| Abstract [eng] |
TOMKEVICIUS, Kestutis. (2008) Research of Applying Biometrics in E-commerce Authentication Modules. MBA Graduation Paper. Kaunas: Vilnius University, Kaunas Faculty of Humanities, Department of Informatics. 77 p. SUMMARY Contemporary society is highly dependent on machines and technology. As a result, the reassurance in information security becomes a global issue, when going to the internet world. Information security is closely associated with privacy protection and thus, is of high importance nowadays. There are several major tasks in the Master‘s thesis. First, this paper aims to investigate the world-wide standards of authentication systems and the devises used. Second, after the foregoing views on certain effects of the evaluation procedure and its possible drawbacks, it aims to explore, whether or not the speaker recognition based on the discrimination of phoneme is considered to be one of the most promising areas for the future research. In other words, the process of investigation seeks to find out whether it is possible to provide a thorough analysis of the speaker verification system and distinguish the significant phonemes in the subsets of different speaker phonemes that would allow forming the characteristic password phrases and if they would, how these phonemes were to be selected. In the feature-extraction stage of the speaker verification system, the Mel frequency cepstral coefficients (MFCC) give the bases for the research paper and various most suitable phonemes of the speaker are counted, while using the likelihood methods. The results of experiments indicate that variations in gender of speaker-based phoneme features as well as the proper recognition of the digital likelihood differ greatly. And the best way for the speaker recognition is the whole set of phonemes of the second and fourth quarter phonemes. The Master‘s thesis consists of 77 pages. This paper includes 23 tables and 37 pictures. |
