Electronic and computational olfaction

Biological gas detection and classification systems achieve a level of molecular sensitivity and robustness as yet unmatched by electronic devices. Development efforts continue to narrow the gap in functionality between electronic and biological olfaction, driven by the demonstrated utility of devices which provide objective, quantitative and reproducible odor measurements. We developed an artificial olfactory system to test whether incorporating biological design features into artificial olfactory systems would yield significant gains in functionality. Our initial system used an array of conducting polymer sensors with band pass filtering and amplification of sensor signals to optimize signal to noise in the relevant frequency range. Several pattern recognition algorithms were applied to the sensor signals. We used different cultivars of common produce items as test odor objects. Our device was able to differentiate between different cultivars of apples and oranges. We are now exploring the use of novel materials for sensor fabrication to allow implementation of new algorithms for odor identification which depend on inputs from 50-100 different classes of odor sensors having a range of binding affinities of at least three orders of magnitude.