A New Turbo Multi-User Receiver With Recursive SIC-MMSE Detection

In joint iterative (turbo) multi-user detection and decoding, or turbo multi-user receiver, suboptimal inner detection schemes are often used due to the prohibitively high complexity of the optimal soft multi-user detection (MUD). A popular suboptimal scheme is the so-called soft interference cancellation plus minimum mean-square error (SIC-MMSE) detection that uses the a priori information from the outer decoder(s) to perform soft interference cancellation before the linear MMSE detection. In addition to the outer decoder outputs, channel observation also provides useful information for the interferences. Motivated by this observation, a new soft-output multi-user detection scheme was proposed that uses both the decoder and the detector outputs of the previous iteration as the a priori information. A nonlinear MMSE estimator is developed that takes these two "a priori" information sources to produce soft interference estimates before interference cancellation and linear MMSE detection. In one iteration, the proposed scheme has comparable complexity as that of the SIC-MMSE scheme. Extrinsic information transfer (EXIT) chart analysis and simulation results show that the proposed scheme leads to a convergence point with higher output mutual information but with significantly less iterations. As a result, considerable performance improvement is achieved with lower complexity when compared to conventional SIC-MMSE-based turbo multi-user receiver.