A Method for Modulation Recognition of PSK/APSK Signals Based on Optimal Clustering

In order to mitigate the sensitivity of existing PSK/APSK signal modulation recognition methods to prior parameters, this paper presents a blind recognition approach based on optimal clustering. Firstly, the sampled signal is preprocessed based on estimated carrier frequency, bandwidth, and code rate; subsequently, the feature of constellation clustering entropy is defined for optimal clustering of the constellation diagram in order to recover its distribution. Finally, the matching degree between the constellation diagram distribution and pre-set templates of different modulation types is calculated for pattern recognition, encompassing BPSK, QPSK, 8PSK, 16PSK, and 16APSK signals. The method leverages inherent differences in overall constellation diagram distributions for stable and practical engineering implementation. Simulation results demonstrate that at a signal-to-noise ratio of 10 dB or higher, the recognition rate for all signal types exceeds $\mathbf{9 7 \%}$, validating the effectiveness of our proposed method.