To deal with the data from a sensor array it is necessary to use different methods,
which allow fitting the multidimensional output of the sensor set.
Depending on the task, some of the methods shown in the table can be used.
In some cases several methods are equally applicable, in the other cases the
certain method gives the best results.
|
DATA PROCESSING TECHNIQUES
|
Pattern recognition methods
Principal component analysis
Cluster analysis
Fuzzy logic
Artificial neural networks (back-propagation, SOM)
|
Multivariate calibration methods
Multi linear regression
Principal component regression
Partial least square regression
Non-linear regression
Artificial neural networks (back-propagation)
|
An important principle we strictly follow - adequate data processing is the essential part
of the electronic tongue approach but even advanced math methods cannot improve the results.
The data produced by the sensor array must be reproducible and reliable. This is ensured by
the responsible design of the sensor array and thoroughly elaborated measuring procedure.
