Download the pdf version.
"From the Ball Field to the Dance Hall - A Decade of Body Sensor Networks in Medicine, Sports, Entertainment, and the Arts"
Abstract
Wearable systems are often vaguely described as electronics somehow integrated with clothing. We argue that while this is the way many systems are implemented, the definition of a wearable system is much more broad. Wearable systems are better defined through their functionality as systems that are usable always and everywhere. It is only from such functional definition that viable architectures and concepts for the integration of electronics with the users outfit can be developed.
Starting from this thesis we describe a multilayer wearable architecture concept that addresses on of the key questions of on body electronics: "What sort of integration between clothing and electronics makes sense for which application ?".
The architecture is based on a system partitioning that makes sure that different components are embedded in the user's outfit to a different degree, adequate to their functionality, their relation to a particular part of the outfit, and implementation technology.
It views the user's outfit as a complex, hierarchical system that combines different 'device' classes with a wide ranges of application domains and functionalities. For each device class the user has well defined ideas about their expected life cycle, price ranges and the way he needs to treat it. The architectural concepts are illustrated through examples of specific systems and applications.
Biography
Paul Lukowicz has a MSc (Diplom) in Computer, a MSc (Diplom) in Physics and a Ph.D in Computer Science all from the University of Karlsruhe in Germany. After his Ph.D Paul Lukowicz went to ETH Zurich where built up the wearable computing group with a strong focus on on body computer and sensor systems and using them for activity and context recognition.
During this time Paul Lukowicz has lead the development of two generations of Wearable Computers and a number of different wearable sensor platforms.
He then went on to Professorship in Computer Engineering at the University of Medical Informatics and Technology in Hall in Tirol, Austria (UMIT) where his group worked on health related applications of pervasive computing and context recognition.
Since 2006 Paul Lukowicz is Full Professor of Computer Science at the University of Passau in Germany where he heads the Embedded Systems Lab with a focus on pervasive and wearable systems and applications.
In addition to his work in pervasive computing Paul Lukowicz has research interest in opto-electronic computing and self organizing systems.
Abstract
This paper presents an autonomic sensing framework for distributed inferencing, which consists of several self-contained machine learning techniques. A multi-objective Bayesian framework for feature selection is used for learning the relationship of the variables.
To cater for fault tolerance and minimal resource utilisation, feature redundancy and network complexity measures have been introduced. We demonstrate how factor graphs and the sumproduct algorithm can be used for model representation and inferencing.
We will also show how they can be used to facilitate the mapping of model architecture onto the physical sensor networks.
Abstract
Body area networks are becoming more and more popular in addressing health care application due to advances in sensing technologies and the fact the these networks lie within close proximity of the body. We have developed a general purpose wearable platform using lightweight embedded system to address various medical applications. This architecture is composed of tiny processors/microcontrollers equipped with non-invasive sensors. In addition, an on-body terminal enables the system to be reconfigurable and to communicate with medical enterprises. Since our architecture is made of software programmable blocks, it becomes a reconfigurable system. We introduce different levels of reconfiguration for body area networks and illustrate how reconfiguration can address several design challenges such as adaptability, reliability and power consumption. Finally, we formulate sampling rate assignment as a means of power reduction while meeting performance specification. Through our formulation, power dissipation can be minimized and at the same time, the desired accuracy of the system is achieved.