3D Industrial Human Action Recognition Based on the Synergy Between Convolutional Neural Networks and Long Short-Term Memory

This paper illustrates a significant contribution to Human Activity Recognition (HAR) within industrial environments. We propose a deep learning model that combines Long Short-Term Memory (LSTM) and Convolutional Neural Networks (CNN) with Recurrent Neural Networks (RNN), offering a superior method for processing a novel and multicamera dataset. The proposed LSTM-Conv1D and RNN models achieved classification accuracies of 92% and 87%, respectively, illustrating the effectiveness of the LSTM-Conv1D architecture in recognizing human actions with higher precision compared to the RNN model. This paper focuses on the HAR based on generating a comprehensive dataset with joints cartesian position data for 3D motion analysis, specifically emphasizing the upper segments within macro action movements.