Near-Far Resistant Linear Multiuser Detection

摘要

This dissertation studies linear detectors for Code-Division Multiple-Access channels perturbed by additive white Gaussian noise, under the assumption that the ensemble of user code waveforms is known to the receiver. In particular the near-far problem of Direct-Sequence Spread-Spectrum Multiple-Access channels, i.e., the problem of demodulating a weak transmitter in the presence of powerful interferers, is solved with a linear receiver. Two multiuser performance measures are used to quantify the performance degradation due to the presence of multiuser interference: The asymptotic efficiency, which is equivalent to bit-error-rate in the low background noise region, and the near-far resistance, which measures a detector's robustness to the near-far problem. Conventional single-user detection in a multiuser channel is not near-far resistant, while the substantially higher performance of the optimum multiuser detector requires exponential complexity in the number of users. We show that the near-far resistance of optimum multiuser detection is achieved by a linear receiver (whose complexity per demodulated bit is only linear in the number of users). The optimum linear detector for worst-case energies - the decorrelating detector - is found, along with existence conditions, which are always satisfied in the models of practical interest. Its implementation does not require knowledge of the received energies, its bit-error-rate is energy-independent, and it achieves optimum near-far resistance. (JHD)