This page describes Discrete Hidden Markov Models (DHMM) toolbox for pattern recognition which was written as a diploma thesis in February 2000 and has been updated since. The toolbox provides algorithms for recognition, supervised and unsupervised learning, which works with Discrete Hidden Markovian Models of sequences and acyclic stuctures. Used algorithms were published in the monograph M.I. Schlesinger, V. Hlavac: Ten lectures from the statistical and structural pattern recognition, in chapter 8 (further referenced as SH10 ). Toolbox funtions work with four types of sequence statistical models and two types of statistical models of acyclic stuctures. This is a public domain software. The toolbox is built on top of the Matlab, version 5.2 or higher.
Master thesis (in czech) / Diplomova prace (cesky) DownLoad
Toolbox DownLoad
(Last change April 2001)
Application of the toolbox to the OCR example DownLoad
Toolbox description (On line help)