A hierarchical model for the evaluation of biometric sample quality

The evaluation of biometric sample quality is of great importance in the evaluation of biometric algorithms. In this paper, we propose a novel hierarchical model to compute the sample quality at three levels. This model is developed on the basis of three types of influencing factors: global factors, subjective factors and variable factors. We adopt different strategies to compute the corresponding three level qualities: database level quality, class level quality and image level quality. The database level quality is estimated by experience. Then, we compute the mean value of variable number of normalized genuine scores, the quantiles of which are used to determine the class level quality. On the image level quality evaluation, a novel concept of subset frequency is proposed.