Facial Expression Recognition using Locally Linear Embedding with LBP and HOG Descriptors

Facial expression recognition intervenes in various fields of applications such as human-computer interaction. Despite the fact that several methods are regularly proposed, designing an efficient automatic facial expression recognition method remains challenging. In this paper, we propose a method through which we compare the performance of two common and well-known image descriptors namely Local Binary Patterns and Histogram of Oriented Gradients. Both are used by two distinct manners; global which uses the whole face while the local exploits predefined sub-regions. Moreover, we employ a specific dimensionality reduction technique namely Locally Linear Embedding. As for the recognition part, we choose to employ a multiclass Support Vector Machine classifier for its generalization capabilities in order to recognize the expressed emotion among the six basic ones. Finally, we assess the performances of the proposed method using three different and common datasets namely KDEF, JAFFE and RafD. The obtained results are promising with corresponding recognition rates; 85.48%, 96.05% and 93.54%, respectively.