MDTP: a novel multi-directional triangles pattern for face expression recognition

Abstract

Face expression recognition is a key-subject of machine learning, and the primary issue on it is lack of accuracy. This paper proposes a novel face expression recognition method using Decision Based Rule-Oriented Median Filter (DBROMF) and Multi-Directional Triangles Pattern (FER-MDTP). The DBROMF is a novel noise reduction method which removes the impulse noise from the facial images. This FER method enriched with a new image descriptor (MDTP) which is structured via multi-directional triangle pattern to provide a superior image description. An unparalleled novel algorithm to locate the human face organ viz. lip and eyeball stuffed in this research by getting assistance with fuzzy edge strength and abbreviated as an MDTP-FES method. These landmarks of features extracted, and the histogram oriented features tailored with the classification division. The support vector neural network classifier (SVNN) is integrated to conduct the classification job. The JAFFE, CK, TFEID, and ADFES databases are linked to perform the simulation which telescoped with six face expressions. The proposed method lifts the accuracy to a significant range than the existing state-of-the-art methods.