Introduction to Pattern Recognition: A Matlab Approach

CHAPTER

6

Hidden Markov Models

6.1 INTRODUCTION

In general,a hidden Markov model (HMM) is a type of stochastic modeli ng appropriate for nonstationary

stochastic sequences whose statistical properties underg o distinct random transitions among a set of,

say, k different stationary processes. In other words, HMMs are used to model piecewise stationary

processes. A stationary process i s one whose statistical propert i es do not change with t i me.

We assume that we are given a set of observations (feature vectors), x

1

,x

2

,...,x

N

∈ R

l

. In contrast to

what has been assumed in previous chapters, here we allow each observation to be generated (emitted)

by