Hidden Markov Models for Spatio-Temporal Pattern Recognition and Image Segmentation
Brian C. Lovell
The Intelligent Real-Time Imaging and Sensing (IRIS) Group
The School of Information Technology and Electrical Engineering
The University of Queensland, Australia QLD 4072
lovell@itee.uq.edu.au
Abstract
Time and again hidden Markov models have been demonstrated to be highly effective in one-dimensional pattern recognition and classification problems such as speech recognition. A great deal of attention is now focussed on 2-D and possibly 3-D applications arising from problems encountered in computer vision in domains such as gesture, face, and handwriting recognition. Despite their widespread usage and numerous successful applications, there are few analytical results which can explain their remarkably good performance and guide researchers in selecting topologies and parameters to improve classification performance.