On the jointly unsupervised feature vector normalization and acoustic model compensation for robust speech recognition.
Luis BueraAntonio MiguelEduardo LleidaOscar SazAlfonso OrtegaPublished in: INTERSPEECH (2007)
Keyphrases
- speech recognition
- feature vectors
- noisy environments
- hidden markov models
- speech synthesis
- speech processing
- language model
- speech signal
- speech recognition technology
- speech recognizer
- automatic speech recognition
- speaker identification
- speech retrieval
- pattern recognition
- speech understanding
- image features
- speech recognition systems
- speech recognizers
- neural network
- semi supervised
- feature space
- speaker independent
- speech recognition errors
- gaussian mixture model
- keyword spotting
- non stationary
- unsupervised learning
- em algorithm
- natural language
- feature extraction
- machine learning