Deep Learning-Based Action Recognition

• Format: eBook
• Language: English
• Published: MDPI - Multidisciplinary Digital Publishing Institute 2022
• Subjects: History of engineering & technology; Technology: general issues; 3D skeletal; 3D-CNN; action recognition; activity recognition; artificial intelligence; class regularization; class-specific features; CNN; continuous hand gesture recognition; convolutional receptive field; data augmentation; deep learning; dynamic gesture recognition; Dynamic Hand Gesture Recognition; embedded system; feature fusion; feedforward neural networks; fusion strategies; gesture classification; gesture spotting; graph convolution; hand gesture recognition; hand shape features; high-order feature; human action recognition; human activity recognition; human-computer interaction; human-machine interface; Long Short-Term Memory; multi-modal features; multi-modalities network; multi-person pose estimation; n/a; partition pose representation; partitioned centerpose network; pose estimation; real-time; spatio-temporal differential; spatio-temporal feature; spatio-temporal image formation; spatiotemporal activations; spatiotemporal feature; stacked hourglass network; transfer learning
• Online Access: Open Access: DOAB: description of the publication; Open Access: DOAB, download the publication

 The classification of human action or behavior patterns is very important for analyzing situations in the field and maintaining social safety. This book focuses on recent research findings on recognizing human action patterns. Technology for the recognition of human action pattern includes the processing technology of human behavior data for learning, technology of expressing feature values ​​of images, technology of extracting spatiotemporal information of images, technology of recognizing human posture, and technology of gesture recognition. Research on these technologies has recently been conducted using general deep learning network modeling of artificial intelligence technology, and excellent research results have been included in this edition.