Table 1.

The proposed model performance comparison with different input data setups.

No	Input Setup	Accuracy
1	3DCNN + LSTM + RAW RGB data without sequence alignment	73%
2	3DCNN + LSTM + RGB data with background removal + body + sequence alignment	90%
3	3DCNN + LSTM + RGB data only focusing on hand + sequence alignment	95.8%
4	3DCNN + LSTM + Depth data only focusing on hand + sequence alignment	94.4%