Feature generation based on maximum normalized acoustic likelihood for improved speech recognition.
Xiang LiRichard M. SternPublished in: ICASSP (1) (2004)
Keyphrases
- speech recognition
- feature generation
- speech recognition systems
- speaker independent
- speech recognizers
- text categorization
- acoustic models
- inductive learning
- hidden markov models
- speech recognizer
- automatic speech recognition
- pattern recognition
- language model
- speech signal
- statistical approaches
- generation method
- maximum likelihood
- speech synthesis
- inductive logic programming
- similarity measure
- feature representations
- feature selection
- word sense disambiguation
- information extraction
- speaker identification
- neural network
- co training
- acoustic features
- image representation
- image classification
- semantic features
- knn