Modulation recognition based on spectral correlation function

As a very important inherent characteristic of modulated signals, Cyclostationarity can be exploited for various signal processing tasks such as detection, classification, especially for signal buried in noise or masked interference. In this paper, cyclostationarity of some typical digitally modulated signals are analyzed and Spectral correlation functions (SCF) of these signals are simulated. Based on two features extracted from cycle frequency profile of SCF, three common digitally modulated signals (BPSK, QPSK, MSK) are recognized with a decision tree classifier. The proposed scheme for modulation recognition is easy for implementation and doesn’t require any prior knowledge such as carrier frequency or symbol rate. Simulations prove that the proposed scheme is able to achieve 90% recognition rate when signal to noise ratio (SNR) higher than -3dB.