ABSTRACT
The distinction of facial feature and section semantic is derived from face image recognition technology, which can be divided into tensor space and facial feature location method. The classical methods of PCA and LDA extracting image features and practicing recognition through analyzing the latitude variation of face image in tensor space, though with better precision, omit the geometrical characteristics of human face, and are complicated in calculation. Facial feature location method analyzes geometrical morphology and changes rule by detecting the facial features and utilizing topology of each feature. However its weakness lies on its less explanation on feature data. It is of higher superiority as for fuzzy rule-based method on tackling with the above problem, its models are widely used on data analysis because of its priority on explaining the data attribute, and it is easy to be understood. But on the contrary, excessive rules would decrease the interpretability on data; meanwhile they increase the difficulty in model establishment and the liability to fitting phenomenon, which would affect the
accuracy. Axiomatic Fuzzy Set[3-5] is a new method proposed on base of fuzzy rule, which can establish the fuzzy assembly according to the original assignment and fuzzy semantics. In the framework of AFS theory, based on the subjective and objective uncertainty, the information included in the data store is transferred from subset membership to logic operator. All the above will be helpful to meet human recognition by making description on data.
