Face Expression Recognition Using Weber Local Descriptor and F-RBFNN
I. Michael Revina, W. R. Sam Emmanuel · 2018
Face Expression Recognition (FER) is the challenging and interesting task in image processing and achieved by human or computer. This paper proposes the FER system is developed by using the combined approach of Weber Local Descriptor (WLD) and Firefly excited Radial Basis Function Neural Network (F-RBFNN). The WLD contains two phases such as image separation and feature extraction. In the first phase, the broad pools of sub-areas are formed by separating the entire face image into non-overlapping. In the second phase, from the large pool of sub-areas having one subarea as the center. The WLD histogram features are extracted by selecting the similar group of area. F-RBFNN classifier is used for recognition of facial expressions. In F-RBFNN, the natural associate groups of training face expression images due to differences in illumination are acquired by using the firefly algorithm and mainly for selection of centers for RBFNN. The WLD-F-RBFNN base FER system recognizes the six expressions such as disgust, sad, smile, surprise, fear and anger. The proposed WLD-F-RBFNN method is established with the facial expression database namely JAFFE and achieves improved performance than other existing methods.