Combination of GMM-based speech estimation method and temporal domain SVD-based speech enhancement for noise robust speech recognition.
Masakiyo FujimotoYasuo ArikiPublished in: Systems and Computers in Japan (2007)
Keyphrases
- speech recognition
- noisy environments
- noisy speech
- speech enhancement
- speech signal
- spectral subtraction
- speaker recognition
- speaker identification
- hidden markov models
- single channel
- noise reduction
- additive noise
- vocal tract
- speech recognizer
- background noise
- similarity measure
- speech synthesis
- pattern recognition
- gaussian mixture model
- edge detection
- linear prediction
- prediction error
- signal to noise ratio
- language model
- denoising