Generalized single cycle classifier with applications to SQPSK vs. 2kPSK

A binary signal classification problem, where the received signal has different spectral correlations for different hypotheses, is investigated. A simple classifier, designated the generalized single-cycle classifier, is developed under additive white Gaussian noise for (1) ideal, (2) unknown phase, and (3) unknown phase and epoch environments. As a specific example, the developed classifier is applied to the SQPSK/2^kPSK (k⩾1) modulation classification problem. The optimum choice of cycle and the relation between the single-cycle classifier and the likelihood ratio for SQPSK/2 ^kPSK classification are discussed. Emphasis is placed on low signal-to-noise ratio. The performance of the classifier is predicted by the classification probability for a fixed rejection probability