Thursday, November 12th @ 4PM
Neel Sundaresan, Sr. Director and Head, eBay Research Labs
"Trust and Technology"
The use of mobile sensors is becoming increasingly commonplace in the everyday lives of people. Examples of sensors range from accelerometers, cameras and microphones on cell phones to special-purpose devices that capture physiological (e.g., EKG, GSR) or contextual information (e.g., temperature, location). These sensors can be used to automatically infer a range of human behavioral states (e.g., physical activities, emotional states, social interactions, and activities of daily living such as cooking or cleaning), the knowledge of which would be useful in a variety of applications, such as detecting anomalous and suspicious behavior, supporting emergency response efforts, tracking physical and cognitive health of individuals, and enhancing the user experience of social and communication technologies.
Much of the activity recognition research has so far focused on the analysis of one individual's sensor data in isolation. However, understanding the trends in activity patterns requires the analysis to move beyond individual to groups, which we believe will not only enable a broader range of applications but also reveal the privacy issues that are present in these scenarios. For example, can an individual be uniquely identified by his activities? Or are there enough similarities across people that can be exploited to anonymize a person's identity? In this project, we will focus on developing algorithms for discovering activity trends, deriving quantitative metrics for finding people who are behaviorally alike, and identifying possible strategies to address some of the privacy concerns that this research uncovers. Furthermore, we will explore the use of sensors on commercially available sensor-equipped mobile devices (such as the iPhone) and the possibility of tying some of our results into popular social applications (such as Facebook or ContextAwareIM).
