Static Human Facial Expression Recognition Using Facial Components Detection And Hog Features

Abstract

An appropriate approach to resolve the limitation of facial expression detection is introduced in this research paper. The system detects nose and ears, hair, brows and mouth components. Considering that facial expressions are a function of muscle or deformations, and that Oriented Gradient (HOG) histograms are very sensitive to things, the researcher prefers to use the HOG for the encryption of those facial parts. A linear SVM is then educated in the classification of face expression. In the JAFFE data set combined with an expanded CHANADE dataset, the investigator seeks to evaluate the expected technique. The average values for both datasets amount to 94.3% and 88.7% respectively. Our proposed approach demonstrates the competitive accuracy of classification by experimental results.