Wearable System Architectures
Wearable Computing is often associated with electronics somehow integrated with clothing or at least worn on the body. While the above is true for many systems, a more general definition refers to system functionality: systems that are usable and useful always, everywhere and under all circumstances. In addition to strict constraints on the form factor and power consumption this implies that the need for unconventional user interfaces and the ability to sense and analyse user activitiy and enviromental conditions. As a consequence a wearable will in general consist of a dnymically configurable set of networked sensors, computing nodes, and IO components distributed to different body locations. The design of such an architecture must consider a variety of conflicting criteria such as miniaturization, power-awareness, global low-power design and suitability for an application. In our research we model, implement and evaluate such architectures. The work is done in cooperation with the wearable computing Lab of ETH Zurich.
Functionality-power-packaging considerations in context aware wearable systems, Nagendra Bhargava Bharatula, Paul Lukowicz, Gerhard Tröster, To appear Personal and Ubiquitous Computing, Springer London, ISSN: 1617-4909
Systematic Approach to the Design of Distributed Wearable Systems U. Anliker, J. Beutel, M. Dyer, R. Enzler, P. Lukowicz, L. Thiele, G. Tr¨oster, A IEEE Transactions on Computers, Vol. 53, No. 8, Aug. 2004, pages:1017 - 1033
Multimodal, Wearable Tasks Tracking
The ability to automatically track the progress of a task is an important aspect of many wearable computing applications. Thus for example in an industrial maintenance scenario it would allow automatic delivery of correct manual parts (e.g. on a head mounted display) as well as automated procedure logging for later verification.
We are currently investigating different approaches to task tracking using wearable sensors. To date we have studied different setups of on body motion sensors combined with sound analysis, ultrasonic hand tracking and user position tracking. In future we will also look at RFID based tools recognition and new sensor modalities such as muscle activity analysis and more accurate hand motion tracking using modulated magnetic fields. From the algorithmic point of view the key challenges are separating the relevant actions from the 'NULL' class in a continuous data stream and the development of appropriate models for the fusion of the different sensor modalities. Furthermore the research needs to move away from simple, controlled lab environments to complex, real life settings.
Activity Recognition of Assembly Tasks Using Body-Worn Microphones and Accelerometers, J.A. Ward and P. Lukowicz and G. Troester and T. Starner, IEEE Trans. Pattern Analysis and Machine Intelligence, 28:10, pp 1553-1567, October 2006
Combining Motion Sensors and Ultrasonic Hands Tracking for Continuous Activity Recognition in a Maintenance Scenario, Thomas Stiefmeier, Georg Ogris, Holger Junker, Paul Lukowicz, Gerhard Tröster, 10th IEEE International Symposium on Wearable Computers (ISWC), Montreux, Switzerland, October 11-14, 2006
Pervasive Healthcare Systems
Beyond pure technology the issue of convincing real life applications is increasingly becoming a central topic in the field of Pervasive Computing. Here the area of health and healthcare has emerged as a promising domain. So called Pervasive Healthcare Systems encompasses a broad range of topics such as advanced hospital information and logistics systems, mobile health monitoring, assisted living for the elderly and the handicapped, and lifestyle and wellness related personal systems. All of the above are both socially and economically highly relevant. The demographic trend towards a more elderly society and the rising healthcare costs lead to a strong demand for solutions that provide adequate care at affordable cost. Furthermore consumers are increasingly health conscious and looking for lifestyle, wellness and health related products.
As promising as it is as an application domain, the field of Pervasive Healthcare is scientifically and technologically highly challenging. Healthcare applications involve complex processes often within strict regulatory constraints. Reliability, security and privacy are central to many applications. In addition since the systems are often closely integrated with the user’s everyday activities and relate to very private issues subtle, human computer interaction issues play an important role in the system design.
Developing and evaluating applications in all areas of Pervasive healthcare is a major research focus of our group. This ranges from systems for hospital personnel (in the WearIt@Work EU project) to assisted living applications (in the MonAmi EU project and in cooperation with the group UMIT in Hall i. Tirol).
Paul Lukowicz, T¨unde Kirstein, Gerhard Tröster, Wearable Systems for Health Care Applications, Methods Inf Med 3/2004
Pervasive computing has begun to transform the previously inanimate physical objects in the environment around us into aware, physical information resources. Similarly, the objects that we wear and carry in our pockets are becoming small computational devices, that are capable of storing information, sensing, and communicating wirelessly.
This projects explores how to utilize those 'objects' to support a diverse range of human acitvities, how to manage complexity moving away from dedicated sensory inputs towards coordination and activity recognition in a more opportunistc fashion. more.
OPF: A Distributed Context-Sensing Framework for Ubiquitous Computing Environments. Max Van Kleek, Kai Kunze, Kurt Partridge, James Bo Begole. In Proceedings of the 1st International Symposium on Ubiquitous Computing, p. 82-97, October 2006. Seoul, Korea.
Distributed Modular Toolbox for Multi-modal Context Recognition, David Bannach, Kai Kunze, Paul Lukowicz, Oliver Amft, Proc. Architecture of Computing Systems - ARCS 2006, pp 99-113, Springer LNCS Volume 3894/2006.
Where am I: Recognizing On-Body Positions of Wearable Sensors. Kai Kunze, Paul Lukowicz, Holger Junker, Gerhard Tröster. International Workshop on Location- and Context-Awarness, May 2005, Oberpfaffenhofen, Germany.
