Generating identities with mixture models for speaker anonymization.
Henry TurnerGiulio LovisottoIvan MartinovicPublished in: Comput. Speech Lang. (2022)
Keyphrases
- mixture model
- gaussian mixture model
- em algorithm
- density estimation
- probabilistic model
- generative model
- expectation maximization
- model selection
- maximum likelihood
- mixture modeling
- language model
- unsupervised learning
- finite mixture models
- finite mixtures
- probability density function
- information loss
- automatic model selection
- prior knowledge
- image processing
- dirichlet prior
- generalized em algorithm
- mixture components
- log linear models
- model based clustering
- image segmentation