Beamforming has been considered a crucial technique in 5G era. To perform received beamforming, one must know the information of angle-of-arrival (AoA). This paper considers AoA estimation in single-input-multiple-output (SIMO) pilot-assisted OFDM systems. Exploiting the sparsity of the wireless channel, we can first efficiently estimate channel impulse response with a compressive-sensing based technique. However, existing works often do not take the pulse-shaping effect into account, resulting in poor performance in real-world systems. We propose a basisadaptive block sparse Bayesian learning framework to solve the problem. Once the CIRs corresponding to received antennas are obtained, AoAs can then be estimated accordingly.