Combining perceptually-motivated spectral shaping with loudness and duration modification for intelligibility enhancement of HMM-based synthetic speech in noise.
Cassia Valentini-BotinhaoJunichi YamagishiSimon KingYannis StylianouPublished in: INTERSPEECH (2013)
Keyphrases
- speech recognition
- signal to noise ratio
- hidden markov models
- noisy environments
- variable duration
- speech enhancement
- speech signal
- spectral features
- noise reduction
- image processing
- noise level
- linear prediction
- image enhancement
- median filter
- noisy data
- background noise
- random noise
- image noise
- signal processing
- speech synthesis
- text to speech
- higher level
- speech quality
- real world
- speaker adaptation
- short time fourier transform
- linear predictive coding
- automatic speech recognition
- spoken language
- dialogue system
- noise model
- additive noise
- image restoration
- image quality