Detection of Human Behavioral Events Combining Perceptual Causality
Wei Tian, Jingyuan Lv · 2021 IEEE Asia-Pacific Conference on Image Processing, Electronics and Computers (IPEC) · 2021
In this paper, we propose a computational model of human behavior events analysis that integrates perceptual causality. Design and establish a causal rule representation method based on default logic, which provides a concise syntax and semantic formal tool for potential causality. Through the analysis of the internal structure of the event, the technical framework of causality with the smallest action unit as the entity is established, which makes the traditional different entities, objects, events, states and actions naturally interrelated. By establishing the perceptual causal relationship between human goals and objects, events, reasoning the intention of human goals, and predicting their behavior, it can effectively solve the problem of human action occlusion and action error detection. The effectiveness of this method is evaluated on a new video dataset.