An On-Line Classification Approach of Visitors' Movements in 3D Virtual Museums

Overview & Synopsis

Recommender systems, in virtual museums and art galleries, providing the personalization and context awareness features require the off-line synthesis of visitors' behaviors therein and the off-line training stage of those synthetic data. This paper deals with the simulation of four visitors' styles, i.e., ant, fish, grasshopper, and butterfly, and the classification of those four styles using an Adaptive Neuro-Fuzzy Inference System (ANFIS). First, we analyze visitors' behaviors related to a visit time and an observation distance. Then, the proposed synthesis procedure is developed and used in the off-line training stage of ANFIS. The training and testing data are the average and variance of a set of visitors' attention data computing by the proposed function of the visit time and observation distance variables. Therefore, the trained ANFIS can identify the behavior style of an on-line visitor using the training set of synthetic data and its memberships can describe degrees of uncertainty in behavior styles.