Time-domain approach using multiple Kalman filters and EM algorithm to speech enhancement with nonstationary noise.
Ki Yong LeeSouhwan JungPublished in: IEEE Trans. Speech Audio Process. (2000)
Keyphrases
- feature extraction
- em algorithm
- non stationary
- speech enhancement
- kalman filter
- expectation maximization
- speech signal
- frequency domain
- mixture model
- noisy environments
- noise reduction
- maximum likelihood
- single channel
- gaussian mixture model
- parameter estimation
- signal to noise ratio
- generative model
- random fields
- background noise
- likelihood function
- feature selection
- maximum a posteriori
- probabilistic model
- additive noise
- mean shift
- signal processing
- higher order
- data analysis
- particle filter
- wiener filter
- edge detection