Signal classification using statistical moments

An automatic modulation classification algorithm utilizing the statistical moments of the signal phase is developed and used to classify the modulation type of general M-ary PSK signals. It is shown that the nth moment (n even) of the phase of the signal is a monotonic increasing function of M. On the basis of this property, the authors formulate a general hypothesis test, develop a decision rule, and derive an analytic expression for the probability of misclassification. Two examples are given to demonstrate the performance of the algorithm. The algorithm is compared with the quasi-log-likelihood radio (qLLRC), square-law (SLC), and phase-based (PBC) classifiers. The algorithm is outperformed by q LLRC at low CNR but is comparable to SLC and is better than PBC. The qLLRC algorithm is only valid at CNR<0 dB and can be used only to discriminate between BPSK and QPSK signals, whereas the moments algorithm is more general