Study of the temperature and humidity dependence of a metal oxide semiconductor sensor array

This paper characterizes the environmental factors based on a sensor array which is composed of four MOS sensors and a temperature-humidity sensor module. A mathematical model of humidity and ambient temperature effects (drift) on gas sensors is first proposed and then used for the on-line drift compensation in an electronic nose (E-nose) to identify indoor/in-car air pollutants. Principle component analysis (PCA) and probabilistic neural network (PNN) methods are used to complete pattern recognition of pollution gases, benzene, ammonia and nitrogen dioxide. After drift correction, these three kinds of pollution gases can be completely discriminated.