Open-view human action recognition based on linear discriminant analysis