Mel cepstral coefficient modification based on the Glimpse Proportion measure for improving the intelligibility of HMM-generated synthetic speech in noise.
Cassia Valentini-BotinhaoJunichi YamagishiSimon KingPublished in: INTERSPEECH (2012)
Keyphrases
- speech recognition
- noisy environments
- cepstral coefficients
- noisy speech
- hidden markov models
- speech signal
- automatic speech recognition
- speech synthesis
- speech processing
- speech recognizer
- language model
- speech enhancement
- speaker independent
- automatic speech recognition systems
- speaker identification
- background noise
- keyword spotting
- signal to noise ratio
- noise level
- pattern recognition
- figure of merit
- recognition engine
- speech recognition systems
- noise reduction
- linear prediction
- noise model
- distance measure
- missing data
- automatically generated
- real world
- multi modal
- noisy data
- speaker verification
- random noise
- handwriting recognition
- median filter
- correlation coefficient
- gaussian noise