Articulatory feature classifiers trained on 2000 hours of telephone speech.
Joe FrankelMathew Magimai-DossSimon KingKaren LivescuÖzgür ÇetinPublished in: INTERSPEECH (2007)
Keyphrases
- speech recognition
- adaboost classifier
- feature set
- svm classifier
- training set
- vocal tract
- speech signal
- single feature
- multiple features
- speech synthesis
- training data
- multi stream
- trained classifiers
- feature weights
- feature vectors
- feature values
- decision trees
- text to speech
- test set
- machine learning algorithms
- support vector machine
- acoustic features
- automatic speech recognition
- lexical features
- multi layer perceptron
- multilayer perceptron
- linear prediction
- feature ranking
- naive bayes
- machine learning
- support vector
- linear classifiers
- class labels
- classification algorithm
- roc curve
- classification method
- data sets
- call center
- training samples
- language model
- image features
- supervised training
- hidden markov models
- noisy environments
- learning algorithm