Training log-linear acoustic models in higher-order polynomial feature space for speech recognition.
Muhammad Ali TahirHeyun HuangRalf SchlüterHermann NeyLouis ten BoschBert CranenLou BovesPublished in: INTERSPEECH (2013)
Keyphrases
- acoustic models
- speech recognition
- higher order
- discriminative training
- feature space
- speech recognizer
- hidden markov models
- automatic speech recognition
- speaker independent
- speech signal
- language model
- maximum likelihood
- training set
- pattern recognition
- noisy environments
- feature vectors
- support vector machine
- probabilistic interpretation
- image retrieval
- high dimensional
- pairwise
- feature set
- speech recognition systems
- feature selection
- posterior probability
- feature extraction
- speaker identification
- information retrieval
- neural network
- data points
- spoken language
- training data
- similarity measure