Abstract: In order to improve the precision of sparse channel estimation and decrease the pilot numbers in OFDM systems, the paper transforms the estimation of sparse frequency selective fading channel in time-domain for OFDM systems into the reconstruction of complex sparse signal existing the noise interference and without the prior information of the sparsity in compressed sensing. We proposed two methods of sparse channel estimation based on basis pursuit denoising (BPDN) and sparsity adaptive matching pursuit (SAMP) for OFDM systems. Under the same conditions, we compared the two methods with the other channel estimation methods including conventional least square (LS), MP-LS which firstly estimate the positions of the most significant taps through match pursuit (MP), then to estimate the numerical value of the positions by LS method. Simulation results show that the proposed methods do not require the prior information of sparsity, and have the merits of smaller normalized mean square error, lower bit error ratio. Between the two proposed methods, the one based on SAMP has the merits of running faster, nearer to the Cramer-Rao bound than the another, and the BER of the developed method based on SAMP can satisfy the practice applications when the pilot subcarriers is 12.5% of the all carriers and SNR is higher than 10 dB.