Unsupervised speech enhancement with spectral kurtosis and double deep priors.
Hien OhnakaRyoichi MiyazakiPublished in: CoRR (2024)
Keyphrases
- speech enhancement
- linear prediction
- noise reduction
- noisy environments
- signal to noise ratio
- speech signal
- single channel
- image coding
- vocal tract
- wiener filter
- spectral analysis
- prediction error
- background noise
- subband
- bayesian networks
- smoothing algorithm
- spectral features
- pattern recognition
- non stationary
- prior information
- lossless compression
- additive noise
- image segmentation
- autoregressive
- noisy images
- independent component analysis