Facial Expression Pervasive Analysis Based on Haar-Like Features and SVM

Abstract

The pervasiveness and applicability of computational models is important for facial expression analysis within different races. In this paper, an ingenious approach is proposed to verify the pervasiveness of facial expression model based on Haar-like features, Adaboost and SVM. The approach is constructed with three characteristics, including facial features extraction with Haar-like features, weak classifiers training with Adaboost, and facial expression recognition with SVM. The contribution is that the approach above can be applied to distinguish the differences of identified facial expression models within two human racial groups. Experiments are conducted on both eastern and western facial expressions recognition on the basis of a facial feature template being trained by western face databases. Results show that high accuracy rate is achieved from the same race template, while a general accuracy rate for the other senses. It is indicated that the approach improves the pervasiveness of facial expression model with high recognition accuracy and verifies the computational differences of facial expressions between different races.