Data Augmentation with Adversarial Autoencoders for the Clustering of Electromagnetic Interference Signals

The widespread application of telecommunication technologies urgently calls for an effective method to analyze the interferences in signals. To utilize machine learning algorithms to discern different types of EMI, several models have been proposed in the past, unfortunately almost all of which suffer from the scarcity of training data. In this paper, we propose a method to augment the training data set. With data generated by pre-trained adversarial autoencoders, models are enabled to perform better across multiple metrics.