作者: Juxiang Zhou , Yunqiong Wang , Tianwei Xu , Wanquan Liu
关键词:
摘要: Curve let transform has been recently proved to be a powerful tool for multi-resolution analysis on images. In this paper we propose new approach facial expression recognition based features extracted via curve transform. First is presented and its advantages in image are described. Then the coefficients of selected scales angles used as analysis. Consequently Principal Component Analysis (PCA) Linear Discriminate (LDA) reduce optimize features. Finally use nearest neighbor classifier recognize expressions these The experimental results JAFFE Cohn Kanade two benchmark databases show that proposed outperforms PCA LDA techniques original pixel values well counterparts with wavelet