Variational-distance-based modulation classifier

A variational-distance-based scheme is proposed for the modulation classification problem. It decides on the modulation that minimizes the variational distance between the theoretical and empirical probability density of the received signal. Simulation suggests that it outperforms some existing featured-based classifiers, namely the cumulant classifier, K-S classifier and Kuiper classifier. Its computational complexity is comparable to those classifiers but it is more robust to the error in estimating the noise power.