Portable Electronic Nose System for Aroma Classification of Black Tea

Abstract

A portable electronic nose system has been developed with an array of five commercially available Metal Oxide Semiconductor (MOS) sensors, where a microcontroller (muc) is used for the pattern recognition. The classification of black tea aroma is carried out in the muc (PIC18F4520) and is based on feed forward multilayer perceptron (FF-MLP) algorithm. With the samples collected from the different gardens of north-east and eastern India, the MLP is trained first using the back-propagation algorithm with the fingerprint from the sensor array and the corresponding tea tasters' mark in a PC to obtain the optimum architecture and weights and biases of the neurons. Once it is trained, the computed weights and biases of the neurons are programmed in the muc and it then becomes a portable instrument, which gives the aroma index directly for new unknown tea samples. It is observed from the results that the performance of the muc-based electronic nose is at par with that of the PC-based electronic nose system when compared with unknown finished black tea samples.