Acoustic modeling for under-resourced languages based on vectorial HMM-states representation using Subspace Gaussian Mixture Models.
Mohamed BouallegueEmmanuel FerreiraDriss MatroufGeorges LinarèsMaria GoudiPascal NoceraPublished in: SLT (2012)
Keyphrases
- gaussian mixture model
- covariance matrices
- mixture model
- hidden markov models
- feature space
- maximum likelihood
- speaker recognition
- em algorithm
- gaussian mixture
- probability density
- expectation maximization
- feature vectors
- low dimensional
- semi supervised
- variational bayes
- speaker identification
- speech recognition
- information retrieval
- mel frequency cepstral coefficients
- gaussian distribution
- vector space
- high dimensional data
- image representation
- language model
- prior knowledge