A SVM-based model for the evaluation of biometric sample quality

One of the main factors affecting the performance of biometric systems is the quality of the acquired samples. Poor-quality samples increase the enrollment failure, and decrease the system performance. Therefore, it is important for a biometric system to estimate the quality of the acquired biometric samples. Toward this goal, we present in this paper a multi-class SVM-based method to predict sample quality. The proposed method uses two types of information: the first one is based on the image quality and the second is a pattern-based quality using the SIFT keypoints extracted from the image. For the experiments, we use four large and significant face databases to show the efficiency of the proposed method in predicting the system performance illustrated by the Equal Error Rate (EER).