On Loss Functions for Supervised Monaural Time-Domain Speech Enhancement.
Morten KolbækZheng-Hua TanSøren Holdt JensenJesper JensenPublished in: IEEE ACM Trans. Audio Speech Lang. Process. (2020)
Keyphrases
- loss function
- speech enhancement
- noisy environments
- noise reduction
- single channel
- signal to noise ratio
- pairwise
- frequency domain
- speech signal
- support vector
- linear prediction
- semi supervised
- learning to rank
- vocal tract
- smoothing algorithm
- loss minimization
- wiener filter
- background noise
- multi channel
- risk minimization
- supervised learning
- convex loss functions
- feature selection
- reproducing kernel hilbert space
- additive noise
- pattern recognition
- learning algorithm
- speech recognition
- markov random field
- machine learning