Classification of digital modulations using the LPC

Abstract

We propose a new digital modulation classification method based on the continuous-time wavelet transformation (CWT) and the linear predictive coding (LPC) method. The LPC coefficients extracted from the LPC model of the CWT for a modulated signal is chosen as the feature used to classify the modulation types of BPSK, QPSK, FSK and jammer. By using several reference features per modulation type we can make our algorithm robust to the influence of noise. To verify the proposed modulation classification algorithm, simulations are performed, which demonstrate excellent classification rates.