Training of reduced-rank linear transformations for multi-layer polynomial acoustic features for speech recognition.
Muhammad Ali TahirHeyun HuangAlbert ZeyerRalf SchlüterHermann NeyPublished in: Speech Commun. (2019)
Keyphrases
- speech recognition
- multi layer
- acoustic features
- automatic speech recognition
- speech signal
- linear transformations
- hidden markov models
- mel frequency cepstral coefficients
- language model
- noisy environments
- speaker identification
- pattern recognition
- neural nets
- neural network
- broadcast news
- training set
- visual information
- parameter space
- music information retrieval
- image representation
- supervised learning
- active learning
- audio features
- speaker verification