An intelligent system for odour discrimination

The paper describes the design of an "electronic nose" - an intelligent system for discrimination of odours. The system is designed using tin-oxide semiconductor gas sensors. The sensor array produces a response vector corresponding to the type of the volatile organic vapours to which it has been exposed. The response is acquired through a 16-bit data acquisition board and displayed on the PC monitor using the virtual instruments of the LabView software. The LabView software organises the data in a suitable format into a file, which is exported to another software (EinSight front Infometrix) for pattern recognition. This software using Principal Component Analysis (PCA) technique helps to reduce the dimensionality of the sensory data and classifies it in different classes in a 2-D/3-D plot. The system has been trained to discriminate different types of coffee, aroma oils, perfumes and alcohol. A program using Visual basic has been written to (1) replace the functions of the LabView Software and (2) set tip a database for storing the test patterns which will be used to train the neural network being used for identification of the samples