In order to further improve the rate of action recognition based on human skeleton and break the limitation that the inputs of most methods based on deep learning are human joint coordinates

a skeleton-based action recognition algorithm combining geometric features with long short-term memory（ LSTM） network is proposed. The geometric features based on the distances between joints and selected lines are selected as the input of the network. Then

a simple network structure based on LSTM

time-concerned LSTM network

is introduced. By utilizing the ability of time-concerned LSTM network to focus on the most recognizable frames

99. 25 % and98. 79 % recognition rates are achieved on SBU Interaction dataset and UT Kinect dataset

respectively. The experimental results show that the method is effective for skeleton-based action recognition.