Facial expression description and recognition based on fuzzy semantic concepts

This paper proposes a novel approach to identifying various expressions using semantic concepts. Based on the framework of the axiomatic fuzzy set, facial features are transformed into semantic concepts, which are then considered as a ruleset to differentiate expression categories. This method has two main advantages. First, it bridges the descriptors between image features and semantic concepts, according to which facial geometric features can be mirrored directly. Second, it alters the description patterns of fuzzy rulesets, which can reduce the dimensionality of expression features. We establish optimization criteria for selecting salient semantic concepts to represent expression characteristics. Experiments are conducted using the proposed method on the FEI and CK+ databases. Semantic concepts are considered as a ruleset to describe the differences between various expressions. The performances of state-of-the art classifiers and the proposed method are compared and analyzed. The results demonstrate that the proposed method provides excellent interpretability and classification performance for facial expressions.