AFE-CNN: 3D Skeleton-based Action Recognition with Action Feature Enhancement

Existing 3D skeleton-based action recognition approaches reach impressive performance by encoding handcrafted features to image format and decoding CNNs. However, such methods are limited in two ways: a) the difficult handle challenging actions, b) they generally require complex CNN models improve accuracy, which usually occur heavy computational burden. To overcome these limitations, we introduce a novel AFE-CNN, devotes enhance of actions adapt actions. We propose feature modules from key joint, bone vector, frame temporal perspectives, thus AFE-CNN is more robust camera views body sizes variation, significantly accuracy on Moreover, our adopts light-weight model decode images with enhanced, ensures much lower burden than state-of-the-art methods. evaluate three benchmark datasets: NTU RGB + D, D 120, UTKinect-Action3D, extensive experimental results demonstrate outstanding AFE-CNN.