Speech Enhancement Based on Bayesian Low-Rank and Sparse Decomposition of Multichannel Magnitude Spectrograms.
Yoshiaki BandoKatsutoshi ItoyamaMasashi KonyoSatoshi TadokoroKazuhiro NakadaiKazuyoshi YoshiiTatsuya KawaharaHiroshi G. OkunoPublished in: IEEE ACM Trans. Audio Speech Lang. Process. (2018)
Keyphrases
- speech enhancement
- single channel
- linear prediction
- speech signal
- multi channel
- low rank and sparse
- noisy environments
- noise reduction
- signal recovery
- vocal tract
- frequency domain
- signal to noise ratio
- convex optimization
- speech recognition
- prior information
- wiener filter
- bayesian networks
- independent component analysis
- low rank
- bayesian inference
- additive noise
- image coding
- brain activity
- signal processing
- prediction error
- machine learning
- sound source
- maximum likelihood
- multiresolution
- hierarchical bayesian model
- pattern recognition
- information retrieval