Coherent detection and demodulation need channel state information at receivers for orthogonal frequency division multiplexing systems. Conventional channel estimations, such as least squares (LS), are widely adopted under the assumption of rich multipath. Recently, numerous physical measurements have verified that many wireless multipath channels encountered in practice tend to exhibit sparse structures. Exploiting the inherent sparsity, a novel channel estimation method based on compressive sensing is proposed, which reduces the number of pilots so as to improve the bandwidth efficiency. Simulation results confirm that the proposed method has a 5-dB lower mean square error in channel estimation when compared to the conventional approach. In addition, the complexity of the proposed method is acceptable.