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