Robust QAM modulation classification via moment matrices

We discuss a method for classification of digitally modulated signals based on performing subspace decomposition on a positive definite matrix of higher order moments of the received signals. Specifically, we specialize a general approach originally introduced for detection and classification of noise contaminated patterns to the case of digitally modulated signals such as M-ary PSK and QAM. We consider two different classifiers: one that provides only satisfactory performance for high signal-to-noise ratio, and one that performs also well in the low SNR regime. The former has the additional advantage of being invariant to both unknown phase angle (rotation) and signal amplitude, and can be used for all QAM signal constellations (including M-ary PSK), whereas the latter is only used for discrimination of M-ary PSK signals. Using simulation, we analyze the performance of the proposed classifier for transmission over the additive white Gaussian noise channel and both coherent and non-coherent reception. Moreover, the robustness of the classifier against mismatched noise modeling is discussed.