Abstract: The flutter-derivative identification of bridge decks can be converted into a least-square optimization problem. This problem is solved using the presented gradient declining algorithm (GDA). The 18 flutter derivatives of bridge deck are extracted subsequently. For GDA, the feedback mechanism is introduced into the stochastic search progress, by which the optimum solution can be searched rapidly. The GDA is applicable to the system parameter identification, and the satisfactory precision can be ensured. The 18 flutter derivatives of Suramadu Bridge deck are identified using GDA, and compared with the results extracted by stochastic subspace identification (SSI) technique. The reasons for poor stability of flutter derivatives at higher wind speed and relative unsatisfactory precision of lateral flutter derivatives extracted from the free vibration method with the existent spring suspension system are offered. For the identification precision of flutter derivatives, experiment procedure is more important than the extraction approach.