| Multi-user detection focuses on the problem of demodulating the information symbols of multiple users sharing a common communication medium, e.g., a wireless communication system. An extensive literature currently exists on the various aspects of multi-user detection, including linear detection, successive and parallel interference cancellation, joint detection, and group detection, among others. The focus of this research is to address the problem of multi-user detection from the perspective of the physical structure of the signal constellation. Many popular methods approach the problem of multi-user detection in terms of algebraic relationships between the user signals. We have instead taken a more geometric approach to the problem of interference suppression and joint detection in the sense that the structural geometry of the signal constellation is used to derive more efficient detection algorithms. We have also considered the case where the combined multi-user signal is passed through a known memoryless non-linearity, e.g., the RF amplifier in a satellite communication system. The non-linearity breaks down any assumptions that currently existing detectors make about the multi-user channel model, e.g., the mutual interference of the user signals is additive, or that the user signals are linearly independent. Our approach has been to generalize the algebraic methods to include the possibility of non-linear and, hence, non-additive signal interference. We have come up with new detection algorithms that work effectively on non-linear as well as linear multi-user systems. We have shown that under sufficient conditions, some of our proposed detectors reduce to currently existing popular multi-user detectors, like the decorrelator. We have also posed the non-linear multi-user system model as a linear system with a higher number of users under certain assumptions on the non-linearity. Finally we have derived a joint detection algorithm that collectively demodulates all the user symbols by using the energy distribution of the signal constellation. The key motivation behind this research is to derive useful information based on the structural geometry of the signal constellation and design detectors that use this information to achieve higher performance gains in terms of the slope of the probability of error in the high SNR regime. |
