Hierarchical digital modulation classification using cumulants

TLDR

These statistics naturally characterize the shape of the noisy baseband I and Q sample distribution in digital modulation classification. The authors propose a simple method based on elementary fourth‑order cumulants for classifying digital modulation schemes. The method employs elementary fourth‑order cumulants, operates with O(N) complexity in the number of baseband samples, and its theoretical validity is confirmed by extensive simulations and comparisons with existing approaches. Cumulant‑based classification proves effective in a hierarchical scheme, enabling subclass separation at low SNR with few samples, serving as a preliminary classifier, remaining robust to carrier phase and frequency offsets, and outperforming existing methods in simulations.

Abstract

A simple method, based on elementary fourth-order cumulants, is proposed for the classification of digital modulation schemes. These statistics are natural in this setting as they characterize the shape of the distribution of the noisy baseband I and Q samples. It is shown that cumulant-based classification is particularly effective when used in a hierarchical scheme, enabling separation into subclasses at low signal-to-noise ratio with small sample size. Thus, the method can be used as a preliminary classifier if desired. Computational complexity is order N, where N is the number of complex baseband data samples. This method is robust in the presence of carrier phase and frequency offsets and can be implemented recursively. Theoretical arguments are verified via extensive simulations and comparisons with existing approaches.