A novel method for reducing the dimensionality in a sensor array

doi:10.1109/19.744338

Return

A novel method for reducing the dimensionality in a sensor array

1998-06-01

PRE

Bahram G. Kermani

Susan S. Schiffman

H. Troy Nagle

DOI：10.1109/19.744338

Original

Original request for help

Save

Abstract

En 中文

Specific types of gas sensors are normally produced by adding different dopants to a common substrate. The advancement of technology has made the fabrication of many dopants and consequently various sensors possible. As a result, in each family of gas sensors, one can find tens of different sensors which are only slightly different in the spectrum of response to various volatile compounds. The wide variety of available gas sensors creates a selection problem for any specific application. Sensor selection/reduction becomes even more important when cost and technology limitations are issues of concern. Accordingly, a methodology by which one can tailor a sensor array to a specific need is highly desirable. In this paper, a novel method is introduced to address this task using data from an electronic nose that uses polymer gas sensors, This method has been delineated based on the geometry of eigenvectors in Karhunen-Loeve expansion. The methodology is general and therefore suitable for many other feature selection problems.

Keywords:

aroma

artificial neural networks

data compression

electronic nose

Karhunen-Loeve

odor recognition

sensor array

smell

AI Summary

Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.