Blocked Gibbs sampling based multi-scale mixture model for speaker clustering on noisy data.
Naohiro TawaraTetsuji OgawaShinji WatanabeAtsushi NakamuraTetsunori KobayashiPublished in: MLSP (2013)
Keyphrases
- noisy data
- mixture model
- multiscale
- markov chain monte carlo
- generative model
- gaussian mixture model
- em algorithm
- probabilistic model
- model selection
- expectation maximization
- maximum likelihood
- missing data
- unsupervised learning
- markov random field
- edge detection
- scale space
- language model
- natural images
- image processing
- high dimensional
- training data
- bayesian inference
- object recognition
- input data
- semi supervised
- machine learning
- missing values
- high dimensionality