Abstract: In this paper,we present a method of facial expression recognition based on weighted clustering of grouped features.At first,we design a feature-grouping method for normalizing feature values to account for different effects of the features for the expression classification.Secondly,a fuzzy clustering algorithm by weighting the grouped features is introduced to tackle the expression uncertainty.The method is computationally simple,easy to implement,and can be used for dynamic updating for an on-line training and recognition system.