Speaker adaptive joint training of Gaussian mixture models and bottleneck features.
Zoltán TüskePavel GolikRalf SchlüterHermann NeyPublished in: ASRU (2015)
Keyphrases
- gaussian mixture model
- speaker recognition
- mel frequency cepstral coefficients
- feature vectors
- speaker identification
- mixture model
- feature extraction
- feature space
- training set
- neural network
- decision trees
- extracted features
- gaussian mixture
- density estimation
- training examples
- image features
- low level
- em algorithm
- training data
- covariance matrix
- image classification
- feature set
- probability density
- audio features
- extracting features
- classification accuracy
- image segmentation
- spectral features
- acoustic features