Exploring Various Aspects of Gabor Filter in Classifying Facial Expression

Abstract

Facial expression detection is a well-studied domain in which facial features are extracted and then classified into six common expressions. One of the most common techniques used for extracting features is the Gabor filter. In literature, for extracting the features, the combined magnitude and phase values of the Gabor filter are used. This paper is exploring the performance of methods using the combined filtering method, using magnitude alone and using phase alone in the domain of facial expression detection. It is observed that considering phase values with the support vector machine classifier yielded an additional 8% accuracy when compared to combined methods.