Discrimination of monofloral honeys using a potentiometric electronic tongue
Conference Paper
Overview
Overview
abstract
There is a commercial interest in selling honey according to its color and pollen
classification because the pollen content is related to the honey quality [1]. The honey
color classification is expressed in the Pfund scale (mm), based on absorbance readings at
560 nm, and ranges between extra-white and dark [2].
The objective of this work is to investigate whether a potentiometric electronic tongue (Etongue)
is able to classify honeys according to their three main colors (white, amber and
dark) and to discriminate between monofloral honeys within each of these groups.
The E-tongue had an Ag/ Agel reference electrode with double junction, and a multi-sensor
device with 20 chemical sensors, based on all-solid-state electrodes with lipid polymeric
membranes formed on solid supports of conducting silver. Two identical systems of multisensors
were used, which allowed recording a profile of 40 signals for each sample.
Honeys samples (65), provided by the National Federation of Beekeepers of Portugal,
were classified according to the color (spectrophotometric method), pollen profile
composition (microscopic method) and also analyzed using the E-tongue.
The honeys of each group contained the following monofloral honeys: white honeys (20
samples), Lavandu/a sp. or Echium sp. ; amber honeys (30 samples), Lavandu/a sp., Echium
sp., Prunus sp. or Rubus sp.; dark honeys (15 samples), Castanea sp., Erica sp. or Rubus sp.
The data were treated by linear discrimination analysis (LDA) using forward stepwise
variable selection and the leave-one-out cross-validation technique.
Selected E-tongue signals together with LDA allowed 100%, 90% and 80% of correct
classifications of monofloral honeys within the amber, white and dark groups,
respectively.
The results showed that the E-tongue can be used as a practical tool for discriminating
monofloral honeys, though more robust classifications are expected by using different
heuristic techniques for variable selection.
