Mixed graph convolution and residual transformation network for skeleton-based action recognition | Litlas