Abstract

Background and Objective: Video-based human activity recognition is a prominent area of research due to a wide range of applications from intelligent video surveillance to human-computer interaction. Recent work on video analysis is focused on applying deep learning approach to accomplish the task of activity recognition.

Conclusion: Deep networks can dramatically improve the recognition performance because of its hierarchical nature to exploit the video frame structure in reducing the search space of the learning model. This motivated us to provide a comprehensive survey of the state-of-art deep models for recognizing human actions/activities.

Keywords: Activity recognition, convolution, deep learning, handcrafted features, spatio-temporal dimension, algorithams.

Graphical Abstract