ACL RD-TEC 1.0 Summarization of H90-1066
Paper Title:
RECOGNITION OF NOISY SPEECH: USING MINIMUM-MEAN LOG-SPECTRAL DISTANCE ESTIMATION
RECOGNITION OF NOISY SPEECH: USING MINIMUM-MEAN LOG-SPECTRAL DISTANCE ESTIMATION
Authors: A. Erell and M. Weintraub
Primarily assigned technology terms:
- algorithm
- approximation
- classification
- database
- estimation algorithm
- estimator
- factorization
- hidden markov
- hidden markov model
- hmms
- markov model
- matching
- maximum likelihood
- minimum-mean log-spectral distance estimation
- mmlsd estimator
- modeling
- parameterization
- preprocessing
- processing
- recognition
- recognition system
- recognition system \
- recognition systems
- recognizer
- resource management
- robust speech recognition
- search
- signal processing
- spectral estimation
- speech enhancement
- speech recognition
- speech recognition systems
- speech-recognition
- template matching
- tuning
- vector quantization
Other assigned terms:
- additive noise
- approach
- case
- class probability
- conditional probability
- continuous-speech
- correlation
- correlations
- distance metric
- distribution
- error rate
- estimation
- euclidean distance
- feature
- feature vector
- grammar
- implementation
- labeling
- large training
- likelihood
- method
- mixture models
- nist
- noise
- noisy speech
- parameter values
- perplexity
- posteriori probability
- probabilities
- probability
- probability distribution
- probability distributions
- procedure
- recognition accuracy
- recognition task
- relation
- research and development
- resource management task
- sentences
- signal
- spectral probability distribution
- speech data
- speech recognition task
- speech signal
- speech vector
- statistics
- term
- test set
- training
- training set
- vocabulary
- white noise
- word
- word-pair grammar
- words
