Maximum likelihood approach to speech enhancement for noisy reverberant signals.
Takuya YoshiokaTomohiro NakataniTakafumi HikichiMasato MiyoshiPublished in: ICASSP (2008)
Keyphrases
- noisy environments
- speech enhancement
- maximum likelihood
- speech signal
- speech recognition
- additive noise
- single channel
- spectral analysis
- background noise
- noise reduction
- automatic speech recognition
- vocal tract
- em algorithm
- non stationary
- multi channel
- maximum a posteriori
- signal processing
- hidden markov models
- sound source
- expectation maximization
- multiscale
- image processing
- image sequences