Sensor networks for social networks

Abstract

This thesis outlines the development of software that makes use of Bayesian belief networks and signal processing techniques to make meaningful inferences about real-world phenomena using data obtained from sensor networks. The effectiveness of the software is validated by applying it to the problem of detecting face-to-face social interactions between groups of people, given data readings from sensors that record light, temperature, acceleration, sound, and proximity. This application represents a novel method for social network construction which is potentially more accurate and less intrusive than traditional methods, but also more meaningful than newer methods that analyze digitally mediated communication.