Nonnegative Tensor-Based Linear Dynamical Systems for Recognizing Human Action from 3D Skeletons

Table 4

Recognition rates on the G3D-gaming dataset.

Approach	Recognition Accuracy
GB-RBM+HMM[19]	86.40
SE[12]	91.09
SO[20]	87.95
LieNet[21]	89.10
tLDS[6]	90.60
Our approach	92.56