Enhanced Depth Motion Maps for Improved Human Action Recognition from Depth Action Sequences

Human action recognition based on depth action sequences is a well-known study that can be used in many fields.Compared to RGB Videos, depth action videos are more robust as they are not affected by changes in lighting.This study proposes the Enhanced Depth Motion Map (EDMM), a new action descriptor to overcome the challenges of the conventional DMM, which cannot handle the presence of some undefined regions in depth maps.Contrary to DMM's global motion description, the EDMM meticulously scans individual pixels and then accurately identifies those in motion.We extracted the EDMM from a series of video sequences and then used a convolutional neural network (CNN) model to simplify the motions accurately.The CNN model, equipped with nine layers, accurately recognizes activities based on maximum movement similarity.The method underwent testing using two standard and publicly available datasets; MSR Action 3D and UTD-MHAD.The test results through True Positive Rate (TPR), Positive Predictive Value (PPV) or Precision, False Discovery Rate (FDR), False Negative Rate (FNR), F1-score, and accuracy demonstrated the superiority of the proposed method over numerous state-of-the-art methods like DMM, DMM with Local Binary Pattern and DMM with Histogram of Oriented Gradients (HOGs).