Signal Data Augmentation Algorithm Research Based on Fractal Theory

Machine learning algorithms alleviate overfitting through data augmentation. However, current methods such as Generative Adversarial Networks for data augmentation suffer from issues such as prolonged training time and slow convergence. Therefore, this paper proposes a data augmentation method based on fractal theory to address these issues. This method is based on relevant fractal theory and proceeds in the order of initialization, data transformation, construction of fractal codes, Iterated Function System (IFS) iteration, and output of augmented results. To verify the effectiveness of this method, it was compared with DCGAN and WGAN-GP by augmenting a segment of audio signal data. The results show that, under similar MSE conditions, the fractal augmentation method takes ten times less time than the other two methods. This indicates that the proposed fractal augmentation method is effective and its performance is significantly better than the other two methods. That is to say, the method proposed in this paper provides an effective approach to address the issue of time consumption in the data augmentation process.