A fuzzy-convolution model for physical action and behaviour pattern recognition of 3D time series

In this paper an innovative model architecture is proposed based on the fusion of a multi auto-adjusted TSK fuzzy logic classifier and a signal convolver classifier to model physical actions and behaviours without any giving prior knowledge of the modeled activities. Three different hypotheses are being tried to investigate such as the classification accuracy of 3D time series activity data, the discrimination clarity of a novel convolution classifier, and a vast number of experimental testing to tune the classifiers´ internal structure by revealing optimal configuration attributes such as features, distances, and functions. The fuzzy-convolution model is being used by a mobile robot for remote surveillance within a smart environment. The hardware configuration incorporates an ubiquitous 3D marker based tracker which establishes an interface between the robot and the actor. The data form is time series based which is fetched to the robot throughout an off-line process.