This study develops a new inverse analysis technique for estimating damage patterns in notched composite laminates using an embedded fiber Bragg grating (FBG) sensor. The damage pattern near the notch was investigated by using a layer-wise finite-element model with cohesive elements for cracks and delamination. The reflection spectrum of the FBG sensor was simulated from the strain distribution obtained along the gage length. We approximated the typical damage pattern (splits, transverse cracks, and delamination) near the notch using some variables; these were then estimated through mathematical programming by minimizing the differences of spectrum shapes between the inputs (e.g. experiments) and the estimations. This damage identification procedure was applied to numerical examples, i.e. simulation results, for a notched CFRP cross-ply laminate. The damage patterns estimated from the calculated reflection spectra were almost identical to those obtained by the simulation. Finally, we estimated the damage pattern of a notched laminate from a measured reflection spectrum of an embedded FBG sensor and demonstrated the capability of the developed identification procedure for practical applications.