Mining Key Skeleton Poses with Latent SVM for Action Recognition

Table 4

Comparison of our method with the others on AS1, AS2, and AS3.

Action subset	Li et al. [11]	Xia et al. [24]	Yang and Tian [25]	Ours
AS1	72.9%	89.8%	80.5%	89.1%
AS2	71.9%	85.5%	73.9%	88.7%
AS3	79.2%	63.5%	95.5%	94.9%
Average	74.7%	79.6%	83.3%	90.9%