Analysis of slap segmentation and HBSI errors across different force levels

摘要

In a recent study, fingerprint data was collected across six different force levels. A total of 8,877 slap samples were ground truthed, and subsequently processed using a commercially available fingerprint segmentation tool. This paper delves deeper into understanding segmentation errors across the force levels. Out of the 8,877 slaps, 370 slaps failed to segment. In order to understand why there were errors, a detailed analysis was undertaken by ground-truthing the slap and segmented datasets. In addition to the ground-truthing, video analysis of the interactions enabled specific failures to be replayed and identified as performance or ergonomic, interaction, and usability (Human Biometric Sensor Interaction) errors. This paper will identify determining factors that would cause the slap segmentation tool to error, and provide guidance to those undertaking data collection activities where they would need to the use a segmentation tool.