Two-Person Mutual Action Recognition Using Joint Dynamics and Coordinate Transformation

Abstract

. Skeleton-based action recognition has attracted lots of attention in computer vision. Human mutual interaction recognition relies on extracting discriminative features for better understanding details. In this work, we propose two vectors to encode joint dynamics and spatial interaction information. The proposed model shows remarkable performance at handling sequential data. Experimental results demonstrate that our model outperforms state-of-the-art approaches with much less overheads.