Odour-sensing system using a quartz-resonator sensor array and neural-network pattern recognition

Abstract

It is difficult to realize an odour or gas sensor with a high selectivity. From a biomimetic viewpoint, it is promising to make a sensor array and analyse the output pattern to recognize the various sorts of gases. We use six quartz resonators, with different coating materials, whose oscillation frequencies decrease when gas molecules are adsorbed on the sensing membranes over them. The pattern analysis method used in the present study is neural network pattern recognition. This network has been trained to identify the types of odours using the back-propagation algorithm. The system is trained to identify 11 kinds of liquors on the market and its recognition probability is 73% when the liquor signals used in the training are input. In order to enhance the odour recognition ability, the data vectors for the liquors are input to the network after subtracting those for aqueous ethanol solutions that have the same ethanol concentrations as the liquors. The recognition probability is then improved to 88%.