Ultrasound-coupled semi-supervised nonnegative matrix factorisation for speech enhancement.
Tom BarkerTuomas VirtanenOlivier DelhommePublished in: ICASSP (2014)
Keyphrases
- matrix factorisation
- speech enhancement
- semi supervised
- pairwise
- noise reduction
- noisy environments
- signal to noise ratio
- single channel
- em algorithm
- speech signal
- multi channel
- linear prediction
- nonnegative matrix factorization
- vocal tract
- wiener filter
- smoothing algorithm
- background noise
- pattern recognition
- face recognition