Performance optimization for multicarrier code-division multiple-access systems
Abstract
In this thesis, multicarrier code-division multiple-access (MC-CDMA) systems employing minimum mean-squared error (MMSE) receivers are investigated. The effect of the window function on the error performance, a low-complexity parameter acquisition scheme, a carrier-tracking algorithm based on minimum output energy (MOE) criterion, and a density estimation algorithm for cognitive radios with MC-CDMA waveforms are examined. First, the windows for achieving good multiple-access performance of MC-CDMA systems employing MMSE receivers are studied. In a simplified two-user system, the rectangular window is shown to be the optimum. Necessary conditions are given for improving the performances of spectrally efficient windows by exploiting the cyclic prefix and oversampling in the frequency domain. Second, a least mean-squared error based acquisition scheme for the symbol timing, carrier phases, and resolvable path gains for MC-CDMA systems is proposed. A new coefficient called quasi-mean-squared accuracy coefficient is introduced to reduce the computation complexity. Third, a MOE-based carrier-tracking scheme is proposed. The proposed non-data-aided gradient-descent algorithm utilizes an exponentially weighted empirical correlation to promptly reflect the variation of the carrier frequency onto the correlation matrix needed by the algorithm. To speed up the tracking without sacrificing the performance of the steady-state tracking error, a simple finite state machine is used to form an adaptive forgetting factor that dictates the exponential weights for the empirical correlation. Finally, cognitive radio schemes are examined with the goal of using the radio-frequency spectrum efficiently. A collaborative sensing of spectral occupancy is considered based on the collected sensing statistics for the optimum likelihood ratio test. A parametric density estimation scheme using the expectation-maximization (EM) algorithm is proposed to estimate the parameters of densities of the sensing statistics. When the log-likelihood function for the EM algorithm satisfies a certain condition, the maximization procedure is shown to require only a weighted sum of the collected sensing statistics.
Degree
Ph.D.
Advisors
Lehnert, Purdue University.
Subject Area
Electrical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.