Year of Graduation
A Program for Multifactorial Biometric Person Identification
School of Software Engineering
Pages: 28, 5 chapters, 3 images, 10 sources, 4 appendices.Keywords: identification, biometric factors, eigenface, fisherface, mel-frequency cepstral coefficients, vector quantization.Object of this project is a program for multifactorial biometric person identification by two factors: person face and voice.The main goal of the project is to develop a system for multifactorial biometric person identification, which will use an image of face and a recording of voice as identification factors; work out an approach for combining results of identification. Using both factors the increase in correct identification rate can be achieved, and a range of environmental conditions, in which identification can be performed, widened.In a research section an analysis of the existing approaches for identification person by face and voice was performed, algorithms were chosen for implementation. In a scope of practical section an identification system, which allows user to obtain necessary biometric data from web-camera and microphone, was developed.As a result, a practical increase of correct identification rate of 7,5 % in comparison with an image-identification and 2,5 % in comparison with a voice identification was achieved.This system can be applied in different spheres, starting with a natural machine-human interfaces, finishing with fast identification at enterprise.This project has great potential for extension and cross-platform development. Firstly, this program can be extended into a complete biometric identification complex, which will use additional factors, such as fingerprints and retinal image. Current rates of mobile platforms development allow to adopt this program for them also.