Recursive speech enhancement using the EM algorithm with initial conditions trained by HMM's.
Ki Yong LeeByung-Gook LeeIickho SongJisang YooPublished in: ICASSP (1996)
Keyphrases
- em algorithm
- initial conditions
- speech enhancement
- speech signal
- expectation maximization
- hidden markov models
- speech recognition
- noisy environments
- maximum likelihood
- noise reduction
- mixture model
- signal to noise ratio
- parameter estimation
- gaussian mixture model
- differential equations
- vocal tract
- linear prediction
- single channel
- fixed point
- automatic speech recognition
- hyperparameters
- generative model
- maximum a posteriori
- wiener filter
- probabilistic model
- conditional random fields
- non stationary
- image coding
- machine learning
- edge detection
- image processing