A Variable Initialization Approach to the EM Algorithm for Better Estimation of the Parameters of Hidden Markov Model Based Acoustic Modeling of Speech Signals.
Md. Shamsul HudaRanadhir GhoshJohn YearwoodPublished in: ICDM (2006)
Keyphrases
- em algorithm
- maximum likelihood estimation
- expectation maximization
- parameter estimation
- maximum likelihood
- speech signal
- log likelihood function
- update equations
- expectation maximisation
- mixture model
- likelihood function
- hyperparameters
- maximum likelihood estimates
- gaussian mixture
- speech sounds
- gaussian mixture model
- automatic speech recognition
- density estimation
- generative model
- sound source
- parameter learning
- k means
- speech recognition
- probabilistic model
- probability density function
- mixture distribution
- hidden variables
- non stationary
- unsupervised learning
- maximum a posteriori
- speech enhancement
- mixture modeling
- hidden markov models
- automatic speech recognition systems
- formant frequencies
- closed form
- model selection
- model based clustering
- multi modal
- image segmentation