A new approach for speech enhancement based on a constrained Nonnegative Matrix Factorization.
Nasser MohammadihaTimo GerkmannArne LeijonPublished in: ISPACS (2011)
Keyphrases
- nonnegative matrix factorization
- speech enhancement
- noisy environments
- noise reduction
- single channel
- signal to noise ratio
- matrix factorization
- data representation
- negative matrix factorization
- speech signal
- least squares
- wiener filter
- vocal tract
- spectral clustering
- background noise
- multi channel
- principal component analysis
- linear prediction
- document clustering
- original data
- objective function
- additive noise
- image data
- speech recognition
- edge detection
- computer vision