Maximum-likelihood modulation classification with incomplete channel information

This paper presents a discussion of the classification of digital communication signals given incomplete knowledge of the channel. Through a maximum-likelihood framework, modulation classifiers are presented which assume no or limited a priori knowledge of the fading experienced by the signal (including time offset, phase shift, and amplitude) and/or the distribution of the noise added in the channel. A recently published asynchronous classifier for digitally modulated signals, which uses a new channel estimator that is blind to the modulation scheme of the received signal, is introduced and analyzed. In addition, results are presented of our recent work on the classification of digitally modulated signals in flat fading non-Gaussian channels.