A blind MMSE multi-user detection based on NOOja algorithm

Abstract

A new blind adaptive MMSE multi-user detection (MUD) based on subspace tracking is presented. The new detector doesn't employ signal eigenvalue estimation but the signal subspace estimation, and it avoids performance deterioration induced by eigenvalue estimation error. The proposed MUD exploits the normalized orthogonal Oja (NOOja) subspace tracking algorithm for subspace estimation, since it guarantees the orthogonality of the weight matrix spanned by the singnal subspace in every iteration, which must be meet in the new detector. The simulation results the proposed MMSE detector has faster convergence rate, better output SINR (signal-to-interference-and-noise ratio) and bit error rate (BER) and lower the computational complexity.