Journal of the Chinese Institute of Engineers (JCIE)

Efficient human skeleton-based action recognition using a two-stream LSTM framework with limited datasets