The automatic distinction (domain separation) between handwriting (textual domain) and freehand drawing (graphical domain) elements into the same layer is a topic of great interest that still requires further investigation. This paper describes a machine learning based approach for the online separation of domain elements. The proposed approach presents two main innovative contributions. First, a new set of discriminative features is presented. Second, the use of a Support Vector Machine (SVM) classifier to properly separate the different elements. Experimental results on a wide range of application domains show the robustness of the proposed method and prove the validity of the proposed approach