Modulation classification for QAM/PSK using a soft clustering algorithm

We propose a novel algorithm for digital modulation classification (MC) by exploiting the constellation shape as a key signature of the modulation type. Viewing the constellation shape as a geometrical pattern, we change the MC problem to a general pattern classification issue. Moreover the algorithm enables us to identify the modulation type under non-coherent conditions. Our approach can be viewed as a method for blind estimation of the reference phase with modulation identification.