作者: Rawan Alharbi , Mariam Tolba , Lucia C. Petito , Josiah Hester , Nabil Alshurafa
DOI: 10.1145/3351230
关键词:
摘要: Activity-oriented cameras are increasingly being used to provide visual confirmation of specific hand-related activities in real-world settings. However, recent studies have shown that bystander privacy concerns limit participant willingness wear a camera. Researchers investigated different image obfuscation methods as an approach enhance privacy; however, these may varying effects on the utility image, which we define ability human viewer interpret activity wearer image. Visual is needed annotate and validate for several behavioral-based applications, particularly cases where loop method label (e.g., annotating gestures cannot be automatically detected yet). We propose new type obfuscation, activity-oriented partial methodological contribution researchers interested obtaining wild. tested this by collecting ten diverse realistic video scenarios involved performing while bystanders performed could concern if recorded. Then conducted online experiment with 367 participants evaluate effect degrees utility. Our results show (1) maintains wearer's activity, especially when object present hand, even extreme filters applied, (2) significantly reducing enhancing privacy. Informed our analysis, further discuss impact filter concerns.