University of Twente Student Theses


Analysis of a Likelihood Ratio Classifier for Histogram Features

Tijink, M.L. (2019) Analysis of a Likelihood Ratio Classifier for Histogram Features.

This is the latest version of this item.

[img] PDF
Abstract:In this paper an extended research is done on a new likelihood ratio (LLR) classifier as proposed by [1]. We analyzed the assumption that histograms, representing biometric features, can be modeled as draws from a multinomial distribution and the parameters of this distribution can be modeled as the Dirichlet probability density. We extracted the estimated parameter vector of the Dirichlet density from real data by raining it with the LLR classifier and used this parameter vector to create new synthetic data. This synthetic data was used to train and test the classifier. In contrary to the results of training and testing with real data, the results with synthetic data were nearly perfect. This lead to the conclusion that the assumption of a multinomial distribution was not correct. Besides this, we compared the LLR classifier with another trained classifier: the support vector machine (SVM). In order to do 1:1 comparisons with the SVM, we concatonated the histograms. The results showed that the LLR classifier performed better, however after a simple experiment we concluded that concatenating the histograms does not result in a well functioning SVM.
Item Type:Essay (Bachelor)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:53 electrotechnology
Programme:Electrical Engineering BSc (56953)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page