Semi-tied covariance matrices for acoustic models based on random forests of phonetic decision trees.
Jian XueL. CheYunxin ZhaoPublished in: ICASSP (2009)
Keyphrases
- random forests
- covariance matrices
- acoustic models
- decision trees
- speech recognition
- maximum likelihood
- discriminative training
- random forest
- covariance matrix
- hidden markov models
- speech recognizer
- automatic speech recognition
- training data
- broadcast news
- speaker independent
- machine learning algorithms
- ensemble methods
- logistic regression
- gaussian mixture model
- vector space
- distance measure
- decision tree ensembles
- gaussian distribution
- naive bayes
- training set
- spoken language
- speech signal
- feature vectors
- machine learning
- base classifiers
- sample size
- language model
- principal component analysis
- pattern recognition
- em algorithm
- training process
- dimensionality reduction
- classification accuracy