Parameter Estimation of a State-Space Model of Noise for Robust Speech Recognition.
Stefan WindmannReinhold Haeb-UmbachPublished in: IEEE Trans. Speech Audio Process. (2009)
Keyphrases
- parameter estimation
- speech recognition
- noisy environments
- state space model
- maximum likelihood
- hidden markov models
- additive noise
- model selection
- speech signal
- least squares
- markov random field
- em algorithm
- language model
- random fields
- pattern recognition
- noise reduction
- expectation maximization
- state estimation
- speech recognition systems
- autoregressive
- noise level
- kalman filter
- neural network
- signal to noise ratio
- semi supervised
- optical flow
- information retrieval