Technological approaches for addressing privacy concerns when recognizing eating behaviors with wearable cameras

Abstract

First-person point-of-view (FPPOV) images taken by wearable cameras can be used to better understand people's eating habits. Human computation is a way to provide effective analysis of FPPOV images in cases where algorithmic approaches currently fail. However, privacy is a serious concern. We provide a framework, the privacy-saliency matrix, for understanding the balance between the eating information in an image and its potential privacy concerns. Using data gathered by 5 participants wearing a lanyardmounted smartphone, we show how the framework can be used to quantitatively assess the effectiveness of four automated techniques (face detection, image cropping, location filtering and motion filtering) at reducing the privacyinfringing content of images while still maintaining evidence of eating behaviors throughout the day.

Privacy behaviors of lifeloggers using wearable cameras

“…Thomaz et al [30] proposed a privacy saliency matrix to guide which images created by a lifelogger may pose privacy threats and should therefore be protected. Their model focuses on a specific setting related to eating behaviors in the context of a user study whereas we consider lifeloggers who share images for social reasons.…”

“…Complementary work includes defensive frameworks in which users define policies based on physical location, so that photos taken in predefined sensitive spaces can be recognized and then deleted or quarantined for review [27]. Researchers have also qualitatively studied reactions of bystanders to wearable camera devices [10,20], as well as the sensitivity of lifelogging data and how it could be automatically altered to enable privacy-preserving processing [30]. However, we do not have a clear understanding of the privacy behaviors and attitudes of the 'lifeloggers' using these devices, or the kinds of images that people will perceive as sensitive, either for themselves or others.…”

Collecting Shared Experiences through Lifelogging: Lessons Learned

“…Reviewing lifelogging data offers a chance to reflect on everyday experiences in a manner not easily afforded otherwise, providing the opportunity to draw new insights and learning, informing future behavior. Unsurprisingly, studies of lifelogging experiences are already an important focus for research [9,10,11,19], particularly in the areas of human autobiographical memory [9,15,17,19] and health-behavior monitoring [13,18].…”

“…This highlights the difficulty of relying on users filtering captured data and suggests that solutions that (i) automatically filter data, or (ii) those that restrict capture, may be more appropriate. Prior work indicates that automatic filtering of lifelogging photos based the presence of faces can be effective at reducing bystander privacy risks, as can cropping techniques that remove the bystanders but keep the remaining image content [18]. However, privacy infringements vary with person and context, and in many cases those very items that pose a privacy risk are themselves most valuable as a memory trigger (e.g.…”

“…For example, photo capture technologies such as Microsoft's SenseCam [17] and more recently Narrative Clips [16] can be used to provide photographic records of activities, while GPS tracking and health appliances capture physical motion. The proliferation of such data enables researchers to explore new applications and interaction patterns with computers such as memory enhancement [5,9] and health-behavior monitoring [13,18]. However, obtaining pervasive sensing datasets, particularly photographic datasets, is extremely challenging.…”