Facial Expression Recognition Based on Subregion Weighted Fusion and LDA

The facial expression recognition occupies an increasingly important position. In this paper, the static emoticon images are analyzed, and the geometrical priori based weighting strategy is used for feature fusion. The disadvantages of the preliminary features extracted by the traditional methods lack the discriminant characteristics. An improved Linear Discriminant Analysis (LDA) algorithm based on intraclass divergence matrix correction is proposed. The data is sent to the Generalized Regression Neural Network (GRNN) classifier for identification and has achieved good results.