ACL RD-TEC 1.0 Summarization of H94-1065
Paper Title:
ADAPTATION TO NEW MICROPHONES USING TIED-MIXTURE NORMALIZATION
ADAPTATION TO NEW MICROPHONES USING TIED-MIXTURE NORMALIZATION
Authors: Anastasios Anastasakost and Francis Kubala and John Makhoul and Richard Schwartz
Primarily assigned technology terms:
- algorithm
- byblos speech recognition
- continuous speech recognition
- database
- feature analysis
- feature mapping
- hidden markov
- hidden markov model
- hmms
- markov model
- mean subtraction
- modeling
- normalization
- processing
- recognition
- recognition system
- recognition system \
- recognition systems
- signal processing
- smoothing
- speaker adaptation
- speaker adaptation \
- speech recognition
- speech recognition system
- speech recognition systems
- statistical methods
- transfer function
- vector quantization
- weighting
Other assigned terms:
- acoustic models
- approach
- case
- cluster
- cluster centroid
- clusters
- codebook
- continuous speech
- corpora
- density function
- distribution
- error rate
- evaluation test
- feature
- frame
- gaussian distribution
- gaussian mixture
- generation
- hmm model
- implementation
- index
- language models
- mapping
- maps
- measure
- method
- noise
- probabilities
- probability
- probability density
- probability density function
- process
- recognition task
- sentences
- signal
- speech corpus
- speech data
- speech recognition task
- statistics
- system performance
- technique
- technologies
- test data
- test set
- training
- training data
- transformation
- utterance
- vocabulary
- word
- word error rates
