The frequency response function (FRF) estimation of the shaking system is adopted without any modification in the traditional process of vibration test. For the time-variant characteristic of the vibration control system, an adaptive method of FRF estimation is described in this paper. This strategy adjustes the variation of the system in every control cycle, which is based on the least me an square(LMS) algorithm, and then improves the performance of shaking control system. The classical means based on the periodogram is lack of resolution to es timate the power spectrum density (PSD) of the response signal of the shaking table. A modern method for PSD estimation named auto-recursive (AR) model is introduced. The parameters of the AR model are solved by Yule-Walker equation, and the method to determine the orders of the AR model is also formed. The experime nt is carried out using the signal processing platform constructed by DSPs and ARM, and the results show that the algorithm provided in this paper is suitable.