No abstract

No abstract

The purpose of our work is to mine streaming data from a variety of hundreds of automotive sensors in order to develop methods to minimize driver distraction from in-vehicle communications and entertainment systems such as audio/video devices, cellphones, PDAs, Fax, eMail, and other messaging devices. Our endeavor is to create a safer driving environment, by providing assistance in the form of warning, delaying, or re-routing, incoming signals if the assistance system detects that the driver is performing, or is about to perform, a critical maneuver, such as passing, changing lanes, making a turn, or during a sudden evasive maneuver. To accomplish this, our assistance system relies on maneuver detection by continuously evaluating various embedded vehicle sensors, such as speed, steering, acceleration, lane distance, and many others, combined into representing an instance of the "state" of the vehicle. One key issue is how to effectively and efficiently monitor many sensors with constant data streams. Data streams have their unique characteristics and may produce data that is not relevant or pertinent to a maneuver. We propose an adaptive sampling method that takes advantage of these unique characteristics and develop algorithms that attempt to select relevant and important instances to determine which sensors to monitor and how to provide quick and effective responses to this type of mission critical situations. This work can be extended to many similar sensor applications with data streams.

No abstract

Non-invasive identity inference in the home environment is a very challenging problem. A practical solution to the problem could have far reaching implications in many industries, such as home entertainment. In this work, we consider the problem of identity inference using a TV remote control. In particular, we address two challenges that have so far prevented the work of Chang et al. (2009) from being applied in a home entertainment system. First, we show how to learn the patterns of TV remote controls incrementally and online. Second, we generalize our results to partially labeled data. To achieve our goal, we use state-of-the-art methods for max-margin learning and online convex programming. Our solution is efficient, runs in real time, and comes with theoretical guarantees. It performs well in practice and we demonstrate this on 4 datasets of 2 to 4 people.