Feature analysis has been increasingly considered as
an important
way to enhance the discrimination performance of gas sensors. In this
work, a new analytical method based on alternating current noise spectrum
is developed to discriminate chemically and structurally similar gases
with remarkable performance. Compared with the conventional analytics
based on the maximum, integral, and time of response, the noise spectrum
of electrical response introduces a new informative feature to discriminate
chemical gases. In experiment, three chemically and structurally similar
gases, mesitylene, toluene, and o-xylene, are tested
on ZnO thin film gas sensors. The result indicated that the noise
analytics assisted by the support vector machine algorithm discriminated
these similar gases with 94.2% in precision, about 20% higher than
those obtained by conventional methods. Such a new alternating current
noise analytics is very promising for application in sensors for high
discrimination precision.