Abstract: The minimum mean square error(MMSE) multiuser detectors(MUDs) based on subspace tracking are deeply investigated,and an improved algorithm,i.e.,a modified MMSE multiuser detector is designed to solve the problems of detecting performance degradation on traditional algorithm caused by eigenvalue estimation errors.The orthogonal projection approximation subspace tracking(OPAST) algorithm with good orthogonality is used to track the subspace,and a modified MMSE multiuser detection algorithm based on OPAST is proposed.The simulation results show that,compared with the OPAST-based MMSE MUDs,the proposed modified MMSE algorithm based on OPAST offers faster convergence rate,its output SIR(signal to interference ratio) and bit error rate(BER) are close to those of the SVD(singular value decomposition) MUD method,and it reduces the computational complexity.