This paper investigates employment of Subspace Gaussian Mixture Models (SGMMs) for acoustic model adaptation to-wards different accents for English speech recognition. The SGMMs comprise globally-shared and state-specific param-eters which can efficiently be employed for various kinds of acoustic

We describe an acoustic modeling approach in which all phonetic states share a common Gaussian Mixture Model structure, and the means and mixture weights vary in a subspace of the total parameter space. We call this a Subspace Gaussian Mixture Model (SGMM). Globally shared parameters define

This paper studies cross-lingual acoustic modelling in the context of subspace Gaussian mixture models (SGMMs). SGMMs factorize the acoustic model parameters into a set that is globally shared between all the states of a hidden Markov model (HMM) and another that is specific to the HMM states. We

We consider a family of Gaussian mixture models for use in HMM based speech recognition system. These "SPAM" models have state independent choices of subspaces to which the precision (inverse covariance) matrices and means are restricted to belong. They provide a flexible tool for robust

This paper presents a generalized i-vector representation frame-work using the mixture of Gaussian (MoG) factor analysis for speaker verification. Conventionally, a single standard factor anal-ysis is adopted to generate a low rank total variability subspace where the mean supervector is assumed

We are surrounded by electronic devices that take advantage of wireless technologies, from our computer mice, which require little amounts of information, to our cellphones, which demand increasingly higher data rates. Until today, the coexistence of such a variety of services has been guaranteed by a fixed assignment of spectrum resources by regulatory agencies. This has resulted into a blind alley, as current wireless spectrum has become an expensive and a scarce resource. However, recent measurements in dense areas paint a very different picture: there is an actual underutilization of the spectrum by legacy sys-tems. Cognitive radio exhibits a tremendous promise for increasing the spectral efficiency for future wireless systems. Ideally, new secondary users would have a perfect panorama of the spectrum usage, and would opportunistically communicate over the available re-sources without degrading the primary systems. Yet in practice, monitoring the spectrum resources, detecting available resources for opportunistic communication, and transmit-ting over the resources are hard tasks. This thesis addresses the tasks of monitoring, de-

1 Introduction Mixture of Gaussians (MoG) are commonly used as the state representation in hidden Markov model (HMM) based speech recognition. These Gaussian mixture models are easy to train using expectation maximisation (EM) techniques [4] and are able to approximate any distribution given a

We consider a family of Gaussian mixture models for use in HMM based speech recognition system. These “SPAM ” models have state independent choices of subspaces to which the precision (inverse covariance) matrices and means are restricted to belong. They provide a flexible tool for robust, compact

This paper presents an efficient approximation of the Gaussian mixture state probability density functions of continuous observation density hidden Markov models (CHMM 's). In CHMM 's, the Gaussian mixtures carry a high computational cost, which amounts to a significant fraction (e.g. 30

This paper copes with the problem of nonlinear Bayesian state estimation. A nonlinear filter, the Sliced Gaussian Mixture Filter (SGMF), employs linear substructures in the nonlinear measurement and prediction model in order to simplify the estimation process. Here, a special density