Modern enterprise applications are forced to deal with unreliable, inconsistent and imprecise information. Probabilistic databases can model such data naturally, but SQL query evaluation on probabilistic databases is difficult: previous approaches have either restricted the SQL queries, or computed approximate probabilities, or did not scale, and it was shown recently that precise query evaluation is theoretically hard. In this paper we describe a novel approach, which computes and ranks efficiently the top-k answers to a SQL query on a probabilistic database. The restriction to top-k answers is natural, since imprecisions in the data often lead to a large number of answers of low quality, and users are interested only in the answers with the highest probabilities. The idea in our algorithm is to run in parallel several Monte-Carlo simulations, one for each candidate answer, and approximate each probability only to the extent needed to compute correctly the top-k answers. The algorithms is in a certain sense provably optimal and scales to large databases: we have measured running times of 5 to 50 seconds for complex SQL queries over a large database (10M tuples of which 6M probabilistic). Additional contributions of the paper include several optimization techniques, and a simple data model for probabilistic data that achieves completeness by using SQL views. 1

In this paper we present results related to achieving finegrained activity recognition for context-aware computing applications. We examine the advantages and challenges of reasoning with globally unique object instances detected by an RFID glove. We present a sequence of increasingly powerful probabilistic graphical models for activity recognition. We show the advantages of adding additional complexity and conclude with a model that can reason tractably about aggregated object instances and gracefully generalizes from object instances to their classes by using abstraction smoothing. We apply these models to data collected from a morning household routine. 1.

Learning patterns of human behavior from sensor data is extremely important for high-level activity inference. We show how to extract a person’s activities and significant places from traces of GPS data. Our system uses hierarchically structured conditional random fields to generate a consistent model of a person’s activities and places. In contrast to existing techniques, our approach takes high-level context into account in order to detect the significant places of a person. Our experiments show significant improvements over existing techniques. Furthermore, they indicate that our system is able to robustly estimate a person’s activities using a model that is trained from data collected by other persons. 1

Abstract. We are developing a personal activity recognition system that is practical, reliable, and can be incorporated into a variety of health-care related applications ranging from personal fitness to elder care. To make our system appealing and useful, we require it to have the following properties: (i) data only from a single body location needed, and it is not required to be from the same point for every user; (ii) should work out of the box across individuals, with personalization only enhancing its recognition abilities; and (iii) should be effective even with a cost-sensitive subset of the sensors and data features. In this paper, we present an approach to building a system that exhibits these properties and provide evidence based on data for 8 different activities collected from 12 different subjects. Our results indicate that the system has an accuracy rate of approximately 90 % while meeting our requirements. We are now developing a fully embedded version of our system based on a cell-phone platform augmented with a Bluetooth-connected sensor board. 1

Learning patterns of human behavior from sensor data is extremely important for high-level activity inference. We show how to extract and label a person’s activities and significant places from traces of GPS data. In contrast to existing techniques, our approach simultaneously detects and classifies the significant locations of a person and takes the high-level context into account. Our system uses relational Markov networks to represent the hierarchical activity model that encodes the complex relations among GPS readings, activities and significant places. We apply FFT-based message passing to perform efficient summation over large numbers of nodes in the networks. We present experiments that show significant improvements over existing techniques. 1

Abstract. We study activity recognition using 104 hours of annotated data collected from a person living in an instrumented home. The home contained over 900 sensor inputs, including wired reed switches, current and water flow inputs, object and person motion detectors, and RFID tags. Our aim was to compare different sensor modalities on data that approached “real world ” conditions, where the subject and annotator were unaffiliated with the authors. We found that 10 infra-red motion detectors outperformed the other sensors on many of the activities studied, especially those that were typically performed in the same location. However, several activities, in particular “eating ” and “reading ” were difficult to detect, and we lacked data to study many fine-grained activities. We characterize a number of issues important for designing activity detection systems that may not have been as evident in prior work when data was collected under more controlled conditions. 1

Abstract. Recognition of everyday physical activities is difficult due to the challenges of building informative, yet unobtrusive sensors. The most widely deployed and used mobile computing device today is the mobile phone, which presents an obvious candidate for recognizing activities. This paper explores how coarse-grained GSM data from mobile phones can be used to recognize high-level properties of user mobility, and daily step count. We demonstrate that even without knowledge of observed cell tower locations, we can recognize mobility modes that are useful for several application domains. Our mobility detection system was evaluated with GSM traces from the everyday lives of three data collectors over a period of one month, yielding an overall average accuracy of 85%, and a daily step count number that reasonably approximates the numbers determined by several commercial pedometers. 1

The UW Dynamic Social Network study is an effort to automatically observe and model the creation and evolution of a social network formed through spontaneous face-to-face conversations. We have collected more than 4,400 hours of data that capture the real world interactions between 24 subjects over a period of 9 months. The data was recorded in completely unconstrained and natural conditions, but was collected in a manner that protects the privacy of both study participants and non-participants. Despite the privacy constraints, the data allows for many different types of inference that are in turn useful for studying the prosodic and paralinguistic features of truly spontaneous speech across many subjects and over an extended period of time. This paper describes the new challenges and opportunities presented in such a study, our data collection effort, the problems we encountered, and the resulting corpus. Index Terms — Data acquisition, privacy, speech analysis, oral communication 1.

Given sensors to detect object use, commonsense priors of object usage in activities can reduce the need for labeled data in learning activity models. It is often useful, however, to understand how an object is being used, i.e., the action performed on it. We show how to add personal sensor data (e.g., accelerometers) to obtain this detail, with little labeling and feature selection overhead. By synchronizing the personal sensor data with object-use data, it is possible to use easily specified commonsense models to minimize labeling overhead.

Automated reasoning about human behavior is a central goal of artificial intelligence. In order to engage and intervene in a meaningful way, an intelligent system must be able to understand what humans are doing, their goals and intentions. Furthermore, as social animals, people’s interactions with each other underlie many aspects of their lives: how they learn, how they work, how they play and how they affect the broader community. Understanding people’s interactions and their social networks will play an important role in designing technology and applications that are “socially-aware”. This paper introduces some of the current approaches in activity recognition which use a variety of different sensors to collect data about users ’ activities, and probabilistic models and relational information that are used to transform the raw sensor data into higher-level descriptions of people’s behaviors and interactions. The end result of these methods is a richly structured dataset describing people’s daily patterns of activities and their evolving social networks. The potential applications of such datasets include mapping patterns of information-flow within an organization, predicting the spread of disease within a community, monitoring the health and activity-levels of elderly patients as well as healthy adults, and allowing “smart environments ” to respond proactively to the needs and intentions of their users. 1