Stochastic-deterministic signal modelling for the tracking of pitch in noise and speech mixtures using factorial HMMs.
Matthew McCallumBernard J. GuilleminPublished in: INTERSPEECH (2013)
Keyphrases
- hidden markov models
- noisy speech
- speech recognition
- noisy environments
- speech signal
- fundamental frequency
- additive noise
- speech quality
- automatic speech recognition
- random noise
- signal detection
- source signals
- stochastic optimization problems
- low snr
- low signal to noise ratio
- independent component analysis
- stochastic methods
- background noise
- particle filter
- signal processing
- phoneme recognition
- acoustic features
- noise level
- finite state transducers
- formant frequencies
- received signal
- speech enhancement
- multi stream
- blind source separation
- signal to noise ratio
- high frequency
- source separation
- stochastic resonance
- additive gaussian noise
- wide band
- white noise
- speech recognizer
- low frequency
- noisy data
- impulse response
- speaker verification
- image sequences
- independent components
- power spectrum
- noise reduction
- noisy images
- non stationary
- mixture model
- expectation maximization
- hidden semi markov models
- markov chain
- direct sequence spread spectrum