Data Augmentation for Blind Signal Classification

Abstract

The Army Rapid Capabilities Office (RCO) sponsored a Blind Signal Classification Competition seeking Artificial Intelligence/Machine Learning (AI/ML) algorithms to automatically identify the modulation schemes of RF signal from complex-valued IQ (in-phase quadrature) symbols. Our team, dubbed “Deep Dreamers”, participated in the competition and placed 3rd out of 42 active teams across industry, academia, and government. Deep learning methods such as CNN, Residual Neural Network (ResNet), and Long Short-Term Memory (LSTM) are the fundamental neural network models we used to develop a multi-class classifier. The key to our success was to use ensemble learning to average the outputs of multiple diverse classifiers. In this following study, we apply Data Augmentation (DA) to the data set in order to further increase the performance of our models. The goal of data augmentation is to push the decision boundary learned from the data set toward a better decision boundary by adding more meaningful data points. An effective data augmentation method for RF signals is to add white Gaussian noise to the existing RF signals. Individual DL models and ensemble learning methods such as blending trained over the augmented data set significantly improve the prediction accuracies for weak RF signals and achieve comparable results to the two leading teams in the competition.