Assessing image characteristics for user feedback in biometric fingerprint identity verification tasks

摘要

Biometric systems seek to establish or verify individual identity based on the measurement and processing of physiological or behavioural characteristics. However, the performance of such systems may be compromised for a variety of reasons, and a common source of inaccuracy arises from poor quality data capture at the biometrie sensor. This leads to a requirement for appropriate strategies to manage this problem. In fingerprint processing, for example, poor quality of the captured image may impose a requirement for exclusion of some samples or for automatic image enhancement, but a further option is to provide user feedback to improve the quality of the interaction itself. This paper investigates a range of strategies for the provision of feedback, and compares these with respect to their effectiveness in practice.