: Facial expression recognition helps human-machine interaction, machine can behave better by using human emotions. In this project, we use if-then rules and type-2 fuzzy logic for facial expression recognition. If-then rules helps machine to perform facial expression recognition process similar to human behavior. Human face can have mixture of multiple expressions at the same time. Due to this uncertainty, we use fuzzy logic to model imprecise knowledge. We will face two types of uncertainty, intra-personal and inter-personal, which can be modeled by type-2 fuzzy logic. In this project, we present two methods for facial expression recognitions. First proposed method is a mamdani fuzzy inference system which its membership functions are interval type-2 fuzzy sets. We developed a mamdani type-1 fuzzy inference system, then convert it to a mamdani type-2 fuzzy inference system. Genetic algorithm is used to optimize membership function parameters of these systems, then these two systems has been compared. To experiment these systems, Cohn-Kanade and JAFFE databases is used. The other proposed method is a sugeno neuro-fuzzy system which membership functions of its inputs are interval type-2 fuzzy sets. We developed a sugero neuro-fuzzy system with type-1 membership function to compare with our method. Back-propagation algorithm is used for optimizing these systems’ membership function parameters. Fuzzy logic rules of these systems are extracted experimentally. JAFFE database is used to experiment these systems. Experiments represent in both methods, type-2 systems have better recognition rate compared with corresponding type-1 systems. Keywords:Facial Expression Recognition, Type-2 Fuzzy Sets, Genetic Algorithm, Back-propagation
