Further results in likelihood classification of QAM signals

Maximum likelihood decision theory is applied to the problem of classification of quadrature-modulated digital communication signals. Several aspects of the existing low signal-to-noise ratio (SNR) results are extended to the moderate and high SNR environments. An approximate probability density function (PDF) for the single-term approximation to the average log-likelihood-ratio (ALLR) which is valid at all SNR values is presented and its superior accuracy, compared to the low SNR pdf, is verified via computer simulation. Computer simulation is also used to show that multiple-term approximations to the ALLR may provide significant performance gains relative to their individual terms. A simple, practical method for setting the threshold of the ALLR test is presented and it is shown, through simulation, that little performance degradation is suffered relative to the optimal setting, which is difficult to determine analytically in most cases. True signal pre-processing techniques are also presented, and it is demonstrated that their use significantly improves the robustness of the classification algorithms for phase-shift-keying signals in frequency-uncertain environments