Neutrosophic Concept Lattice based Approach for Computing Human Activities from Contexts

Abstract

Abstract Complex human activity recognition suffers from ambiguity of interpretation problem. A novel neutrosophic formal concept analysis method has been proposed to quantify non-determinism leading to ambiguity of interpretation and utilize it in activity recognition. The method works by penalizing performance of non-deterministic activities and rewarding the deterministic ones. Thus, non- deterministic activities are identified during testing due to significantly reduced performance and contexts can be redesigned to improve their description. The proposed method has been implemented on benchmark dataset having both types of activities. Our approach successfully identified nondeterminism in activities description without compromising recognition performance of deterministic activities. It has also been shown that other approaches fail to identify non deterministic activities. Overall accuracy of activity recognition of our approach was comparable to other approaches.