Abstract

Propelled by a rapidly expanding market in personal devices, sensing and processing elements have become small and efficient enough to enable new wearable technologies that can detect when the wearer performs particular activities, including how well, and how frequently they occur. This is on one hand of interest for machine learning researchers as these type of systems are challenging since they need to be fast, adaptive, small-scale, robust, and power-efficient, to be deployed for long stretches of time in the real-world. For researchers in psychology and biology on the other hand these systems could be used as instruments to obtain information that might provide new insights in the behaviors of living subjects in their natural environments. This talk will focus on the current state of the art in wearable activity recognition research and will show, using several own projects at TU Darmstadt, what is currently feasible, where the promises and bottlenecks in this field lie, as well as what achievements can be expected in the next years.

About Kristof Van Laerhoven

Kristof Van Laerhoven heads the Embedded Sensing Systems group at the Technical University of Darmstadt (Germany) and combined this till recently with a guest professorship at the University of Passau (Germany). He obtained his PhD at the University of Lancaster (UK), and his M.Sc. at the University of Brussels (Belgium). Kristof is active in wearable and ubiquitous computing research, as well as wireless sensor networks research, and has specialized particularly in sensor-driven systems that are able to detect complex events through the use of pattern recognition and machine learning algorithms. He is in 2013 active as general chair for the 17th international symposium of wearable computers (ISWC) in Zurich, Switzerland, and program co-chair of the 12th international conference on mobile and ubiquitous multimedia in Luleå, Sweden.