Advanced Data Analysis Methods for Analyte Recognition from Optical Sensor Arrays

摘要

The objectives of this project were, simply put, to develop advanced statistical pattern recognition methodologies for the Tufts University artificial nose (and other sensors of interest -- notably, hyperspectral imagers). This effort had significant positive impact on the Tufts University artificial nose. In particular, I claim that the paper: C.E. Priebe, 'Olfactory Classification via Interpoint Distance Analysis', IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. No. 4, pp. 404-413, April 2001, is among the most important papers ever published on statistical pattern recognition for artificial olfactory sensor systems. Additional publications detail advancements made which positively impact the Tufts nose (and are applicable to many other sensor systems). Furthermore, this effort produced workable initial versions of a methodology for jointly optimizing classification with sensing and processing, in terms of adaptive dimensionality reduction. This latter concept is relevant to a wide variety of adaptive sensors, and will be pursued in the future.