The problem of reducing a Hidden Markov Model (HMM) to one smaller dimension that exactly reproduces the same marginals is tackled by using system-theoretic approach. Realization theory tools are extended HMMs leveraging suitable algebraic representations probability spaces. We propose two algorithms return coarse-grained equivalent obtained stochastic projection operators: first returns models...

This paper introduces a new form of observation distributions for hidden Markov models (HMMs), combining subvector quantization and mixtures of discrete distributions. We present efficient training and decoding algorithms for the discretemixture HMMs (DMHMMs). Our experimental results in the airtravel information domain show that the high-level of recognition accuracy of continuous mixture-dens...

AND TURBO CODES Javier Garcia-Frias and John D. Villasenor Electrical Engineering Department University of California, Los Angeles Abstract|Hidden Markov models have been widely used to statistically characterize sources and channels in communication systems. In this paper we will consider their application in the context of turbo codes. We will describe simpli ed techniques for modifying a dec...

In this paper, we describe the first Mandarin/Taiwanese (Min-nan) bi-lingual, continuous speech recognition system for large vocabulary or vocabulary-independent applications. A phonetic transcription system called Tong-yong Phonetic Alphabet (TYPA) is described and used to transcribe the bilingual Mandarin/Taiwanese lexicons. The Right-ContextDependent (RCD) phonetic continuous-density Hidden ...

Linear discriminant or Karhunen-Lo eve transforms are established techniques for mapping features into a lower dimensional subspace. This paper introduces a uniform statistical framework, where the computation of the optimal feature reduction is formalized as a Maximum-Likelihood estimation problem. The experimental evaluation of this suggested extension of linear selection methods shows a slig...

Abstract In this paper, we study the use of continuous-time hidden Markov models for network protocol and application performance evaluation. We develop an algorithm to infer the continuous-time hidden Markov model from a series of end-to-end delay and loss observations of probe packets. This model can then be used to simulate network environments for network performance evaluation. We validate...

The health status evolving from normal to broken condition of wear tool are needed as an object of assessment in condition-based maintenance (CBM). This paper proposes a continuous Hidden Markov Models (CHMM) to assess the status of the wear tool online based on the normal dataset in the same case. A waveletpackets technology is used to feature extraction and the CHMM is trained by Baum-Welch a...

The paper studies large sample asymptotic properties of the Maximum Likelihood Estimator (MLE) for the parameter of a continuous time Markov chain, observed in white noise. Using the method of weak convergence of likelihoods due to I.Ibragimov and R.Khasminskii [14], consistency, asymptotic normality and convergence of moments are established for MLE under certain strong ergodicity conditions o...

Recently, there has been a surge of interest in using spectral methods for estimating latent variable models. However, it is usually assumed that the distribution of the observations conditioned on the latent variables is either discrete or belongs to a parametric family. In this paper, we study the estimation of an m-state hidden Markov model (HMM) with only smoothness assumptions, such as Höl...