BACHELOR Assignment
LR histogram classification
Type: Bachelor EE
Period: Apr, 2018 - TBD
Student: Tijink, M.L. (Melissa, Student B-EE)
Supervisors:
Introduction:
In biometrics, often histograms are used to represent biometric samples (example pictures of faces). Subsequently the histograms are compared by a classifier to determine whether the biometric samples originate from the same individual or not. Nowadays several classifiers are used for this purpose, however most of them are relatively simple. Our starting point is a new classifier that computes a likelihood ratio, assuming that the histograms can be modelled as a sample from a multinomial distribution.
The assignment is:
- To check model assumptions by generating synthetic data from a multinomial distribution with the parameters estimated from real data and comparing the recognition performances obtained with real and synthetic data.
- To compare the proposed classifier with other trained classifiers (e.g. SVM, PCA/LDA).