A blind reduced-rank multi-user detector(MUD)based on a new variable threshold reduced-rank subspace selection and an improved dimension estimation is proposed.By using the new reduced-rank subspace selection

a proper reduced-rank subspace can be obtained quickly and the subspace can be reused.The improved Akaike Information Criterion(AIC)is adopted to estimate the dimension of the signal subspace in the subspace tracking.The computation of dimension estimation can be reduced by using the criterion in the same error probability.Because AIC has a risk of overestimating the dimension of the signal subspace

the performance of the blind reduced-rank multi-user detector is specially analyzed when the signal subspace dimension is overestimated.The simulation results show that the proposed reduced-rank subspace selection algorithm can achieve the desired system performance with lower computational complexity.