With the development of high-speed railway, the future wireless communication system is facing the high-speed mobile environment, where rapid and significant variations will emerge in the instantaneous wireless channel conditions. The main challenge on the channel estimation is that plenty of pilots should be inserted to track the fast- varying channel. In this paper, we propose a method to estimate the fast-varying channel with the aid of Basis Expansion Model (BEM) and uplink Sounding Reference Signal (SRS). BEM is used to model the fast-varying channel with the combination of a finite number of basis functions, and the combination coefficients are slow-varying. Therefore, the sparse SRS can be used to estimate the channel, which will reduce the pilot number compared to the estimation with demodulation reference signals. Simulation results show that our proposed method provides better performance than the traditional one in the fast- varying channel, and can accommodate to different Doppler frequency shift with changing the BEM order.