Maximum likelihood based noise covariance matrix estimation for multi-microphone speech enhancement.
Ulrik KjemsJesper JensenPublished in: EUSIPCO (2012)
Keyphrases
- speech enhancement
- covariance matrix
- maximum likelihood
- covariance matrices
- additive noise
- estimation error
- noisy environments
- noise reduction
- parameter estimation
- signal to noise ratio
- maximum likelihood estimation
- wiener filter
- single channel
- speech signal
- automatic speech recognition
- background noise
- principal component analysis
- sample size
- linear prediction
- gaussian distribution
- em algorithm
- noise level
- smoothing algorithm
- expectation maximization
- speech recognition
- vocal tract
- maximum a posteriori
- sound source
- correlation matrix
- spatial domain
- noisy images
- mixture model
- multi channel
- noise model
- point spread function
- hyperparameters
- computer vision
- gaussian mixture model
- hidden markov models
- evolutionary algorithm