Gas identification in electronic nose by using similarity measure between latency patterns

Recently, implementation friendly bio-inspired coding schemes have been developed for an electronic nose system to recognise different gases. In these schemes, a logarithmic timedomain encoding technique is used to covert the response vector of the sensor array in an electronic nose into a latency pattern. These schemes assume a unique temporal sequence of latencies, referred to as a rank order, for each target gas. However, poor repeatability and sensor drift limit the performance of these schemes. In this paper, we use angular separation between the latency patterns of the sensor array for gas identification. An electronic nose system containing an array of commercially available gas sensors and a radio frequency module is developed and characterized in the laboratory environment with four gases. Experimental data is used to compare the performance of our coding scheme with existing bio-inspired coding schemes and commonly used pattern recognition algorithms.