A novel method for automatic modulation classification under non-Gaussian noise based on variational mode decomposition

Automatic modulation classification (AMC) plays an important role in identifying the modulation format of a signal. Most of the existing modulation classifiers assume the signal to be contaminated only by additive white Gaussian noise (AWGN). However, the performances of these traditional classifiers degrade in the presence of non-Gaussian impulsive noise. In this paper, we present a novel automatic modulation classification algorithm based on variational mode decomposition (VMD) in presence of non-Gaussian impulsive noise and additive white Gaussian noise. Our proposed method is a three step process. In the first step, received signal is decomposed into modes. It is clear from simulation results that impulse noise is effectively captured in the first mode. In the second step, all other modes are added except first mode to reduce the impact of impulse noise in the received signal. Then in the third step, cyclostationary feature based classifiers are employed to identify the modulation type. The proposed algorithm is evaluated using well-known existing classifiers for different digital modulation techniques. Comparative results depict the superior performance of our proposed method over other traditional methods under different noise conditions.