Subspace Gaussian Mixture Models Based on Noise Compensation for Speech Recognition.
Mohamed BouallegueDriss MatroufGeorges LinarèsMickael RouvierPublished in: INTERSPEECH (2012)
Keyphrases
- speech recognition
- gaussian mixture model
- speaker identification
- speaker recognition
- noisy environments
- mixture model
- covariance matrices
- feature space
- noisy speech
- language model
- speech signal
- hidden markov models
- speech recognizer
- mel frequency cepstral coefficients
- speech recognition systems
- speaker independent
- automatic speech recognition
- pattern recognition
- maximum likelihood
- feature vectors
- probabilistic neural network
- noise level
- em algorithm
- expectation maximization
- feature extraction
- speaker verification
- background noise
- probability density function
- speech synthesis
- gaussian distribution
- bayesian information criterion
- low dimensional
- noise reduction
- data mining
- dimensionality reduction
- high dimensional
- computer vision