Human Action Recognition based on GMM-UBM supervector using SVM with non-linear GMM KL and GUMI

摘要

In recent years, Human Action Recognition (HAR) has attracted much attention from the research community due to its challenges as well as wide applications. In this paper, we investigate GMM supervector based Universal Background Model (UBM) and Support Vector Machine (SVM) with dense trajectories and motion bound features for HAR system. A GMM supervector is obtained by adapting with UBM and cascading all the mean vector components. After that, supervectors are applied as input features to SVM classifier. Moreover, we also adopted two modified GMM KL and GUMI kernels in this research. Then we make a comparison and critical analysis of our method with previous systems. Experimental results demonstrate that the proposed approach performs more efficient than current systems.