Skeleton-Based Action Recognition Based on Distance Vector and Multihigh View Adaptive Networks

Table 1

Results on the NTU-RGB + D datasets.

Methods	X-Sub (%)	X-View (%)
HBRNN-L [9]	64.0	59.1
Part-aware LSTM [26]	70.3	62.9
Trust gate ST-LSTM [27]	77.7	69.2
Two-stream RNN [28]	79.5	71.3
STA-LSTM [29]	81.2	73.4
Ensemble TS-LSTM [30]	81.3	74.6
Visualization CNN [11]	82.6	76.0
VA-LSTM [21]	87.6	79.4
ST-GCN [14]	88.3	81.5
SR-TSL [22]	92.4	84.8
HCN [6]	91.1	86.5
PB-GCN [23]	93.2	87.5
AGC-LSTM [31]	95.0	89.2
Ours	95.3	89.5