Computationally efficient blind multiuser detection for multipath CDMA channels

Abstract

In this paper a blind, adaptive minimum mean squared error (MMSE) receiver, which jointly performs data detection and channel estimation for multipath CDMA channels, is considered. The receiver is implemented using the constrained minimum output energy (MOE) criterion and the least mean square (LMS) algorithm. The computational complexity of the proposed receiver is only of order N (O(N)) per symbol, where N is the processing gain. The proposed receiver requires only the signature code of the user of interest (UOI) and does not require any training sequences, the channel parameters of any users and signature codes of the interfering users (Timing information is assumed to be known by the receiver). Though the channel parameters of the UOI are not explicitly calculated, the energies of multiple paths are advantageously combined as done in a conventional RAKE receiver. Here multiple access interference (MAI) is first removed and SNR is subsequently maximized using the min-max principle. The bit-error performance of the proposed receiver closely approaches that of the non-blind, non-adaptive MMSE detector. We are also introducing a blind, adaptive, decorrelating receiver, which can be obtained even when noise is zero.