Master Thesis - Department of Computer Science
Master Thesis - Department of Computer Science
Master Thesis - Department of Computer Science
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
CHAPTER 5<br />
Enhancing Decision Combination <strong>of</strong> Face<br />
and Fingerprint by Exploitation <strong>of</strong><br />
Individual Classifier Space: An approach<br />
to Multimodal Biometry<br />
This chapter presents a new approach to combine decisions from face and fingerprint<br />
classifiers by exploiting the individual classifier space on the basis <strong>of</strong> availability <strong>of</strong><br />
class specific information present in the classifier’s output space. We use the response<br />
vectors on a validation set for enhancing class separability (using linear discriminant<br />
analysis and nonparametric linear discriminant analysis) in the classifier output space<br />
and thereby improving performance <strong>of</strong> the face classifier. Fingerprint classifier <strong>of</strong>ten<br />
does not provide this information due to high sensitivity to minutiae points, pro-<br />
ducing partial matches across subjects. The enhanced face and fingerprint classifiers<br />
are combined using sum rule. We also propose a generalized algorithm for Multiple<br />
Classifier Combination (MCC) based on our approach. Experimental results show<br />
superiority <strong>of</strong> the proposed method over other already existing fusion techniques like<br />
sum, product, max, min rules, decision template and Dempster-Shafer theory. The<br />
rest <strong>of</strong> the chapter is organized as follows: Section 5.1 provides brief description on a<br />
few relevant works attempted for decision combination and a succinct introduction to<br />
our proposed method <strong>of</strong> decision fusion. Section 5.2 gives the theoretical formulation<br />
<strong>of</strong> our proposed classifier combination technique; Section 5.3 describes our method <strong>of</strong><br />
exploitation <strong>of</strong> base classifier using prior information and then combine them; Sec-<br />
tion 5.4 presents the experimental results to support our methodology and Section 5.5<br />
concludes this chapter. The description <strong>of</strong> other decision combination strategies used<br />
in our experimentation for comparison are provided in Section 2.3.3.3.