Hidden Markov Models (HMMs) are statistical models of sequential data that have been used successfully in many machine learning applications, especially for speech recognition. Furthermore, in the last few years, many new and promising probabilistic models related to HMMs have been proposed. We first summarize the basics of HMMs, and then review several recent related learning algorithms and extensions of HMMs, including in particular hybrids of HMMs with artificial neural networks, Input-Output HMMs (which are conditional HMMs using neural networks to compute probabilities), weighted transducers, variable-length Markov models and Markov switching state-space models. Finally, we discuss some of the challenges of future research in this very active area. 1 Introduction Hidden Markov Models (HMMs) are statistical models of sequential data that have been used successfully in many applications in artificial intelligence, pattern recognition, speech recognition, and modeling of biological ...

: We present a method for utilizing soft information in decoding of variable length codes (VLCs). When compared with traditional VLC decoding, which is performed using "hard" input bits and a state machine, the soft-input VLC decoding offers improved performance in terms of packet and symbol error rates. Soft-input VLC decoding is free from the risk, encountered in hard decision VLC decoders in noisy environments, of terminating the decoding in an unsynchronized state, and it offers the possibility to exploit a priori knowledge, if available, of the number of symbols contained in the packet. 1 Introduction In most applications of variable length codes (VLCs), decoding is performed bit by bit, with the input to the entropy decoder assumed to be a sequence of "hard" bits about which no soft information is available. However, in noisy environments, soft information can be associated with each information bit, either by direct use of channel observations in the case of uncoded transmission...

vi Abbreviations vii Notation viii 1 Introduction 1 2 Hidden Markov Models 4 2.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2.2 HMM Modelling Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.3 HMM Topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.4 Finding the Best Transcription . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.5 Setting the Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 2.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 3 Objective Functions 19 3.1 Properties of Maximum Likelihood Estimators . . . . . . . . . . . . . . . . . . . 19 3.2 Maximum Likelihood . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 3.3 Maximum Mutual Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 3.4 Frame Discrimination . . . . . . . . . . . . . . . . ....

Discriminative training techniques for Hidden-Markov Models were recently proposed and successfully applied for automatic speech recognition. In this paper a discussion of the Minimum Classification Error and the Maximum Mutual Information objective is presented. An extended reestimation formula is used for the HMM parameter update for both objective functions. The discriminative training methods were utilized in speaker independent phoneme recognition experiments and improved the phoneme recognition rates for both discriminative training techniques. 1. INTRODUCTION Recently discriminative training techniques for Hidden- Markov Models (HMM) were used successfully for automatic speech recognition. They provide better performance compared to Maximum Likelihood Estimation (MLE), since the training is concentrated on the estimation of class boundaries and not on parameters of assumed model distributions [1,12]. Although MLE and discriminative training are theoretically equivalent (if su...

Most approaches in forecasting merely try to predict the next value of the time series. In contrast, this paper presents a framework to predict the full probability distribution. It is expressed as a mixture model: the dynamics of the individual states is modeled with so-called "experts" (potentially nonlinear neural networks), and the dynamics between the states is modeled using a hidden Markov approach. The full density predictions are obtained by a weighted superposition of the individual densities of each expert. This model class is called "hidden Markov experts". Results are presented for daily S&P500 data. While the predictive accuracy of the mean does not improve over simpler models, evaluating the prediction of the full density shows a clear out-ofsample improvement both over a simple GARCH(1,1) model (which assumes Gaussian distributed returns) and over a "gated experts" model (which expresses the weighting for each state nonrecursively as a function of external inputs). Sev...

This paper considers the problems of estimating bigram language models and of efficiently representing them by a finite state network, which can be employed by an hidden Markov model based, beam-search, continuous speech recognizer.

This paper describes an Automatic Speech Understanding (ASU) system used in a human-robot interface for the remote control of a mobile robot. The intended application is that of an operator issuing telecontrol commands to one or more robots from a remote workstation. ASU is supposed to be performed with spontaneous continuous speech and quasi real time conditions. Training and testing of the system was based on speech data collected by means of Wizard of Oz simulations. Two kinds of robustness factors are introduced: the first is a recognition error-tolerant approach to semantic interpretation, the second is based on a technique for evaluating the reliability of the ASU system output with respect to the input utterance. Preliminary results are 90.9% of correct semantic interpretations, and 89.1% of correct detection of out-of-domain sentences at the cost of rejecting 16.4% of correct in-domain sentences. 1. INTRODUCTION This paper describes an Automatic Speech Understanding (ASU) sys...

Stochastic Language Models (LMs) are key for achieving good performance in speech recognition systems. This is confirmed by the numerous LMs that have been proposed recently in the literature. This work compares three different LMs within the robot telecontrol speech understanding system developed at IRST. 1.1 Introduction This work compares three different class-based bigram LMs inside an Automatic Speech Understanding (ASU) system developed at IRST, which provides a voice interface to a robot telecontrol station [1] 2 . The LMs considered are the naive one, simply based on conditional frequencies, the recently proposed backing-off model by Placeway et al. [7] and the interpolated model by Derouault and Merialdo [5]. Robustness of each LM was evaluated against increasing sparseness of the training data, by augmenting the number of word classes assigned to the vocabulary. This issue is crucial for applications in which only small text corpora are available. Results are given in ter...

Central problems in the field of computer vision are learning object models from examples, classification, and localization of objects. In this paper we will motivate the use of a classical statistical approach to deal with these problems: the missing information principle. Based on this general technique we derive the Expectation Maximization algorithm and deduce statistical methods for learning objects from invariant features using Hidden Markov Models and from non-invariant features using Gaussian mixture density functions. The derived training algorithms will also include the problem of learning 3D objects from two-dimensional views. Furthermore, it is shown how the position and orientation of a three-dimensional object can be computed. The paper concludes with some experimental results.

In this paper we describe anomaly-based intrusion detection as a specialized case of the more general behavior detection problem.