We consider the problem of parameter estimation in full-covariance Gaussian mixture systems for automatic speech recognition. Due to the high dimensionality of the acoustic feature vector, the standard sample covariance matrix has a high variance and is often poorly-conditioned when the amount of training data is limited. We explain how the use of a shrinkage estimator can solve these problems, and derive a formula for the optimal shrinkage intensity. We present results of experiments on a phone recognition task, showing that the estimator gives a performance improvement over a standard full-covariance syste
Nowadays, almost all speaker-independent (SI) speech recognition systems use CDHMM with multivariate...
Hidden Markov models (HMMs) for automatic speech recognition rely on high dimensional feature vector...
We propose to use sparse inverse covariance matrices for acoustic model training when there is insuf...
HMM-based systems for Automatic Speech Recognition typically model the acoustic features using mixt...
Recently various techniques to improve the correlation model of feature vector elements in speech re...
Recently various techniques to improve the correlation model of feature vector elements in speech re...
Summarization: A trend in automatic speech recognition systems is the use of continuous mixture-dens...
An estimation of parameters of a multivariate Gaussian Mixture Model is usually based on a criterion...
In conventional hidden Markov model (HMM) based speech recognisers, the emitting HMM states are mode...
Abstract—We address covariance estimation in the sense of minimum mean-squared error (MMSE) when the...
Senones were introduced to share Hidden Markov model (HMM) parameters at a sub-phonetic level in [3]...
Covariance estimation is a key step in many target detection algorithms. To distinguish target from ...
In most HMM-based recognition systems, a mixture of diagonal covariance gaussians is used to model t...
[[abstract]]© 1998 Elsevier - The projection-based likelihood measure, an effective means of reducin...
Speech recognizers often experience serious performance degradation when d ployed in an unknown acou...
Nowadays, almost all speaker-independent (SI) speech recognition systems use CDHMM with multivariate...
Hidden Markov models (HMMs) for automatic speech recognition rely on high dimensional feature vector...
We propose to use sparse inverse covariance matrices for acoustic model training when there is insuf...
HMM-based systems for Automatic Speech Recognition typically model the acoustic features using mixt...
Recently various techniques to improve the correlation model of feature vector elements in speech re...
Recently various techniques to improve the correlation model of feature vector elements in speech re...
Summarization: A trend in automatic speech recognition systems is the use of continuous mixture-dens...
An estimation of parameters of a multivariate Gaussian Mixture Model is usually based on a criterion...
In conventional hidden Markov model (HMM) based speech recognisers, the emitting HMM states are mode...
Abstract—We address covariance estimation in the sense of minimum mean-squared error (MMSE) when the...
Senones were introduced to share Hidden Markov model (HMM) parameters at a sub-phonetic level in [3]...
Covariance estimation is a key step in many target detection algorithms. To distinguish target from ...
In most HMM-based recognition systems, a mixture of diagonal covariance gaussians is used to model t...
[[abstract]]© 1998 Elsevier - The projection-based likelihood measure, an effective means of reducin...
Speech recognizers often experience serious performance degradation when d ployed in an unknown acou...
Nowadays, almost all speaker-independent (SI) speech recognition systems use CDHMM with multivariate...
Hidden Markov models (HMMs) for automatic speech recognition rely on high dimensional feature vector...
We propose to use sparse inverse covariance matrices for acoustic model training when there is insuf...