An Evaluation of Local Binary Descriptors for Facial Emotion Classification

Abstract

Feature descriptors are vitally important in the broad domain of computer vision. In software systems for face recognition, local binary descriptors find wide use as feature descriptors. Because they give more robust results in varying conditions such as pose, lighting and illumination changes. Precision depends on the correctness of representing the relationship in the local neighbourhood of a digital image into small structures. This paper presents the performance analysis of various binary descriptors such as local binary pattern (LBP), local directional pattern (LDP), local directional number pattern (LDNP), angular local directional pattern (ALDP), local optimal-oriented pattern (LOOP), support vector machine (SVM), K-nearest neighbour (KNN) and back propagation neural network (BPNN) are used for emotion classification. The results indicate that ALDP + Polynomial SVM on MUFE, JAFFE and Yale Face databases gives better accuracy with 96.00%, 94.44% and 89.00%, respectively.