DAV 3E – a MATLAB toolbox for multivariate sensor data evaluation

Bastuck, Manuel; Baur, Tobias; Schütze, Andreas

We present DAVinline-formula3E, a MATLAB toolbox for feature extraction from, and evaluation of, cyclic sensor data. These kind of data arise from many real-world applications like gas sensors in temperature cycled operation or condition monitoring of hydraulic machines. DAVinline-formula3E enables interactive shape-describing feature extraction from such datasets, which is lacking in current machine learning tools, with subsequent methods to build validated statistical models for the prediction of unknown data. It also provides more sophisticated methods like model hierarchies, exhaustive parameter search, and automatic data fusion, which can all be accessed in the same graphical user interface for a streamlined and efficient workflow, or via command line for more advanced users. New features and visualization methods can be added with minimal MATLAB knowledge through the plug-in system. We describe ideas and concepts implemented in the software, as well as the currently existing modules, and demonstrate its capabilities for one synthetic and two real datasets. An executable version of DAVinline-formula3E can be found at urihttp://www.lmt.uni-saarland.de/dave (last access: 14 September 2018). The source code is available on request.



Bastuck, Manuel / Baur, Tobias / Schütze, Andreas: DAV3E – a MATLAB toolbox for multivariate sensor data evaluation. 2018. Copernicus Publications.


12 Monate:

Grafik öffnen


Rechteinhaber: Manuel Bastuck et al.

Nutzung und Vervielfältigung: