TY - JOUR
T1 - A combined pattern separability and two-tiered classification approach for identification of binary mixtures of VOCs
AU - Polikar, Robi
AU - Jahan, Kauser
AU - Healy, Bryan
N1 - Funding Information:
This work is supported by the New Jersey Water Resources Research Institute administered by Cook College at Rutgers University, NJ; and the National Science Foundation under grant No: ECS-0239090. Experimental sections of this work were conducted at Microanalytical Instrumentation Center (MIC) at Iowa State University, Ames IA. Authors greatly acknowledge the assistance of Dr. Ruth Shinar and Dr. Marc Porter at MIC. Robi Polikar received his BSc degree in electrical engineering from Istanbul Technical University in 1993, and his MS and PhD degrees, both in biomedical engineering and electrical engineering, from Iowa State University, in 1995 and 2000, respectively. Since 2001, he has been with the Department of Electrical and Computer Engineering at Rowan University, New Jersey. His current research interests include computational intelligence, signal processing, pattern recognition and data fusion particularly in applications to biological and chemical signals. Kauser Jahan received her BSc degree in civil engineering from Engineering University in Dhaka in 1983, her MS degree in environmental engineering from University of Arkansas in 1987, and her PhD degree from University of Minnesota in 1993. She has been with the Department of Civil and Environmental Engineering at Rowan University, New Jersey since 1996. Her primary experience is in fate and transport of organic pollutants in soil and water.
PY - 2006/7/28
Y1 - 2006/7/28
N2 - Several classification techniques have been developed with varying degrees of success for automated identification of VOCs, however, the problem becomes considerably more challenging when more than one VOC is present. The reason is two-fold: first, the response of the sensors to certain VOCs may be too strong and mask the response of the sensors to other VOCs in the environment; and second the responses of the sensors to VOCs may not have enough separability information if the specificity of the sensors is not adequate. We propose the following procedures for these two issues in identification of binary mixtures of VOCs: a nonlinear cluster transformation technique or nonparametric discriminant analysis to increase pattern separability, followed by a two-tier classification to aid in identification of dominant and secondary VOCs separately. Results demonstrate the feasibility of the combined approach.
AB - Several classification techniques have been developed with varying degrees of success for automated identification of VOCs, however, the problem becomes considerably more challenging when more than one VOC is present. The reason is two-fold: first, the response of the sensors to certain VOCs may be too strong and mask the response of the sensors to other VOCs in the environment; and second the responses of the sensors to VOCs may not have enough separability information if the specificity of the sensors is not adequate. We propose the following procedures for these two issues in identification of binary mixtures of VOCs: a nonlinear cluster transformation technique or nonparametric discriminant analysis to increase pattern separability, followed by a two-tier classification to aid in identification of dominant and secondary VOCs separately. Results demonstrate the feasibility of the combined approach.
UR - http://www.scopus.com/inward/record.url?scp=33646806834&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=33646806834&partnerID=8YFLogxK
U2 - 10.1016/j.snb.2005.11.079
DO - 10.1016/j.snb.2005.11.079
M3 - Article
AN - SCOPUS:33646806834
SN - 0925-4005
VL - 116
SP - 174
EP - 182
JO - Sensors and Actuators, B: Chemical
JF - Sensors and Actuators, B: Chemical
IS - 1-2
ER -