Statistical approach to enhancing esophageal speech based on Gaussian mixture models.
Hironori DoiKeigo NakamuraTomoki TodaHiroshi SaruwatariKiyohiro ShikanoPublished in: ICASSP (2010)
Keyphrases
- gaussian mixture model
- speaker recognition
- speech signal
- speaker identification
- mixture model
- mel frequency cepstral coefficients
- speech recognition
- feature vectors
- em algorithm
- maximum likelihood criterion
- speech music discrimination
- gaussian mixture
- maximum likelihood
- feature space
- gaussian distribution
- speech quality
- automatic speech recognition
- hidden markov models
- density estimation
- finite mixtures
- multi modal
- vector quantization
- expectation maximization
- noisy environments
- mixture distribution
- probabilistic neural network
- covariance matrices
- variational bayes
- low level
- probability density
- audio visual
- pattern recognition
- statistical methods
- model selection