A fully blind MMSE multiuser detector

Abstract

We propose a blind implementation of the finite-tap linear MMSE detector for asynchronous directsequence CDMA. Unlike partially blind detectors that require knowledge of the signature sequence of the desired user in lieu of a training sequence, the proposed detector requires neither training nor kpowledge of any of the signature sequences. Moreover, the detector need not know the number of interfering users, the size of their QAM alphabets, nor the amount of memory in the channel. The detector first transforms the channel into an equivalent higher-dimensional channel without memory by stacking a sufficient number of receiver observations. The three factors in a singular-value decomposition of the MMSE detector are then implemented one by one, with each factor being adapted blindly and independently. Numerical results demonstrate that, unlike many subspace-based detectors, the proposed detector is robust to inaccuracies in its estimate of the signal subspace dimension.