Accession Number:

ADA543331

Title:

Using Evidence Feed-Forward Hidden Markov Models

Descriptive Note:

Corporate Author:

ARMY TANK AUTOMOTIVE RESEARCH DEVELOPMENT AND ENGINEERING CENTER WARREN MI

Report Date:

2010-05-11

Pagination or Media Count:

7.0

Abstract:

Visual Understanding is an increasing field of research thanks to the advances in image processing, object detection, classification, and advanced computational intelligence techniques. Hidden Markov Models HMM are one of these techniques which have been used extensively for this problem. This paper will introduce a new type of HMM, called Evidence Feed Forward Hidden Markov Models, that not only increase the classification rate for sparse messy data, but outlines a whole new theory towards changing the way HMMs are conceived. Data is taken from simulated images of peoples actions. Over processing is performed to decrease the likelihood of correct classification. Finally, the over processed, sparse data is used to train and test the Evidence Feed-Forward HMM and the standard HMM. Results are compared.

Subject Categories:

  • Statistics and Probability

Distribution Statement:

APPROVED FOR PUBLIC RELEASE