A Proposed Method for Pattern Classification with HMM in the Context of Supervised Learning

Abstract

HMM was used in solving difficult problems in pattern recognition such as speech recognition, handwriting, postures and gestures recognition. This paper presents a proposed method based on HMM for a more general problem of pattern recognition, namely one in which the patterns are represented by points in a space with p dimensions. The first part of the paper is focused on the main elements of HMM theory. The following provides a formalism for the automatic classification of patterns in the context of supervised learning and the method by which this can be solved with HMM. The proposal of methodology is validated using GPL Jahmm library that has been extended with a new class for evaluating the performances of pattern classification with HMM.

Keywords

• Hidden Markov Models
• pattern recognition
• supervised learning
• pattern classification
• Jahmm Library
• Java