Fusing Geometric Features for Skeleton-Based Action Recognition using Multilayer LSTM Networks

Recent skeleton-based action recognition approaches achieve great improvement by using RNN models. Currently these approaches build an end-to-end network from coordinates of joints to class categories and improve accuracy by extending RNN to spatial domains. First, while such well-designed models and optimization strategies explore relations between different parts directly from joint coordinates, we provide a simple universal spatial modeling method perpendicular to the RNN model enhancement. Specifically, according to the evolution of previous work, we select a set of simple geometric features, and then seperately feed each type of features to a 3-layer LSTM framework. Second, we propose a multi-stream LSTM architecture with a new smoothed score fusion techinique to learn classification from different geometric feature streams. Furthermore, we observe that the geometric relational features based on distances between joints and selected lines outperform other features and the fusion results achieve state-of-the-art performance on four datasets. We also show the sparsity of input gate weights in the first LSTM layer trained by geometric features and demonstrate that utilizing joint-line distances as input require less data for training.