Modulation Classification Based on Multifractal Features

A novel method for modulation classification is proposed. The multifractal dimensions of signal, which have preferable distinction capability and robustness to SNR variance, are extracted as the distinctive features. These features are obtained based on the phase space reconstructing theory and the classifier is built by adopting probabilistic neural network. Simulations on CW, BPSK, BFSK, 4ASK and 16QAM signals indicate that the proposed method can classify the modulated signals accurately