A Geometric Approach to Improve Interference Mitigation in Multi-User Detection and Equalization

We apply maximal asymptotic conditional efficiency joint successive interference cancellation (MACE-JSIC) to frequency-selective channel equalization and propose improvements that extend MACE-JSIC detection to the broader paradigm of time-varying and imprecisely known communication systems. The need for expensive redesign to include any change in the interfering signals is a fundamental limitation of MACE-JSIC detection despite its high performance and low detection complexity. In this work, we derive sufficiency criteria that reduce maximum asymptotic efficiency (MAE) detection to low-complexity decorrelation and exploit this relationship to extend the MACE-JSIC approach to time-varying communication systems. Simulations on several multi-user systems and frequency-selective channels demonstrate that the performance of the proposed detectors is consistent with theoretical expectations.