New data reduction algorithm for voltammetric signals of electronic tongue for discrimination of liquids

This paper deals with a data reduction technique for pulse voltammetric signals of electronic tongue for discrimination of different types of liquids. The electronic tongue is a system which can determine the fingerprint of a liquid sample by an array of non-specific/semi-specific electrodes/sensors coupled to a pattern recognition system. The data acquired from electronic tongue is very large in size, in particular, when multiple electrodes are used. A new method or technique for reducing the size of the variables in a voltammetric signal is proposed. The method is based on segmentation of the voltammetric signal into significant and insignificant segments based on the statistical measure, variance, in the samples under investigation for classifying into different classes. The method is simple, fast and good enough to reduce the size of the signal without much loss of significant information needed for classifying the samples. The proposed method is described in the present paper with its application to two different sets of data obtained from the electronic tongue where in data reduction is achieved to a greater extent. Further, principal component analysis (PCA) was applied on the data before and after reduction. The results obtained validate the technique proposed in the present paper as a relatively good option for reducing the dimension or size of the data matrix for voltammetric signals of electronic tongue.

► Voltammetric data of electronic tongue are very large in size.
► A new data reduction method for voltammetric signal is proposed.
► It is based on the segmentation of the voltammetric signal using statistical measures.
► The data reduction method is applied on basic taste solutions and juices signals.
► The method has achieved ∼80% of data reduction on juices and taste samples.