Dynamic Score Combination: A Supervised and Unsupervised Score Combination Method

Abstract

In two-class score-based problems the combination of scores from an ensemble of experts is generally used to obtain distributions for positive and negative patterns that exhibit a larger degree of separation than those of the scores to be combined. Typically, combination is carried out by a “static” linear combination of scores, where the weights are computed by maximising a performance function. These weights are equal for all the patterns, as they are assigned to each of the expert to be combined. In this paper we propose a “dynamic” formulation where the weights are computed individually for each pattern. Reported results on a biometric dataset show the effectiveness of the proposed combination methodology with respect to “static” linear combinations and trained combination rules

Competition on counter measures to 2-D facial spoofing attacks

“…), and we used a set of support vector machines (SVMs) to compute a frame level confidence score of being a real session or not. To obtain an high separation between score distributions we combined these similarity scores by means of the Dynamic Score Combination methodology [25]. This type of analysis can be performed also frame by frame.…”

A Deep Learning Approach for Analyzing Video and Skeletal Features in Sign Language Recognition

“…To this end, we test majority voting (MV), which is a well-known technique that accepts as class of a tested video sequence, the one with the most votes from the employed classifiers. Furthermore, inspired by the work of [22], we employ Dynamic Score Combination (DSC) [23] [24] and Particle Swarm Optimization (PSO) [25]. DSC attempts to combine the individual probabilities in a way that the combined probability distribution exhibits a larger separation than the probability distribution produced by the individual classifiers.…”

The 2nd competition on counter measures to 2D face spoofing attacks

“…For the both stages the team used Dynamic Score Combination [22] as a score-level fusion rule that allows dynamically choosing the best scores and weights to be combined. It requires computation of a weight parameter via majority voting [21].…”