Automatic Modulation Classification under IQ Imbalance Using Supervised Learning

The process of classifying digital modulation schemes given IQ imbalance at the transmitter or receiver is studied using fourth and sixth order cumulants as features. Various methods of supervised learning are proposed in order to mitigate the effect of IQ imbalance at the receiver, including K-Nearest Neighbors (k-NN), Support Vector Machine (SVM), and decision tree learning. The impact of IQ imbalance at the transmitter is also observed, as well as the effect of IQ imbalance on the theoretical cumulant values for each modulation scheme. Through simulation, it is shown that supervised learning approaches are effective at compensating for the IQ imbalances that can occur at the receiver.