The UK Department of Trade & Industry has been the catalyst for a technology agreement between two companies that have developed a computer with a sense of smell. Dubbed the Neural Nose Neotronics Olfactory Sensing Equipment – the computer consists of a sensor developed by Neotronics Technology Plc, Bishops Stortford, Hertfordshire, hooked up to a neural network. The project has sprung out of a three-year, UKP5.25m Department of Trade & Industry initiative to encourage British industry to use neural network processing techniques (CI No 2,123). The sensor has been developed over four years at a cost of around UKP750,000 by Neotronics, and it includes technology from the University of Warwick, that long-time pioneer of the electronic olfactory of the future.
Sugar in its beer
It consists of 12 vapour-sensing polymers that react electro-chemically to particles in the air. The reaction in the polymer creates a conductor that links two electrodes on either side of the polymer. Different particles cause different electro-chemical reactions, and thus different resistances. The different resistances are measured against a standard resistance, and the nature of the vapour thus determined. The vapour can be displayed as a difference plot or solar plot to highlight variances from the standard pattern. When hooked up to the neural network supplied by Neural Technologies, the variances in resistance can be analysed by the neural network to provide identification against a range of different smells. A customer such as a brewery could thus arrive at an application which, for example, told it that it was putting too much yeast or sugar in its beer, or that the beer was simply off. The Neural Technologies network uses a standard 80486 personal computer which houses a board containing the company’s Neural Instruction Set Processor. The chip houses 8,000 neurons and 65,000 connections, and can make 40m interactions between those neurons every second. Technical director Nick Ryman-Tubb estimates that a normal 80486 computer could make a maximum of 1m connections per second. Using the computer and the board, he says that a typical network could be trained in as little as half an hour. The NOSE, which should ship by July, will initially sell into laboratory control applications, but Neotronics is anticipating a production line version for process control within a year, and a portable version within two years. The applications are limitless according to Neotronics, who says that it could be used for everything from detecting fake perfume through to monitoring industrial waste products and air quality. The NOSE is even said to be able to identify the year and the vineyard of your average bottle of plonk. Neotronics, which went public in 1987, turned over UKP21.4m last year, and spends 12.5% on research and development. Now there’s a figure not to be sneezed at.
