Helical hinges are common in transmembrane (TM) proteins. Hinging motions of TM helices are associated with the functional dynamics of many important membrane proteins such as various G-protein-coupled receptors and potassium channels. Although coarse-grained (CG) simulations are useful for studying membrane proteins, they are limited in describing accurately the hinge-related protein dynamics. In this study, we have overcome this limitation through a further development of the MARTINI CG model. The key improvement lies in implementation of additional local structural types and parameterization of the model against a structural library of TM helices obtained from the Protein Data Bank. Through simulations of 12 membrane proteins, we have demonstrated that the improved model not only accurately describes the local hinge structures of these proteins but also reproduces the overall structures of TM domains better than the original MARTINI is able to do. Furthermore, we show that the improved model, when combined with an elastic network, can now be used to explore the deactivation of the human β2 adrenergic receptor and the gating motions of the KcsA potassium channel. The important details of these hinge-associated transitions captured in our simulations agree well with previous experiments and all-atom simulations, and it was concluded that the improved model shows promise as a useful tool for the study of functional dynamics of membrane proteins.