In most applications of touch based humancomputer interaction, multi-touch gestures are used for directlymanipulating the interface such as scaling, panning, etc. In thispaper, we propose using multi-touch gesture as indirectcommand, such as redo, undo, erase, etc., for the operatingsystem. The proposed recognition system is guided by temporal,spatial and shape information. This is achieved using a graphembedding approach where all previous information are used.We evaluated our multi-touch recognition system on a set of 18different multi-touch gestures. With this graph embeddingmethod and a SVM classifier, we achieve 94.50% recognition rate.We believe that our research points out a possibility ofintegrating together raw ink, direct manipulation and indirectcommand in many gesture-based complex application such as asketch drawing application.