Recognizing Actions in 3D Using Action-Snippets and Activated Simplices