Application of Improved Long-short-term Memory Network in Human Morphology Detection
Ming Huang, Tao Wen, Xu Liang · 2019
With behavior recognition technology playing an increasingly important role in many fields such as intelligent monitoring, human-computer interaction, video sequence understanding, medical and health care, researchers are trying to improve the accuracy of behavior recognition technology. By combining LSTM with GRU, this paper simplifies the network structure on the basis of retaining the basic idea of LSTM (forgetting and updating mechanism), and uses update gate to make each unit learn long-term and short-term characteristics, thus reducing the risk of gradient dispersion. Experiments show that this optimization method has better resolution than traditional LSTM.