Using Physical Abstractions to Model and Control Self-Organizing Systems
Studying self-organizing systems is a central topic in all parts of computer science. A self organizing system is a dynamic process where the internal organization of the system increases without a manager outside the process. Typically emergent behaviour also occurs. Emergence is a core topic in self-organizing systems. In most general terms it can be described as the development of global complex system properties from simple local interactions between its entities without a centralized control.
Emergence is an appealing concept for many reasons. For one, it simplifies the system description and accomplishes complex system behaviour with simple rules between the parts of the system.
Up to now, most self-organizing systems in special issues of computer science are imitations of concepts found in the nature. One reason is that there are no modeling procedures to model unexpected behavior.
We investigate basic approaches to model and control self-organizing systems and their emergent behavior by using physical abstractions.
Context Recognition Network Toolbox
The Context Recognition Network (CRN) Toolbox allows to quickly build distributed, multi-modal context recognition systems by simply plugging together reusable, parameterizable components. Thus, the toolbox simplifies the steps from prototypes to final implementations that might have to fulfill real-time constraints on low-power mobile devices. Moreover, it facilitates portability between platforms and fosters easy adaptation and extensibility. The toolbox also provides a set of ready-to-use parameterizable algorithms including different filters, feature computations and classifiers, a runtime environment that supports complex synchronous and asynchronous data flows, encapsulation of hardware-specific aspects including sensors and data types (e.g., int vs. float), and the ability to outsource parts of the computation to remote devices.
David Bannach, Kai Kunze, Paul Lukowicz, Oliver Amft: Distributed Modular Toolbox for Multi-Modal Context Recognition, Proceedings of the ARCS (Architecture of Computing Systems) 2006, March/2006, Frankfurt, Germany
Pervasive Systems for Gaming and Sports
Video analysis and motion capturing are standard tools in professional sports to monitor and improve athletic performance by recognizing and fine-tuning the quality of movement. Cutting-edge systems with high-quality sensors hardly suffice to fulfill these professionals’ needs. Quite often, trainers and other experts still process the recorded data by hand. The whole setup and procedure are not only expensive and time-consuming but also error-prone in the sense that the effectiveness of the analysis depends on the humans doing it. Hence, the large-scale use of similar analyses for the hobbyist and gaming masses requires a completely different approach. We envision to add context awareness and ambient intelligence to edutainment and computer sports/gaming applications in general. This requires mixed-reality setups and ever-higher levels of immersive human-computer interaction. We focus on the automatic recognition of natural human hand gestures recorded by inexpensive, wearable motion sensors.people on their bodies (e.g., integrated into their clothes and other personal accessories like watches or jewelry). more ->
Waving Real Hand Gestures Recorded by Wearable Motion Sensors to a Virtual Car and Driver in a Mixed-Reality Parking Game. David Bannach, Oliver Amft, Kai Kunze, Ernst A. Heinz, Gerhard Tröster, Paul Lukowicz:Proceedings of the IEEE Symposium on Computational Intelligence and Games (CIG), April/2007, Honolulu, Hawaii.
Towards Recognizing Tai Chi - An Initial Experiment. Kai Kunze, Michael Barry, Ernst A. Heinz, Paul Lukowicz, Dennis Majoe, Jeurg Gutknecht. The Third International Forum on Applied Wearable Computing, April 2006, Bremen, Germany.
Real-time Recognition Tasks in Games of Martial Arts - An Initial Experiment,Ernst A. Heinz, Kai Kunze, Matthias Gruber, David Bannach, Paul Lukowicz, Proc. IEEE CIG'06, May 2006, Reno/Lake Taho, Nevada
News Sensing Modalities for Context Recogntion
- force sensing resistors
- LC circuit based sensor
A lot of motion based wearable activity recognition systems rely on a combination of 3d-accelerometer and 3d-gyro (and maybe also 3d-magnetic field sensor). While this sensor combination has proven useful for in different applications, it is known to have a number of problems. In particular tracking absolute orientation and subtle motion variants is difficult. As a consequence our group is currently investigating two alternative sensing modalities
- Monitoring of the muscular activity using force sensitive resistors placed on the muscles surface. Such sensors are very thin, power efficient and have also
been demonstrated as pure textile devices, so that they can be
easily integrated in such garments as elastic underwear or tight
shorts/shirt. The method relies on the fact that muscle contractions are accompanied by changes in muscle shape. At the same time very subtle differences between motions are often associated with clearly distinguishable muscle activation patterns. - We investigate how modulated magnetic field generated
by an simple LC oscilator can be used to measure joint angles, which
are a key element in posture recognition and many motion analysis
applications. Our method uses the same physical principle as large
stationary motion trackers, however it applies the principle in a way
suitable for a small, low power wearable system.
Up to date initial prototypes were demonstrated for both above sensors. Currently work is ongoing on second generation hardware and the use of the sensors for more complex activity recognition.
Detecting and Interpreting Muscle Activity with Wearable Force Senors, Paul Lukowicz, Friedrich Hanser, Christoph Szubski, and Wolfgang Schobersberger, Proc. 4th International Conference, PERVASIVE 2006
Wearable Joint-Angle Measurement with Modulated Magnetic Field from L/C Oscillators , Michael Barry, Agnes Grünerbl, Paul Lukowicz, Proc. 4th International Workshop on Wearable and Implantable Body Sensor Networks, BSN 2007, March 2007, Aachen, Germany