Modulation classification of MFSK signals using the higher-order correlation domain

The problem of classification of M-ary frequency shift keying (MFSK) signals when contaminated by additive white Gaussian noise (AWGN) is addressed. In particular, we propose a novel class of classifiers that utilizes time-domain higher-order correlation (HOC) operations while completely avoiding the explicit determination of the spectrum of the observed signal. Families of classification rules are identified for three different cases that exhaust all possible relations between symbol duration and bandwidth. The performances are numerically established in terms of the probability of correct classification. It is shown that the HOC-based method delivers a performance which tightly lower-bounds that of the optimal likelihood-ratio test. In addition, this intrinsically wideband method possesses an immunity to imperfect knowledge of exact frequency locations. Substantial performance improvement is also reported over the energy-based rule whenever it is applicable.