بازشناسی جلوه‌های هیجانی با استفاده از تحلیل تفکیک پذیری مبتنی بر خوشه بندی چهره

Improvement of Facial expression recognition is aim of proposed method. This is a new formulation to the linear discriminant analysis. In the new formulation within-class and between-class covariance matrix are estimated on the each cluster and in the test phase new samples are mapped to the subspace that is related to the cluster of them. At the first we addressed clustering analysis of faces and three criteria are proposed for clustering of them. Then cluster based discriminant analysis is achieved through the each three clustering approach. Results show that recognition rate is increased by this new approach and the best result is related to clustering based on the facial index that performance of basic system increased from 95.75% to 98.66%. This is an efficient technique to encounter to large scale dataset in facial expression recognition filed