반응형
[Paper 리뷰] SpecGrad: Diffusion Probabilistic Model based Neural Vocoder with Adaptive Noise Spectral Shaping
SpecGrad: Diffusion Probabilistic Model based Neural Vocoder with Adaptive Noise Spectral ShapingDenoising Diffusion Probabilistic Model을 사용하는 neural vocoder는 주어진 acoustic feature에 대한 diffusion noise 분포를 조절함SpecGradTime-varying spectral envelope가 conditioning log mel-spectrogram에 가까워지도록 diffusion noise를 조절하는 neural vocoderTime-varying filter를 통한 high-frequency band에서의 품질 상승논문 (INTERSPEECH 2022) ..
Paper/Vocoder
2024. 1. 16. 15:38
반응형