Facial expression recognition using fuzzified Pseudo Zernike Moments and structural features

Facial expression recognition (FER) is an important part of emotional computing that can be useful in many applications for people's behavior analysis. Recently, some methods have been suggested to recognize facial expressions, but they do not offer a strong approach to facial expression recognition. In this paper, we propose a fuzzy-based approach that incorporates two different types of features to increase the recognition rate of facial expression. These features include locally weighted Pseudo Zernike Moments (LWPZM) and structural features (mouth and eye-opening, teeth existence, and eyebrow constriction). To classify facial expressions, the proposed fuzzy inference system uses fuzzified features. The performance of our proposed method has been assessed using the well-known RaFD database. The experimental results show that the proposed method is not only robust in terms of age, ethnicity, and gender changes that would make our contribution, but also improve the recognition rate of facial expression compared to several state-of-the-art methods.