Reconstruction of missing features by means of multivariate Laplace distribution (MLD) for noise robust speech recognition.
Arash MohammadiFarshad AlmasganjPublished in: Expert Syst. Appl. (2011)
Keyphrases
- speech recognition
- noisy environments
- speech recognition systems
- hidden markov models
- speech signal
- language model
- automatic speech recognition
- speech synthesis
- pattern recognition
- speech processing
- missing data
- cepstral coefficients
- speech recognizer
- background noise
- feature extraction
- speech recognition technology
- speech retrieval
- feature set
- speaker independent
- low level
- noisy speech
- signal to noise ratio
- mel frequency cepstral coefficients
- keyword spotting
- speaker dependent
- speech recognition errors
- neural network
- isolated word
- audio visual speech recognition