Data Augmentation for Environmental Sound Classification Using Diffusion Probabilistic Model with Top-K Selection Discriminator

Despite consistent advancement in powerful deep learning techniques recent years, large amounts of training data are still necessary for the models to avoid overfitting. Synthetic datasets using generative adversarial networks (GAN) have recently been generated overcome this problem. Nevertheless, despite advancements, GAN-based methods usually hard train or fail generate high-quality samples. In paper, we propose an environmental sound classification (ESC) augmentation technique based on diffusion probabilistic model (DPM) with DPM-Solver ++ fast sampling. addition, ensure quality spectrograms, a top-k selection filter out low-quality synthetic According experiment results, samples similar features original dataset and can significantly increase accuracy different state-of-the-art compared traditional techniques. The public code is available https://github.com/JNAIC/DPMs-for-Audio-Data-Augmentation .