Human Action Recognition Based on Projection and Mass Center Movement Features

This paper proposes a simple and effective human action recognition algorithm based on projection and mass center movement features. Firstly, divide the original video into equal length subsequences with overlapping time window, and make moving parts detection using adjacent frame difference. Then, horizontal projection and vertical projection of binary image are made, and in order to get ride of redundant information and reduce computational burden, the principal components are extracted by PCA from a series of projection features in equal length subsequences with overlapping time window. Finally these projection features are combined with mass center movement features to identify the human action, and KNN is adopted to classify. Online video library is used in final experiments, and the experimental results show that this strategy can achieve good performance in terms of recognition rate, computational cost.