Chemical Source Localization Using Electronic Nose Robotics

Data Processing

Experimental Overview
Experimental Setup
Data Acquisition
Data Processing
Discussion and Future
About Us

As mentioned in Data Acquisition, when each of the 3 sensor pairs detected the isopropyl alcohol source, they each outputted a pair of voltage signals, resulting in a 6-component composite signal.  See figures  below for examples of these signals:


Principles Behind Processing

In order to extract information about the actual location of the isopropyl alcohol source from these signal pairs, a variety of data processing methods were considered and attempted, based on the shark model of sensing.  Recall that, according to "The Function of Bilateral Odor Arrival Time Differences in Olfactory Orientation of Sharks," sharks localize odor sources by distinguishing which nostril smells the odor first.  In this project, each signal pair represented a smell detected by a pair of electrical nostrils.  The goal of processing was to determine a method of distinguishing which individual voltage signal in a pair corresponded to the sensor that sensed the isopropyl alcohol source first.

At first, processing was performed based on the hypothesis that in a given pair of sensors, the sensor closer to the isopropyl alcohol source produced a bumpier signal.  This hypothesis was based on the reasoning that the alcoholic fume currents would be more concentrated and unpredictable closer to the source, but would dissipate and waft over the sensors more smoothly as they travelled farther than the source.

However, eventually this approach was abandoned because it yielded unpredictable and often incorrect results.  Ultimately, the signals were processed based on the principle that alcoholic air current would reach the sensors closest to the source first.  In other words, we assumed that, for a given pair of sensors, the sensor closer to the source would produce peaks and troughs in voltage sooner than the sensor that is farther away.