Using AR HMM state-dependent filtering for speech enhancement.
Driss MatroufJean-Luc GauvainPublished in: ICASSP (1999)
Keyphrases
- conditional random fields
- state dependent
- speech enhancement
- hidden markov models
- speech signal
- noisy speech
- noisy environments
- steady state
- adaptive filtering
- noise reduction
- speech recognition
- queueing networks
- single channel
- signal to noise ratio
- optimal policy
- probabilistic model
- augmented reality
- higher order
- single server
- stationary distribution
- markov chain
- automatic speech recognition
- markov models
- arrival rate
- linear prediction
- queue length
- vocal tract
- asymptotically optimal
- background noise
- multi channel
- smoothing algorithm
- prior information
- additive noise
- wiener filter
- neural network