A robust rank estimation algorithm of group-blind MMSE multiuser detectors for CDMA systems

In this letter, we propose an adaptive rank-estimation method for the additive white or. colored Gaussian noise model. The main contribution of this. letter has three parts. 1) We investigate the. rank mismatch problem in the group-blind multiuser detector of Wang and Host-Madsen, and find that underestimating the rank causes. significant performance degradation, whereas, rank overestimation can achieve performance gain in. the low signal-to-noise-ratio (SNR) region. However, rank overestimation can lead to inaccurate channel estimation, which degrades the detector performance significantly in the high SNR region. 2) We propose a heuristic criterion for initial rank estimation which is robust for nonwhite noise cases. 3) In order to mitigate the rank mismatch problem, we introduce a hypothesis testing criterion for rank and signal subspace, decisions, which selects the most probable, rank for the group-blind detectors. Simulation results show,that the performance of the group-blind detector using this adaptive rank estimation algorithm is comparable to the group-blind with perfect knowledge of the rank, and even better in the low-to-medium SNR region.