly affects the subsequent emotional classification.This paper mainly uses multiple scales and multiple directions of Gabor filters toextract facial expression feature. Because of the Gabor wavelet is not sensitive tolight,and to a certain degree of image rotation and deformation has very goodrobustness.Secondly,facial expression image classification.By Using the advantage of faceclassification by flexible template matching and the advantage of similaritycalculation by improved Hausdorff distance in the noise environment,This paperpresents the application of elastic graph matching and improved Hausdorff distancesimilarity calculation algorithm combining.It is testified by experiments that therecognition rate is better than traditional classification algorithm through this method.Finally,design of facial expression recognition classification system.Based on theimproved algorithm proposed in this paper, developing a set of facial expressionrecognition system basing on the JAFFE facial expression database and Cohn-Kanade facial expression database,and compared the improved algorithm andtraditional algorithm through experiments.The experimental analysis shows that under the interference environment the method proposed in this paper is better than traditional algorithm on recognitionrate.