Semi-supervised non-negative tensor factorisation of modulation spectrograms for monaural speech separation.
Tom BarkerTuomas VirtanenPublished in: IJCNN (2014)
Keyphrases
- semi supervised
- speech signal
- tensor factorization
- speech recognition
- semi supervised learning
- labeled data
- high order
- positive and negative
- unlabeled data
- audio visual
- pairwise
- unsupervised learning
- sound source
- automatic speech recognition
- spoken language
- active learning
- diffusion tensor
- recognition engine
- multi view
- manifold learning
- pairwise constraints
- text to speech
- speech synthesis
- semi supervised classification
- metric learning
- co training
- supervised learning
- high dimensional
- tensor decomposition
- broadcast news
- dimensionality reduction
- hearing impaired
- ofdm system
- higher order
- single channel
- non stationary
- signal to noise ratio
- matrix factorization
- low rank
- noisy environments
- feature selection
- subspace learning