Response Standardization for Drift Correction and Multivariate Calibration Transfer in "Electronic Tongue" Studies.

Abstract

The procedures for response standardization in "electronic tongue" (ET) studies are described. The construction of reliable multivariate calibration for "electronic tongue" requires the analysis of a large number of representative samples both with ET and reference techniques. This is a laborious and expensive process. Long-term sensor array operation leads to the changes in sensor response characteristics and thus invalidates the multivariate predictive models. Moreover, due to the individual parameters of each sensor in different sensor arrays, it is not possible to use the calibration model for one system together with the data acquired by another system, even if they have the same sensors. Both of these issues lead to the necessity of frequent sensor array calibration which would be ideal to avoid. Instead of recalibration, these two problems can be handled using mathematical methods intended for sensor response standardization. This chapter describes two popular methods of standardization which can be used for both drift correction and calibration transfer. Thus, significant efforts on measuring representative sample sets for sensor array recalibration can be avoided.