Scientific journal
Scientific Review. Technical science
ISSN 2500-0799
ПИ №ФС77-57440

Personal portfolio

COMPARATIVE ANALYSIS OF HIDDEN MARKOV MODELS STRUCTURES USED IN THE PROBLEM OF THE IDENTIFICATION HUMAN BY FACE IMAGE

Dvoynoy I.R. 1 Salnikov I.I. 2
1 JSC New System
2 Penza State Technological University
In order to identify the structure of a hidden Markov model, which gives the best result in identification of human faces by the image, a comparative analysis was conducted. This article provides an overview of hidden Markov models, a description of specific areas of human face’s image. Algorithms, described in this article, show the usage of the left-right and pseudo-dual-in-line structure. To assess the quality of recognition, four different sets of test images (distributed free of charge in the Internet) have been used. The study found that the usage of pseudo-dual-in-line model gives a slight increase in the recognition quality of the system due to a more precise description of local deformations and mutual arrangement of individual parts of the image. A significant development of computational complexity impinging upon recognition system’s productivity is an additional constraint, restraining the usage of pseudo-dual-in-line models in real applications.