The downlink transmission performance of the massive MIMO Time Division Duplex (TDD) system is bottlenecked by the channel reciprocity errors called antenna reciprocity errors. Antenna reciprocity errors are caused by the mismatch and mutual coupling between antennas. In order to compensate antenna reciprocity errors of the base station, a reciprocity calibration algorithm is proposed in this paper, which can reduce the impact of channel estimate errors by using total least square estimation and increasing channel measurement samples. Rayleigh quotient iteration is also used to reduce the complexity of the reciprocity calibration algorithm in this paper. Simulation results reveal that the algorithm proposed in this paper can achieve 1.8 dB performance gain with respect to the traditional one proposed in references when the antenna reciprocity errors of user equipments are ignored. When the antenna reciprocity errors of user equipments are considered,the performance of the proposed algorithm increases with the decreasing variance of the channel estimation errors.