An air-quality sensor system with fuzzy classification

Abstract

A sensor system for the measurement of air quality is presented and evaluated. It is based on the approach of mimicking the advantages of human olfactory sensing, using gas sensor measurements and fuzzy sensor fusion. Signals from four sensitive metal - oxide - semiconductor field-effect transistors (MOSFET), each with a unique selectivity profile, are used. Each sensor exhibits information which is extracted by signal analysis and used to determine a crisp air-quality profile. The resulting crisp profile is unique, representing a distinct air quality and serves as a basis for the determining of a fuzzy quality profile. A fuzzy measurement profile is then matched pair-wise to the different fuzzy quality profiles. The matches are processed with help of a fuzzy fusion technique to determine the most representative crisp air-quality profile. Preliminary experiments show that the performance of the proposed fuzzy sensor clustering system is very promising. Environmental air-quality tests inside and outside a car driving in different cities during different seasons resulted in an artificial system determining air-quality opinions. The sensor opinions are classified with reference to three air-quality references (clean, medium and polluted air) and appear to be in accordance with human opinions obtained simultaneously.