Novel robust feature extraction based on spectrally masked channel energy ratio (SMaChER) for speech recognition.
Changxue MaPublished in: ICASSP (2) (2003)
Keyphrases
- speech recognition
- feature extraction
- noisy environments
- speaker identification
- pattern recognition
- cepstral coefficients
- hidden markov models
- automatic speech recognition
- speech understanding
- language model
- speech recognition technology
- speech synthesis
- speech processing
- speech signal
- speech recognizer
- speech recognition systems
- multi channel
- keyword spotting
- speech recognizers
- isolated word
- feature vectors
- single channel
- principal component analysis
- wavelet transform
- speech recognition errors
- feature space
- image processing
- feature selection