Constructing ensembles of dissimilar acoustic models using hidden attributes of training data.
Takashi FukudaRyuki TachibanaUpendra V. ChaudhariBhuvana RamabhadranPuming ZhanPublished in: ICASSP (2012)
Keyphrases
- acoustic models
- speech recognition
- training data
- hidden markov models
- automatic speech recognition
- speech recognizer
- broadcast news
- decision trees
- discriminative training
- data sets
- speaker independent
- training process
- speech signal
- training examples
- training set
- learning algorithm
- class labels
- language model
- classification accuracy
- prior knowledge
- information retrieval