Chapter 9

The Use of Hidden Markov Models for Image Recognition: Learning with Artificial Ants, Genetic Algorithms and Particle Swarm Optimization

9.1. Introduction

Hidden Markov models (HMMs) are statistical tools which are used to model stochastic processes. These models are used in several different scientific domains [CAP 01] such as speech recognition, biology and bioinformatics, image recognition, document organization and indexing as well as the prediction of time series, etc. In order to use these HMMs efficiently, it is necessary to train them to be able to carry out a specific task. In this chapter we will show how this problem of training an HMM to carry out a specific task can be resolved with the help of several different population-based metaheuristics.

In order to explain what HMMs are, we will introduce the principles, notations and main algorithms which make up the theory of hidden Markov models. We will then continue this chapter by introducing the different metaheuristics which have been considered to train HMMs: an evolutionary method, an artificial ant algorithm and a particle swarm technique. We will finish the chapter by analyzing and evaluating six different adaptations of the above metaheuristics that enable us to learn HMMs from data which comes from the images.

9.2. Hidden Markov models (HMMs)

HMMs have existed for a long time. They were defined in 1913 when A. A. Markov first designed what we know as Markov chains [MAR 13]. The first efficient HMM algorithms ...

Get Optimisation in Signal and Image Processing now with the O’Reilly learning platform.

O’Reilly members experience live online training, plus books, videos, and digital content from nearly 200 publishers.