A minimum-mean-square-error noise reduction algorithm on Mel-frequency cepstra for robust speech recognition.
Dong YuLi DengJasha DroppoJian WuYifan GongAlex AceroPublished in: ICASSP (2008)
Keyphrases
- speech recognition
- noisy environments
- cepstral coefficients
- minimum mean square error
- signal to noise ratio
- speech signal
- noise reduction
- hidden markov models
- language model
- noisy speech
- speaker identification
- speech synthesis
- automatic speech recognition
- speech recognizer
- pattern recognition
- low frequency
- speech recognition systems
- noise level
- recursive algorithm
- denoising algorithm
- high frequency
- neural network
- speech enhancement
- image quality
- signal processing
- feature set
- feature extraction