BigVSAN: Enhancing GAN-based Neural Vocoders with Slicing Adversarial Network

arXiv:2309.02836v2 Announce Type: replace-cross
Abstract: Generative adversarial network (GAN)-based vocoders have been intensively studied because they can synthesize high-fidelity audio waveforms faster than real-time. However, it has been reported that most GANs fail to obtain the optimal projection for discriminating between real and fake data in the feature space. In the literature, it has been demonstrated that slicing adversarial network (SAN), an improved GAN training framework that can find the optimal projection, is effective in the image generation task. In …

adversarial arxiv cs.lg cs.sd eess.as gan network slicing type