Robust switching detection and active mode recognition are addressed for switched nonlinear systems. The nonlinear modes are represented by Takagi- Sugeno (TS) models where the weighting functions are supposed to be online computable. The proposed switching detection technique and the active mode recognition method lie on indicator signals that are generated using a datadriven projection technique. The main advantage of this method is that it does not need the knowledge of local model parameters values. Only input-output data and the weighting functions of the TS model are used to generate these indicator signals. A numerical example is provided to illustrate the proposed approach.