Blind modulation classification of linearly digitally modulated signals

Hameed, Fahed (2006) Blind modulation classification of linearly digitally modulated signals. Masters thesis, Memorial University of Newfoundland.

[English] PDF - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. Download (6MB)

Abstract

Blind modulation classification (MC) is an intermediate step between signal detection and demodulation, with both military and civilian applications. MC is a challenging task, especially in a non-cooperative environment, as no prior information on the incoming signal is available at the receiver. -- In this thesis, we investigate classification of linear digital modulations in slowly varying flat fading channels. With unknown channel amplitude, phase and noise power at the receive-side, we investigate hybrid likelihood ratio test (HLRT) and quasi-HLRT (QHLRT) -based classifiers, and discuss their performance versus computational complexity. Both classifiers rely on the likelihood function (LF) of the received signal, and the decision is made based on the likelihood ratio test. To compute the LF, the former employs maximum likelihood (ML) estimates of the unknown parameters, whereas the latter uses method-of-moment (MoM) estimates of these parameters. It is shown that the QHLRT algorithm provides a low computational complexity solution, yet yielding performance close to the HLRT algorithm. -- We further study the performance of MoM estimators employed in the QHLRT algorithm, in terms of their variance. We derive Cramer-Rao Lower Bounds (CRLBs) of estimators of the channel amplitude, phase and noise power, for Binary Phase Shift Keying (BPSK) and Quadrature Phase Shift Keying (QPSK) modulated signals. CRLB provides a lower bound on the variance of an unbiased estimator. The CRLB expressions are evaluated for different signal-to-noise ratios (SNRs) and number of processed symbols. Variance of MoM estimators is compared with corresponding CRLB and numerical results reveal reasonable accuracy of MoM method of parameter estimation, for the SNR range in which these estimators remain unbiased. -- An application of the CRLB to MC is presented, by developing an idealized reference for QHLRT-based classifier. This is referred to as the QHLRT-IR (QHLRT-IR) classifier. The QHLRT-IR classifier employs unbiased and normally distributed estimators of the unknown parameters, with mean as the true value of the parameter and variance given by the CRLB of the parameter estimator. Performance of the QHLRT and QHLRT-IR classifiers are compared. QHLRT-IR provides an upper bound on the classification performance in the SNR range where the MoM estimators of the unknown parameters remain unbiased and normally-distributed. In this SNR range, it is observed that the performance of the two classifiers is close to each other, which indicates the reasonable accuracy of MoM method for such SNRs. It is shown that the classification performance improves with an increase in the number of processed symbols, due to the increase in estimation accuracy.

Actions (login required)

View Item