Recognition of facial expressions extracting salient features using local binary patterns and histogram of oriented gradients

摘要

The extraction of accurate facial landmarks play an exceedingly significant role for recognition of facial expressions. This paper proffers an approach in order for recognition of varied expressions of the faces by incipiently implementing the LBP (Local Binary Pattern) descriptor on the facial landmarks and then the HOG (Histogram of Oriented Gradients) descriptor is applied on the extracted LBP features and finally they are classified into seven different expressions of the face: anger, happiness, surprise, disgust, sadness, neutral and fear adopting the Multiclass support vector machine (SVM). The structured model implemented on the Multimedia Understanding Group (MUG) and the Japanese Female Facial Expression (JAFFE) databases signifies 96% and 97.1% rate of recognition respectively and found to be functioning considerably well under diverse illuminating variation conditions.