Spectral subtraction with full-wave rectification and likelihood controlled instantaneous noise estimation for robust speech recognition.
Haitian XuZheng-Hua TanPaul DalsgaardBørge LindbergPublished in: INTERSPEECH (2004)
Keyphrases
- speech recognition
- spectral subtraction
- noise estimation
- speech signal
- noisy environments
- noisy images
- speech synthesis
- automatic speech recognition
- hidden markov models
- language model
- noise level
- noisy speech
- pattern recognition
- speaker identification
- denoising
- additive noise
- noise model
- vocal tract
- speaker verification
- maximum likelihood
- signal processing
- computer vision