Vocos: Closing the gap between time-domain and Fourier-based neural vocoders for high-quality audio synthesis.
Hubert SiuzdakPublished in: CoRR (2023)
Keyphrases
- high quality
- frequency domain
- fourier transform
- network architecture
- neural network
- multimedia
- image quality
- signal processing
- audio signals
- neural model
- fourier spectrum
- neural fuzzy
- ground truth
- spatial domain
- audio video
- low quality
- audio visual
- audio signal
- fourier domain
- fourier analysis
- audio stream
- visual information
- radon transform
- digital video
- feature extraction
- depth map
- hebbian learning
- visual data
- higher quality
- bio inspired
- morphological operators
- biologically plausible
- shift invariant
- program synthesis
- multimedia information
- associative memory
- emotion recognition
- learning rules