A View-invariant Skeleton Map with 3DCNN for Action Recognition

Skeleton based human action recognition is a challenging task in the field of computer vision, which is easily affected by variant camera viewpoint. A novel view-invariant feature is proposed in this paper. We encode the spatial-temporal information of skeleton joint points sequences into a view-invariant skeleton map (VISM), and employ a 3D convolutional neural network (3DCNN) to exploit features from VISM for 3D action recognition. Experimental results on the NTU RGB+D dataset have demonstrated the efficacy of the proposed method, and the robust to the variant viewpoint problem.