Real-time state estimation and parameter identification of physical motion are the basic modules for automatic control system. However, the high-order motion information of nonlinear dynamic system is always hard to be achieved. In order to realize the real-time state estimation, a data-driving method of state estimation is investigated by combing the fast Fourier transform and the sliding recursive least squares method. By analyzing the frequency domain of the sampled data, the frequency and amplitude of the multi-period coupled motion are extracted, based on which a reasonable real-time state estimation method is constructed. The suggested method can overcome the impact of multi-period coupled motion character in navigation algorithm. Compared with traditional methods such as Kalman filtering or averaging value, the suggested method has better performance in navigation system.