One of the key technologies of massive Multiple Input Multiple Output(MIMO) systems is channel estimation. The problem of downlink channel estimation in the Frequency Division Duplexing (FDD) Massive MIMO system is studied. By using the block sparsity of the Massive MIMO-OFDM channel in the angular-frequency domain, a low complexity estimation algorithm based on block matching pursuit is proposed. In addition, the system coherence time may be exceeded due to too much time cost when adopting orthogonal pilot symbol in the time domain. To solve this problem, an antenna grouping transmission scheme is proposed to reduce the total channel estimation time by sacrificing the observed data length. The simulation results show that the proposed algorithm has good anti-noise performance, and it can accurately determine the non-zero position of the sparse vector and can adaptively determine the sparsity.