A neural network model for an electronic nose based on quartz-crystal microbalance sensors

An electronic nose with a radial basis function network that can identify the qualities of vapors has been developed. This electronic nose uses an array of quartz microbalance sensors, each coated with a different organic film. To identify the qualities of vapors, pattern recognition of the sensor responses is necessary because such a sensor does not usually response to a particular type of vapor. The sensor-array responses are normalised to isolate the intensity and quality of the vapor for feature extraction. These normalised responses are sequentially matched with the templates, the centres of the radial basis functions, by using time-delayed connections. The radial basis function network needs less training data than other pattern recognition algorithms. Experiments show that it has better recognition performance than a classical linear classifier