ACL RD-TEC 1.0 Summarization of H91-1049
Paper Title:
A DYNAMICAL SYSTEM APPROACH TO CONTINUOUS SPEECH RECOGNITION
A DYNAMICAL SYSTEM APPROACH TO CONTINUOUS SPEECH RECOGNITION
Authors: V.Digalakis and J.R. Rohlicek and M.Ostendorf
Primarily assigned technology terms:
- algorithm
- classification
- computing
- continuous speech recognition
- database
- em algorithm
- forward pass
- hidden markov
- hidden markov models
- hmms
- identification
- iterativc optimization
- iterative method
- kalman predictor
- likelihood identification
- maximum likelihood
- maximum likelihood identification
- modeling
- optimization
- parameter estimation
- phone classification
- phoneme classification
- predictor
- processing
- recognition
- regression
- search
- segmentation
- segmentation method
- signal processing
- smoothing
- speech production
- speech recognition
- statistical methods
- tile
- time warping
- training algorithm
Other assigned terms:
- approach
- case
- classification performance
- computational complexity
- conditional distribution
- continuous speech
- convergence
- correlation
- distribution
- duration
- estimation
- experimental results
- feature
- feature set
- feature space
- feature vector
- feature vectors
- formalism
- frame
- gaussian distribution
- likelihood
- likelihood function
- log-likelihood
- markov models
- method
- model parameters
- noise
- phoneme
- phonetic sequence
- priori
- probabilities
- probability
- process
- recursion
- segments
- sentences
- signal
- statistics
- stochastic segment model
- symbol
- term
- terms
- test data
- test set
- theory
- timit database
- training
- training data
- training set
- transformation
