Abstract: A subspace-based minimum mean square error(MMSE) group-blind multi-user detection algorithm for CDMA uplink is proposed where the codes of intra-cell users are known but the codes of inter-cell users are unknown.The interferers are suppressed by exploiting the spreading sequences of all known users.An improved projecting approximation subspace tracking with deflation(PASTd) algorithm is proposed for estimating the signal subspace adaptively which enhances the convergence rate of the algorithm.Simulation results show that the proposed algorithm has a faster convergence rate and much less computational complexity,and the output bit error rate(BER) and signal-to-interference-plus-noise-ratio(SINR) performance are superier to the PASTd blind multiuser detection algorithm and close in on the singular value decomposition(SVD) group-blind multiuser detection algorithm.