|
Advanced Data Anaylsis Methods for Olfactory Classification using
Optical Sensor Arrays
An artificial dog brain
For processing output from Tufts'
artificial dog nose
Download a 1-page summary in postscript format
On BEADS...
NEW! S. Stitzel, L. Cowen, K. Albert, D. Walt,
Array-to-Array Transfer of An Artificial Nose Classifier (Preprint).
Download separately: Figure 1,
Figure 2,
Figure 3,
Table 1,
Tables 2 and 3.
Sponsored by DARPA
DARPA Applied and Computational Mathematics
AFOSR
Executive Summary
A Tufts University Chemistry/Neuroscience collaboration has resulted in a
cross-reactive non-specific fiber-optic sensor system which mimics
the design employed in the natural olfactory system. This "artifical
nose" is capable of highly sensitive and broad-band odor
recognition.
However, based soley on the raw sensor output (multi-band multi-wavelength
time series resposes to vapor) it is difficult to determine
quantitatively the detection and identification capabilities of the
sensor.
Our task is to develop a statistical pattern analysis methodology
(mathematics/algorithms/software) which takes as input the
sensor response and acta as the olfactory processing
unit.
Given a database of training observations (sensor
responses to known analytes) statistical pattern analysis
partitions observation space into detection/identification
decision regions.
Goals:
Quantify the "highly sensitive and broad-band" claim
Positively impact further sensor design
Use the sensor in conjunction with the processing unit and the
training database to perform detection and identification in applications.
Technical Contact
Carey E. Priebe
cep@jhu.edu
Department of Mathematical Sciences
The Johns Hopkins University
Baltimore, MD 21218-2682
New Publications
J. Xie and C.E. Priebe,
A Weighted Generalization of the Mann-Whitney-Wilcoxon Statistic.
Technical Report No. 593, Department of Mathematical Sciences, Johns Hopkins University.
J. Xie and C.E. Priebe,
Generalizing the Mann-Whitney-Wilcoxon Statistic.
Technical Report No. 579, Department of Mathematical Sciences, Johns Hopkins University.
C.E. Priebe,
Olfactory Classification.
Technical Report No. 585, Department of Mathematical Sciences, Johns Hopkins University.
C.E. Priebe and L.J. Cowen,
A Generalized Wilcoxon-Mann-Whitney Statistic with Utility for Interpoint Distance Analysis.
Technical Report No. 586, Department of Mathematical Sciences, Johns Hopkins University.
A. Cannon and L.J. Cowen,
Approximation Algorithms for the Class Cover Problem.
M.C. Minnotte, R.W. West, J. L. Solka, "The Data Image for High
Dimensional Exploratory Data Analysis", submitted to the Journal of
Computational and Graphical Statistics.
Previous Publications
L.J. Cowen and C.E. Priebe,
Randomized Non-Linear Projections Uncover High-Dimensional Structure.
Advances in Applied Math 19:319-331, 1997.
C.E. Priebe and L.J. Cowen,
Approximate Distance Clustering.
Computing Science and Statistics 29:337-346, 1997.
C.E. Priebe and L.J. Cowen,
Mine Detection Via Generalized Wilcoxon-Mann-Whitney Classification.
Proceedings of the SPIE 3392: 906-917, 1998.
A. Cannon, L.J. Cowen and C.E. Priebe,
Approximate Distance Classification.
Computing Science and Statistics 30, 1998.
M. C. Minnotte and R. W. West, "The Data Image a Tool for Exploring High
Dimensional Data Sets," 1998 Proceedings of the ASA Section on Statistical
Graphics, in press.
The program manager for the project is:
Dr. Dennis Healy
|
