Embedded pattern recognition methodsPowerful microcontrollers have become more available and
affordable, enabling
the implementation of pattern recognition algorithms in
embedded systems. These
systems consist of electronic components such as sensors,
actors, data
transmission devices and microcontrollers, which demand
restrictions on power
consumption, device dimensions and processing time. Parts of
the pattern
recognition pipeline which are implemented on
microcontrollers include the
preprocessing of sensor data, the feature extraction and the
use of preselected
classification algorithms.
This research project addresses the development of pattern
recognition methods
and algorithms on embedded systems that take into account
the mentioned
requirements on the target platforms. The development
process of these kinds of
systems will be supported in applied research projects. An
analysis of the
processing power and storage capacity needed for pattern
recognition algorithms
to run on a target hardware platform is part of our research
goals. Closely
related is the case where the hardware resources are known
(e.g. memory,
processing power, arithmetic operations) and an adequate
classification
algorithm has to be selected which delivers the best
possible classification
results for the given hardware restrictions. Since embedded
systems are often
battery powered devices, the power consumption of the entire
system is crucial
point in the design process. Therefore, the efficient energy
usage in our
methods is essential for many applications. Example
applications of pattern
recognition in embedded systems can be found in the
automotive, sports or
medical engineering sectors.
| Project manager: Prof. Dr. Björn Eskofier
Project participants: Dr.-Ing. Ulf Jensen, Dipl.-Ing. Gabriel Gomez
Keywords: Pattern recognition; embedded systems; classification
Duration: 1.1.2012 - 31.12.2012
Contact: Jensen, Ulf E-Mail: ulf.jensen@cs.fau.de
| Publications |
---|
Ring, Matthias ; Jensen, Ulf ; Kugler, Patrick ; Eskofier, Björn: Software-based Performance and Complexity Analysis for the Design of Embedded Classification Systems. In: Institute of Electrical and Electronics Engineers (IEEE) (Ed.) : Pattern Recognition (ICPR), 2012 21st International Conference on (21st International Conference on Pattern Recognition Tsukuba, Japan November 11-15, 2012). 2012, pp 2266-2269. - ISBN 978-4-9906441-1-6 |
Institution: Chair of Computer Science 5 (Pattern Recognition)
|
|