3D Motion Recovery via Low Rank Matrix Restoration with Hankel-Like Augmentation

This paper proposes a 3D skeleton recovery model equipped with a joint augmented low-rank and sparse prior and an articulation-graph-based isometric constraint to exploit temporal and spatial correlation, respectively. The corrupted 3D skeleton sequence is represented as a matrix and constrained by several priors in the proposed model. A Hankel-like augmentation is adopted to strengthen the low-rankness and we integrate a decoupling technique to reduce the internal interferences of the data. We solve our model via an alternating direction method under the augmented Lagrangian multiplier framework with a Gauss-Newton solver for the subproblem of isometric optimization. Experimental results on two skeleton datasets demonstrate the effectiveness and superiority of the proposed model in motion reconstruction and skeleton recovery, compared with state-of-the-art methods.