The nonlinear Lamb wave based damage detection method has become a focus recently. However, the extraction and process on expected mode signal were usually difficult due to the dispersive and multi mode characters of Lamb wave. Time frequency process technique and dispersive curves were adopted to develop the extraction method of nonlinear Lamb wave in active damage monitoring which was also validated experimentally. The results indicate that the developed method could distinguish and extract the expected modes effectively. Further simulated damage monitoring experiments validated that nonlinear characteristic signals were sensitive to the structural variation.