Approximate Centroid Estimation with Constellation Grid Segmentation for Blind M-QAM Classification

This paper solves the problem of Automatic Modulation Classification (AMC) without the knowledge of some key signal parameters. The main achievement is the estimation of signal centroids in a non-cooperative environment. The estimation is based on an approximate distribution theory and implemented with automatic constellation grid segmentation. The classification decision is made by finding the modulation candidates which provides the highest density at estimated centroids. The simulation results show that the proposed blind AMC classifier is able to achieve good accuracy in most cases while outperforming stateof-the-art methods under imperfect channel conditions.