Bioelectronic artificial nose using four-channel moth antenna biopotential recordings

The use of insect antennae as an odor sensor array was evaluated as a means to advance the current capabilities of "artificial nose" technology. A given species is highly sensitive to odors of survival interest (e.g. species-specific pheromones), but also to a broad range of other natural and anthropogenic compounds. The sensitivity of the antennae to some odors extends to the parts per billion range [1]. In contrast, the best current artificial nose technology is able to detect compounds in the parts per million range. Here, a system designed to utilize four antenna biopotential signals suitable for field use and a computational analysis strategy which allows discrimination between specific odors, and between odor and background or unknown compounds, with high fidelity and in real time, is described. The automated analysis measures three parameters per odor response. Following a training period, a K nearest-neighbor (KNN) approach is used to classify an unknown odor, assuming equal prior probabilities. The algorithm can also declare an odor as "unknown". System responses to single filaments in an odor plume can be analyzed and classified in less than one second.