QUALITY-BASED RANK FUSION METHOD

The logistic regression method, which is a variation of the Borda count method, calculates the weighted sum of the individual ranks. In this method, the final consensus rank is obtained by sorting the identities according to the sum of their ranks obtained from individual matchers multiplied by the weights.

Algorithm 5 shows the Borda score calculation process in this work. Algorithm 5: Logistic regression

o    Get the ranking lists from different biometric classifiers.

o    Assign different weights to all ranking lists.

o    For all ranking lists –

·         Step 3a: For all identities in the three ranking list

·         Step 3a(i): Find out the total Borda score of each identity utilizing the following equation

Quality-based rank fusion method depends not only on the ranking list of the unimodal classifiers, but also on the quality of the input images. Usually, this method applies on other biometric rank fusion approach with the modification by incorporating the quality of the input image. Quality based fusion methods usually do not have any training phase and hence can be used in other biometric information fusion process, such as fuzzy logic based fusion. There is no specific rule or general equation for quality based fusion method. Researchers can apply this method to any of their existing methods to improve the identification or verification rate. For example, Abaza and Ross introduced a quality based rank fusion method by modifying the existing Borda count method incorporating the quality of the input image into the equation.

In this chapter, different existing methodologies for rank level fusion methods for multimodal, biometric system have been reviewed. The methods for rank level fusion include plurality voting method, highest rank method, Borda count method, logistic regression method, and quality-based rank fusion method. Advantages and disadvantages of all of these rank fusion methods have been discussed in the context of current state of the art in the discipline. Also, with the help of appropriate diagrams, outcomes of different possible rank fusion methods have been shown. In the next chapter, a new rank fusion method, the Markov chain based rank fusion method will be discussed which has several advantages over the traditional rank fusion methods.