HMM-based MAP prediction of voiced and unvoiced formant frequencies from noisy MFCC vectors.
Jonathan DarchBen MilnerPublished in: INTERSPEECH (2006)
Keyphrases
- speech signal
- formant frequencies
- linear prediction
- noisy environments
- hidden markov models
- speech recognition
- automatic speech recognition
- speaker identification
- background noise
- prediction accuracy
- vocal tract
- spectral analysis
- speech quality
- non stationary
- speaker verification
- prediction model
- prediction error
- maximum a posteriori
- acoustic features
- prediction algorithm
- feature vectors
- mel frequency cepstral coefficients
- sound source
- input data
- noisy data
- vector space