A simple and efficient method to align very long speech signals to acoustically imperfect transcriptions.
Germán BordelMikel PeñagarikanoLuis Javier Rodríguez-FuentesAmparo VaronaPublished in: INTERSPEECH (2012)
Keyphrases
- speech signal
- automatic speech recognition
- speech recognition
- background noise
- vocal tract
- noisy environments
- automatic speech recognition systems
- hidden markov models
- non stationary
- linear prediction
- additive noise
- speech enhancement
- adaptive filtering
- broadcast news
- blind source separation
- speech sounds
- language model
- information retrieval
- spectral analysis
- speech quality
- feature selection
- pattern recognition
- fundamental frequency