Efficient Expression Recognition Based on Shape-indexed Features
Ying Wang, Zhang Guoying · 2018
An efficient and accurate facial expression recognition method is proposed based on shape-indexed features and random decision forests. Different from traditional methods that extract features from image coordinate system, the proposed method extracts features based on the location of facial landmarks, which are simple but more robust against pose and inter-personal variations. Assisted with random decision forests that are efficient and less prone to over fitting than decision tree, this method achieves real-time expression recognition and the accuracy is superior over the state of the art.