An Efficient Data Augmentation Method for Automatic Modulation Recognition from Low-Data Imbalanced-Class Regime

The application of deep neural networks to address automatic modulation recognition (AMR) challenges has gained increasing popularity. Despite the outstanding capability learning in feature extraction, predictions based on low-data regimes with imbalanced classes signals generally result low accuracy due an insufficient number training examples, which hinders wide adoption methods practical applications AMR. identification minority class samples can be crucial, as they tend higher value. However, AMR tasks, there is a lack attention and effective solutions problem Imbalanced-class regime. In this work, we present data augmentation method for radio signals, called SigAugment, incorporates eight individual transformations effectively improves performance tasks without additional searches. It surpasses existing mainstream solving imbalanced-class problems multiple datasets. By simply embedding SigAugment into pipeline model, it achieve state-of-the-art benchmark datasets dramatically improve classification trained uniform use different types models works right out box.