GMM and HMM training by aggregated EM algorithm with increased ensemble sizes for robust parameter estimation.
Takahiro ShinozakiTatsuya KawaharaPublished in: ICASSP (2008)
Keyphrases
- em algorithm
- baum welch
- gaussian mixture model
- expectation maximization
- robust parameter estimation
- mixture model
- gaussian mixture
- hidden markov models
- maximum likelihood
- training set
- maximum likelihood estimation
- parameter estimation
- likelihood function
- hyperparameters
- estimate the model parameters
- generative model
- expectation maximisation
- log likelihood
- probability density function
- learning algorithm
- mixture modeling
- supervised learning
- penalized likelihood
- incomplete data
- background subtraction
- feature selection
- density estimation
- ensemble methods
- maximum a posteriori
- missing data
- model based clustering
- model selection
- log likelihood function
- probabilistic model